Datasets:

ccdv
/

arxiv-summarization

Dataset card Data Studio Files Files and versions Community

Subset (2)

section · 216k rows

Split (3)

train · 203k rows

article stringlengths 0 745k	abstract stringlengths 0 92.7k
the goldstone theorem asserts that in a theory with spontaneous breaking of a global symmetry group some massless scalar bosons should exist - the so - called nambu - goldstone bosons ( ng ) - one per global symmetry broken by the vacuum . good candidates for spontaneously broken quantum numbers are the partial lepton numbers @xmath3 and total lepton number @xmath4 . indeed , they are conserved by the standard model ( sm ) interactions , but several experiments going now from solar to atmospheric neutrinos and laboratory oscillation experiments @xcite indicate that the lepton flavors are not conserved . the ng bosons not only have zero mass but also no scalar potential terms such as @xmath5 . futhermore , they also interact with the fermion particles however , they do not mediate long range forces . the reason lies in the very global symmetry that , broken by the vacuum still operates at the lagrangian level . a ng boson transforms as @xmath6 @xmath7 , and therefore the symmetry requires that these fields have nothing but derivative couplings ( this is related to the soft pion low energy theorems @xcite ) _ i.e. _ , all of the ng boson interactions can be expressed with an effective lagrangian of the form @xmath8 the @xmath9 equation of motion is nothing but the conservation law of the total current associated with the @xmath10 symmetry , @xmath11 where @xmath12 is essentially the scale of symmetry breaking and the current @xmath13 is determined in leading order by the quantum numbers of the particles . for instance , if @xmath14 is the electron lepton number @xmath15 , @xmath16 where the dots stand for model dependent radiative corrections and scalar boson contributions . this general result can be obtained using a suitable rephasing of the fields present in the theory @xcite . let @xmath17 and @xmath18 be the quantum numbers of the fermion fields @xmath19 and scalars @xmath20 under an abelian global symmetry@xmath21 which is spontaneously broken by vacuum expectation values @xmath22 . a suitable change of variables namely , @xmath23 makes the lagrangian to be expressed in terms of _ physical weak _ eigenstates , fermions @xmath24 , massive bosons @xmath25 and gauge bosons ( assumed to be singlets of the global symmetry ) all of them invariant under u(1)@xmath26 . the only non - invariant field is the ng boson @xmath9 . it transforms as @xmath27 hence , the lagrangian , symmetric under u(1)@xmath26 , can only depend on @xmath9 through the derivatives @xmath28 . its leading order couplings are derived from the kinetic lagrangians of @xmath19 and @xmath20 by application of eqs . ( [ tpsi]-[tsigma ] ) . one obtains in particular , @xmath29 . the second member of the ng equation of motion , eq . ( [ ddfi ] ) , is clearly not an ordinary scalar density . if one calculates the divergency of a vector or axial - vector current using the dirac equation , one ends up with either scalar but flavor violating terms , @xmath30 or flavor conserving but pseudo - scalar , @xmath31 this kind of source terms do not generate long range @xmath32 fields , but rather spin dependent @xmath33 potentials @xcite . in the intent of escaping the derivative couplings law , models were made with explicit symmetry breaking terms . these might be anomaly terms @xcite , as in the axion case @xcite , or soft symmetry breaking terms @xcite . in this way scalar couplings with the ng field are produced but at the cost of giving a mass to the then , so - called pseudo nambu - goldstone boson . the mass ultimately means a finite range for the ng bosons . i will show next that under certain conditions a true nambu - goldstone boson may develop a macroscopic field @xcite . first notice that the source of a ng field is different from zero if the matter current @xmath13 is not conserved . the volume integral @xmath34 represents the rate of @xmath14-number _ created _ per unity of time ( in all particles except @xmath9 ) and @xmath35 is its volume density . the equation of motion ( [ ddfi ] ) gives @xmath36 it is a long - range coulomb type of field but suppressed by the product of the scale @xmath37 times the interval of time per @xmath14-violating process . to have a significant macroscopic field one needs a large rate of reactions with a net @xmath14 number variation . that calls for astrophysical sources . there is evidence that solar neutrinos change flavor as they come out of the sun . if @xmath1 oscillates into @xmath38 , for instance , then @xmath39 and @xmath40 are not conserved in a large scale in stars and possibly in supernovae as well . if any of them is associated with a spontaneously broken global symmetry then , the respective ng boson acquires classic field configurations . for definiteness suppose that @xmath41 is spontaneously broken and that a fraction of the @xmath1s emitted from a star are resonantly converted into @xmath42 through the mikheyev - smirnov - wolfenstein ( msw ) mechanism @xcite ( also @xmath42 into @xmath1 in a supernova ) in a certain shell of radius @xmath43 inside the star . assuming for simplicity stationary neutrino fluxes and spherical symmetry one obtains the solution for the ng boson @xmath44 : @xmath45 \ , \quad r > r_{{\rm c}}\end{aligned}\ ] ] in terms of the total numbers of converted neutrinos , @xmath46 and @xmath47 , per unity of time . what are the possible consequences of such a ng field ? applying eqs . ( [ tpsi ] ) to the kinetic terms @xmath48 one obtains the interaction @xmath49 the fermions only care about the gradient of the ng field . in the situation described above , denoting by @xmath50 the flux of @xmath51-neutrinos converted to @xmath52 and @xmath53 the reciprocal , @xmath54 \ . \label{ae}\ ] ] the quantity @xmath55 is suppressed relatively to the field @xmath44 by the scale @xmath12 and radius @xmath56 and that is what makes it extremely small . however , as follows from eq . ( [ lint ] ) , it still contributes to the potential energy of an electron - neutrino with velocity @xmath57 as @xmath58 these contributions associated with breaking of partial lepton numbers are flavor dependent by nature and therefore may have a role in neutrino oscillations . one estimates the order of magnitude of theses ng potentials as the neutrino flux divided by @xmath59 . in the case of the sun , evaluating the total flux at the sun radius @xmath60 , @xmath61 it produces a too small variation of the neutrino phase , @xmath62 , thus it does not seem significant for solar neutrinos . in a supernova however the fluxes are many orders of magnitude larger . one has @xmath63 where @xmath64 is the fermi constant , @xmath65 , @xmath66 is the neutrino energy luminosity , and @xmath67 its average energy . @xmath68 gives the neutrino emission rate . this has to be compared with the standard model electro - weak potential @xmath69 @xcite . in the regions of a supernova star with densities typical of the sun the mass density goes as @xmath33 , the constant @xmath70 lying between @xmath71 and @xmath72 depending on the star @xcite then , @xmath73 @xmath74 being the electron abundance and @xmath75 the number density . clearly , in the period of time where the neutrino luminosities are high enough say , @xmath76 , the scale of lepton symmetry breaking can be as high as the fermi scale and yet the ng potentials compete with @xmath77 at the places where the resonance occurs for values of @xmath78 interesting for solar neutrino solutions ( @xmath79 ) . futhermore , the ng potentials decay as @xmath80 and necessarily overcome the local interactions at large enough radius . at this point we can already draw some conclusions . first , if the lepton numbers @xmath81 are spontaneously broken , the neutrino oscillations in a supernova originate the appearance of classic nambu - goldstone fields . second , if the scale of symmetry breaking lies below 1 tev they in turn yield potential energies that are significant for the very neutrino oscillation processes . what are the possible consequences ? the ng coupling are flavor dependent by nature hence , may alter the pattern of neutrino oscillations , with a fundamental difference : the ng potentials are proportional to the neutrino fluxes and that induces a time dependence of the effects in a supernova context . next i show two examples . one @xcite is the improvement of adiabaticity of otherwise @xmath82 non - adiabatic transitions . the other @xcite is the violation of the sm prediction that the anti - neutrinos do not resonantly oscillate if the neutrino masses obey the hierarchy of the charged lepton masses . suppose that in the sun the electron - neutrino oscillates into the muon - neutrino with the parameters of the non - adiabatic , small mixing angle solution @xcite . then , the @xmath83 transitions are also non - adiabatic in a supernova and only a fraction of each neutrino species is converted into the other . furthermore , since the level crossing probability is an increasing function of the energy , the hotter @xmath42s have larger survival probabilities than the cooler @xmath1s . let us assume that the partial lepton number @xmath15 is conserved at the lagrangian level but the respective global symmetry , u(1)@xmath84 , is spontaneously broken . then , a ng boson ( majoron ) , @xmath44 , exists with zero mass . although in principle , the neutrino mass matrix violates all three lepton numbers , i will ignore for simplicity other possibly existing majorons . notice that other ng fields are naturally less intense if the respective scales of symmetry breaking are slightly higher . the equations governing two flavor oscillation @xcite with mixing angle @xmath85 and @xmath86 difference in the vacuum , are given by , up to an universal phase absorbed in the wave functions , @xmath87 @xmath88 is the difference between the @xmath89 and @xmath52 potentials , which is the following in the presence of a @xmath44 field ( see eqs . ( [ vnue],[vw ] ) : @xmath90 assuming spherical symmetry , @xmath55 only has radial component and one obtains from eq . ( [ ae ] ) @xmath91 \ . \label{dv2}\ ] ] @xmath88 would be the same if one considers alternatively a ng boson associated with the breaking of @xmath92 or @xmath93 . the reason is , @xmath94 oscillations only violate @xmath95 not the total number @xmath96 . in order to calculate the radial component @xmath97 one needs to know the fluxes of @xmath42 and @xmath1 as functions of the radius and their energy spectra . let @xmath98 and @xmath99 be the number of neutrinos emitted per unity of time and energy , @xmath100 and @xmath101 respectively . if the transition is non - adiabatic , the number of converted neutrinos with a given energy @xmath102 is only a fraction @xmath103 of the radiated particles , @xmath104 is the level crossing probability , and so @xmath105 the problem of obtaining these rates as functions of the radius is simplified by neglecting the oscillation length of the resonantly converted neutrinos : in that limit a neutrino with energy @xmath102 is converted to the other flavor at the position where the resonance condition is verified : @xmath106 in addition , the level crossing probability @xmath107 can be calculated using the landau - zener approximation @xcite , @xmath108 these approximations change the numbers but not , i believe , the lesson taken by comparing the results with and without a majoron field . eqs . ( [ vw],[dv2 ] - [ pc ] ) establish the differential equations for @xmath109 and @xmath110 . they are non - linear because the derivative of @xmath111 is not independent from the majoron potential @xmath97 . energy . dashed curve : sm with @xmath112 and @xmath113 . dotted and bold curves : a nambu - goldstone field exists with @xmath114 and @xmath115 respectively . the luminosities are @xmath116 for @xmath1 and @xmath117 for @xmath42.,width=302 ] let us examine the @xmath94oscillations with the mixing parameters of the non - adiabatic solar neutrino solution ( for a recent update see @xcite ) , choosing in particular the values @xmath118 , @xmath119 @xmath120 . in a supernova the resonance is non - adiabatic as well and , as eq . ( [ pc ] ) indicates , the survival probability @xmath103 increases with the @xmath0 energy . a detailed study in the framework of the sm was done in @xcite . in fig . 1 , @xmath121 in the landau - zener approximation is plotted against the energy . the dashed curve holds for the sm potential with a constant @xmath122 . it is manifest the aggravation of the non - adiabaticity with the energy . to study the majoron case one has to specify the energy spectra and luminosities . i used fermi - dirac distributions with the following values of temperature and chemical potential @xcite : for @xmath1 , @xmath123 and@xmath124 ; for @xmath42 , @xmath125 . this gives average energies of @xmath126 and @xmath127 respectively . the luminosity intensities are in turn @xmath76 for @xmath1 and @xmath117 for @xmath42 which amount to particle emission rates of @xmath128 and @xmath129 respectively . because the @xmath51-neutrinos are more numerous , the @xmath130 oscillations produce a net destruction of @xmath15-number and a positive potential @xmath131 . the dynamics is the following : the less energetic @xmath1 oscillate to @xmath42 at smaller radius than @xmath132 ; the @xmath133 conversion produces a positive @xmath97which attenuates the fall of the total potential @xmath134 with @xmath56 ; consequently , the adiabaticity improves at larger radius and the most energetic neutrinos change flavor with higher probabilities . as a function of the radius . the dashed curve is the sm potential , the dotted and bold curves are the majoron cases with the same parameters as in fig . 1.,width=302 ] fig . 2 @xcite shows the total potential @xmath88 as a function @xmath56 . the dashed curve stands for the sm potential , @xmath77 , with @xmath135 and @xmath113 . the dotted and bold curves differ in the scale of @xmath15 symmetry breaking , @xmath114 and @xmath115 , respectively . the potential falls more slowly as the field @xmath44 grows which reflects on the level crossing probability plotted in fig . 1 @xcite for the same cases ( dotted and bold curves ) . the effect is clear : the stronger the majoron field , the more efficient is the flavor conversion . it is worth to mention that if the @xmath136-neutrinos were more numerous than the @xmath51-neutrinos the effect would be the opposite because the majoron potential would be negative ( @xmath137 ) . that shows up in the rise of @xmath107 at the @xmath42 energy band around @xmath138 . and @xmath42 as emitted from the neutrinospheres . the dashed curve represents the @xmath1 luminosity after electroweak neutrino oscillations whereas in the bold curve a @xmath44 field exists with @xmath115.,width=324 ] fig . 3 @xcite shows the implications for the outgoing @xmath1energy spectrum . the dotted curves are the assumed luminosity distributions for the emitted @xmath1s and @xmath42s . the dashed curve represents the luminosity of the @xmath51-neutrinos that come out of the star after standard msw oscillations and the bold curve is the same but with a majoron field ( @xmath115 ) . the improvement of adiabaticity makes more @xmath139s convert into @xmath1 and less @xmath1 to survive , and because the @xmath136-neutrinos are more energetic , the outgoing @xmath1spectrum is harder than if there was no majoron field . the average energy of the outgoing @xmath51-neutrinos is @xmath140 for @xmath141 but reaches @xmath142 if @xmath114 and @xmath143 if @xmath144 . these effects on the spectra can in principle be traced in those detectors such as super - kamiokande and sno , capable of detecting supernova electron - neutrinos @xcite . to summarize , if the explanation of the solar neutrino deficit is the msw non - adiabatic oscillation @xmath145(@xmath146 @xmath147 or any other superposition of @xmath42 and @xmath147 ) then , the standard model predicts that in a supernova the @xmath94 transitions are also non - adiabatic . it means that , to a large extent , the @xmath51-neutrinos preserve their lower energy spectrum , unless @xmath1 also mixes to another flavor with a too high or too low @xmath86 to show up in solar neutrinos . if however , @xmath15 is a spontaneously broken quantum number , the associated nambu - goldstone boson , @xmath44 , will acquire a classic field configuration which may be strong enough to produce a back reaction that improves the adiabaticity of the @xmath94transitions . the outcome is a @xmath1 energy spectrum harder than expected . in 1987 , the existing detectors were only able to detect electron anti - neutrinos but the now operating super - kamiokande and sno experiments will be capable of detecting supernova @xmath1 events . the analysis of the energy distribution can in principle reveal or put limits on that kind of effect . the scenarios of neutrino mixing change considerably if one considers the evidences from atmospheric and terrestrial neutrino experiments ( for a review see @xcite ) . the atmospheric neutrino anomaly and the zenith angle dependence observed by super - kamiokande can be explained by @xmath148 oscillations@xcite , @xmath149 excluded chooz @xcite , but that is still consistent with solar neutrinos oscillations into a linear combination of @xmath150 and @xmath147 . that is no longer true if one takes in consideration the lsnd evidences of @xmath151 @xcite and @xmath132 @xcite . the much larger @xmath86 scale involved in lsnd ( @xmath152 ) , calls for a fourth flavor - a sterile neutrino - into which solar @xmath1s should oscillate . that changes some features concerning the effects of a ng field @xmath44 but one thing is maintained : the signature of ng fields is a _ surprise _ _ i.e. _ , an oscillation pattern not consistent with sm weak interactions @xcite . an important point is that the nambu - goldstone fields are proportional to the reaction rates that violate the relevant quantum numbers . in the case of neutrino oscillations this manifests as a dependence on the neutrino luminosity magnitudes . the effects of the majoron fields if any , will be observed in a shorter or longer interval of time depending on the actual scale of lepton number symmetry breaking . thus , the observation of such correlation with the flux intensities , would thus provide a measurement of the scale of spontaneous symmetry breaking . in the framework of the sm weak interactions the anti - neutrinos can not undergo a resonant oscillation if the neutrino masses follow the same hierarchy as the charged leptons . that is because the sm potential @xmath77 is positive in an electron rich star medium and the potential reverses sign for antineutrinos . the ng boson contributions are not only non - universal in flavour but also their signs depend on the flavour content of the neutrinofluxes and dominant flavor transitions . this allows in principle other than the standard msw type of oscillations . if one of the partial lepton numbers is spontaneously broken one expects the same happens for all three flavours . let @xmath153 be the set of scalar fields , singlets of su(2)xu(1 ) , with quantum numbers @xmath154 under the symmetry group @xmath155 . if no residual symmetry is left after spontaneous symmetry breaking there are three ng bosons . by going to the physical basis as in eqs . ( [ tpsi]-[tsigma ] ) , now summing over the quantum numbers @xmath156 , @xmath157 one obtains the lagrangian of the ng fields @xmath158 the sum over @xmath159 is implicit . the matrix @xmath160 is in general non - diagonal , provided that some of the fields @xmath153 have simultaneously more than one non - trivial quantum number . such mixing implies that the ng bosons mediate an action at a distance between two distinct currents as show the equations of motion , @xmath161 where @xmath162 is the inverse matrix of @xmath163 . in a stationary regime , if the neutrinos oscillate causing current variations @xmath164 , the expression for the gradients of the ng bosons , @xmath165 , that generalizes eq . ( [ ae ] ) is @xcite : @xmath166 since the interaction lagrangian of the leptons has the same form as eq . ( [ lint ] ) _ i.e. _ , @xmath167 @xmath168 , the potential energy of a neutrino @xmath169 subject to stationary ng fields is given by @xmath170 as in eq . ( [ vnue ] ) . plugging in eq . ( [ alv2 ] ) , one gets for example , @xmath171 this shows that the violation of a certain lepton number in one place originates fields that acts on other flavors in other places . in the old preferred scenario ( before super - kamikande results on atmospheric neutrinos @xcite ) , the neutrino mixing angles are small , @xmath172 @xmath173 @xmath174 , and the masses obey the hierarchy @xmath175 . then @xcite , as the neutrinos come out of a supernova the first resonant transitions are between @xmath1 and @xmath147 . it is outside their resonance shell , at lower medium densities , that msw transitions @xmath176 , with smaller @xmath86 , occur . in this meeting atsuto suzuki @xcite reported super - kamiokande results on the atmospheric neutrino anomaly and zenith angle dependence that can be explained by @xmath177 oscillations . the best values , confirmed at neutrino 98 , are @xmath178 and @xmath179 . the alternative @xmath149 oscillation is excluded by chooz @xcite . this is consistent with a scenario where the heaviest of the neutrinos is essentially a superposition , @xmath180 ( possibly maximal ) , of @xmath139 and @xmath147 ; @xmath1 is the lightest neutrino and oscillates in the sun into the medium weight orthogonal superposition , @xmath181 , of @xmath42 and @xmath147 . in what follows what is said about @xmath147 , @xmath42 and quantum numbers @xmath182 , @xmath183 can be maintained if replaced by @xmath180 , @xmath184 and respective combinations of @xmath182 and @xmath40 . in a supernova the @xmath1 flux is larger than the @xmath185 flux and consequently more @xmath1 transform into @xmath185 than the reverse which makes a net decrease of @xmath186 . the current densities @xmath187 and @xmath188 suffer the variations @xmath189 that generate the ng fields @xmath190 . applying eq . ( [ alv2 ] ) one gets the vector potentials outside the sphere where @xmath191 takes place . the magnitude of the constants @xmath162 is directly related to the symmetry breaking scales , but the flavor structure is arbitrary and is not constrained by the neutrino mixing . the matrix @xmath162 has only to be symmetric and positive definite as demands the scalar bosons kinetic lagrangian . consider the following hierarchy : @xmath192 @xmath193 . then , @xmath55 is much smaller in magnitude than @xmath194 and @xmath195 and , assuming spherical symmetry , the potential energies are well approximated by @xmath196 omitted the neutral current terms . the anti - neutrino potentials are the symmetric of these . the increment of @xmath197 lepton number per unity of time , @xmath198 , is positive and so is the flux transfer @xmath199 . defining @xmath200 @xmath201 one estimates @xmath202 in the first 1/2 seconds of neutrino emission until the supernova explosion and @xmath203 at 1 second time . see @xcite for more details . the ng potential for @xmath185 is negative but that is not necessarily the case of @xmath52 because the element @xmath204 can be negative ! using eq . ( [ vng ] ) , @xmath205 the schrdinger equation for the anti - neutrinos is the same as for neutrinos , eq . ( [ osc1 ] ) , but with a symmetric potential : @xmath206 where @xmath207 @xmath208 is implicit . within the sm , the effective mass @xmath209 would be negative , and the resonance condition @xmath210 could not be met . in presence of ng fields , the effective mass @xmath211 is negative at the high density regions , but since the ng potentials only decay with the second power of the radius their contributions inevitably overcome at large enough distances . that is a crucial point . if on top of that @xmath204 is negative , the effective mass @xmath212 suddenly turns positive and may reach the value @xmath213 . in that case there is a resonance layer where @xmath214 is met . in fig . 4 it corresponds to the interception of the @xmath51 and @xmath136 curves , @xmath215 and @xmath216 , respectively . the positive part of the potential @xmath212 is plotted in fig . 4 with values @xmath217 , @xmath218 or 3 , @xmath113 , @xmath219 , and a neutrino energy of @xmath220 , the same for all curves . is plotted in the solid and dotted curves with @xmath221 and 3 respectively . the dashed line stands for a particular value of @xmath222 and the squares curve is the adiabatic boundary for @xmath223 in resonance 2.,width=324 ] take @xmath224 for example . the @xmath225 dashed line crosses the relative potential curve at two points : the first , @xmath226 , lies in the remarkably sharp transition from the @xmath227 dominated region to the ng field dominated one and very likely a non - adiabatic level crossing occurs there _ i.e. _ , @xmath228 continues @xmath228 and @xmath229 continues @xmath230 . in the second point , @xmath231 , the potential is smooth enough ( @xmath232 ) for adiabatic transitions . in general terms , the conversion probability is enhanced @xcite if @xmath233 is larger than the rate @xmath234 at resonance . in fig . 4 , @xmath235 is plotted with squares for @xmath236 . one can see that small mixing angle adiabatic conversions are possible for a wide range of neutrino masses ( for more details see @xcite ) . through this mechanism one could have resonant @xmath2 oscillations in a supernova whereas @xmath237 take place in the sun . in particular , with the parameters of the non - adiabatic solution @xcite namely , @xmath238@xmath239 and @xmath240 , provided that the neutrino fluxes are high enough ( compare the dotted @xmath241and solid curves ) . since the predicted energy spectrum of @xmath229 and @xmath242 is harder than the @xmath228 spectrum that kind of oscillation could be manifest in the sn1987a events @xcite . some analyses made @xcite , but not all @xcite , disfavour the occurrence of @xmath243 oscillations . this deserves some discussion @xcite , here i simply stress that the effects caused by the ng fields strongly depend on magnitude of the neutrino fluxes and on the scale of the global symmetry breaking as well . since the fluxes vary with time in a supernova , one could have anti - neutrino oscillations in the first instants of emission that fade away when the fluxes drop below a certain threshold . the observation in the future of such correlation between oscillation and flux magnitudes would provide a measurement of the symmetry breaking scale . we have shown that despite the fact that true nambu - goldstone ( ng ) bosons only have derivative couplings , it is still possible to have coherent ng fields . this happens whenever the quantum numbers they are associated with are not conserved in reactions taking place in a large scale . if the lepton numbers are spontaneously broken charges then , neutrinos oscillations in stars produce classic ng fields . in addition , if the scale of symmetry breaking is below 1 tev , these fields are strong enough in supernovae to change the neutrino flavor dynamics . the observable signatures are : - time variation of neutrino oscillation patterns in correlation with the flux magnitudes . this is a consequence of the ng fields being proportional to the time rate of reactions and therefore to the neutrino fluxes . - incongruence of supernova neutrino oscillation patterns with the evidence obtained from solar , atmosferic and terrestrial neutrinos . the ng fields are not expected to be significant in these cases even if they are in supernovae because the fluxes are much smaller . this is ultimately a result of the derivative coupling nature of the ng bosons . finally , if we will are lucky enough to see such kind of phenomena in the future , then , supernova neutrinos will provide a measurement of the scale of lepton number symmetry breaking . this work was supported in part by the project eso / p / pro/1127/96 . d. o. caldwell , in proceedings of `` cosmo 97 : first international workshop on particle physics and the early universe '' , ambleside , england , 1997 ( edited by l. roszkowski , world scientific , 1998 ) , hep - ph/9804367 ; y. totsuka , in proceedings of the `` xviii international symposium on lepton - photon interactions '' , hamburg , 1997 ( edited by a. de roeck and a. wagner , world scientific , 1998 ) .	it is shown that true nambu - goldstone ( ng ) bosons develop coherent fields whenever the associated charges of the matter particles are not conserved in a macroscopic scale . the sources of the ng fields are the time rates of quantum number violation . if the lepton numbers are spontaneously broken at a scale below 1 tev , the neutrino oscillation processes generate classic ng fields that are strong enough in supernovae to modify the neutrino flavor dynamics . the oscillation patterns may change in the periods of largest @xmath0 fluxes . two examples are given : 1 . the back reaction of a ng field improves the adiabaticity of the @xmath1 resonant conversion ; 2 . @xmath2 oscillations may occur even if @xmath1 is the lightest of the neutrinos .
thermodynamic heat engines convert heat into useful work . they work cyclically between two thermal reservoirs kept at different temperatures @xmath0 and @xmath1 ( @xmath2 ) . the second law of thermodynamics restricts their efficiency to the carnot limit @xcite , @xmath3 . however , this efficiency can only be achieved in the quasistatic limit where transitions between thermodynamic states occur infinitesimally slowly and hence the power output vanishes . curzon and ahlborn ( c - a ) @xcite showed that for finite time endoreversible heat engines , efficiency at maximum power is given by @xmath4 . as yet there is no consensus on this result ( @xcite ) . with the advances in nano - technology , a few - micrometer - sized stirling heat engine has been experimentally realized @xcite.this microscopic heat engine operates in conditions where typical changes in their energies are of the order of the thermal energy per degree of freedom @xcite . an appropriate theoretical framework to deal with these systems has been developed during the past decades within the context of stochastic thermodynamics @xcite . this formalism of stochastic energetics provides a method to calculate work , heat and entropy even for a single particle along a microscopic trajectory . one can obtain average quantities after averaging over respective ensembles . the averaged thermodynamic quantities , work and entropy , obey second law . using this formulation various single particle heat engines have been studied in the literature @xcite . fluctuation relations for heat engines ( frhe ) @xcite operating in a time periodic steady state ( tpss ) have recently been obtained @xcite . frhe are in the form of equality and carnot s inequality for efficiency @xmath5 follows as a direct consequence of this theorem . in the present work we have studied in detail a simple model for a stochastic heat engine described by langevin equation . both underdamped and overdamped regimes are explored and qualitative differences are pointed out . we emphasize on fluctuations of thermodynamic variables including the engine efficiency . we show that fluctuations dominate the mean values even in quasistatic regime . therefore in such situations one needs to study the full probability distribution of the physical variable for the proper analysis of the system . in section [ mod ] , we describe the model of our system and the protocol . in section [ uql ] , we obtain analytical results for relevant average thermodynamic quantities in the quasistatic regime for the underdamped case . in section [ unl ] , engine with finite time cycle in the inertial regime is studied numerically , in detail . the system driven by time asymmetric cycles and various other model systems are also explored . we have verified frhe in this section . sections [ oql ] and [ onl ] are devoted to the analytical and numerical studies of the system in the overdamped limit . finally , we conclude in section [ con ] . each section is self contained . the single particle stochastic heat engine consists of a brownian particle having position @xmath6 and velocity @xmath7 at time @xmath8 , confined in a one dimensional harmonic trap . the stiffness of the trap @xmath9 varies periodically in time as shown in fig.([fig - pro ] ) . for the underdamped case , the equation of motion for the particle is given by @xcite @xmath10 in overdamped limit the equation reduces to @xmath11 in our further analysis , we set mass of the particle m , the boltzmann constant @xmath12 and the frictional coefficient @xmath13 to be unity . @xmath14 is the temperature of the thermal bath . all physical parameters are made dimensionless . the noise is gaussian with zero mean , @xmath15 and is delta correlated , @xmath16 . the internal energies of the particle in the underdamped and the overdamped limit are given by @xmath17 and @xmath18 , respectively . ) . , width=283 ] and @xmath0 . the cycle consists of two isothermal steps and two adiabatic steps according to the time varying protocol k(t ) . the blue line denotes a one dimensional potential v(x , t ) and the filled region denote the corresponding steady state distribution.,width=283 ] operation of the system consists of four steps - two isotherms and two adiabatics . in the first step , the system undergoes an isothermal expansion , during which it is connected to a hot bath at temperature @xmath1 and the stiffness constant is varied linearly with time as @xmath19 for @xmath20 . here @xmath21 is the period of the cycle and @xmath22 is the initial value of the stiffness constant . in the second step , the potential undergoes an instantaneous expansion ( adiabatic ) by decreasing the stiffness constant from @xmath23 to @xmath24 . as the process is instantaneous the distribution before and after expansion will not change and heat absorption will be zero . in the third step , the system is connected to a cold bath with lower temperature @xmath0 and isothermal compression of the trap is carried out by changing the stiffness as @xmath25 for @xmath26 . in the last step , we carry out instantaneous adiabatic compression by varying the stiffness constant from @xmath23 to @xmath22 and simultaneously connecting the system to the hot bath . this cycle is then repeated . the time dependence of the protocol is given in fig.([fig - pro ] ) and a schematic representation of the system within a cycle at its various stages is depicted in fig.([fig - car ] ) . the described protocol differs from those used in earlier studies . in the experimental set up @xcite two adiabatic steps are absent . work optimized protocol is used by schmiedl and seifert @xcite whereas the protocol based on the concept of shortcut to adiabaticity is used by tu @xcite . however , their emphasis is on the possible correlation between efficiency at maximum power and c - a bound . our main motivation , namely to study the fluctuation of physical quantities , is different from earlier studies as mentioned in the introduction . in this section , we analytically calculate the average thermodynamic quantities of our model system in the quasistatic limit . in this limit , the duration of the protocol is much larger than all the relevant time scales , including the relaxation time . hence as protocol is changed , the system immediately adjusts to the equilibrium state corresponding to the value of protocol at that instant . first , we calculate the average work done on the particle in all the four steps of a cycle and the heat absorbed by it in the first isothermal step . finally , we calculate efficiency in the quasistatic limit . in the first isothermal process , average work done on the particle is the same as the free energy change @xmath27 before and after the expansion , i.e. , @xmath28 at @xmath29 , the system is in equilibrium with the bath at @xmath1 with stiffness constant @xmath23 . the second step being instantaneous , no heat will be dissipated and the phase space distribution remains unaltered . correspondingly the average work done on the particle is equal to the change in its internal energy : @xmath30 where @xmath31 , is the normalization constant . similarly in the third step ( i.e. , isothermal compression step ) the average work done on the particle in the quasistatic limit is @xmath32 the average work done in the last step ( i.e. , second adiabatic step ) is given as @xmath33 with @xmath34 . hence , the average total work done in the full cycle of the heat engine in the quasistatic process is @xmath35 to obtain the heat absorption in the first step ( i.e. , isothermal expansion ) , we calculate the average change of internal energy and use the first law . during this process , the particle stays in contact with hot bath at temperature @xmath1 . however , it is to be noted that at time @xmath36 , the system was in contact with low temperature bath at @xmath0 , whereas at @xmath37 the system is in contact with hot bath at @xmath1 . thus the system has to relax into new equilibrium after sudden change in temperature . the time taken for this relaxation process is assumed to be negligible compared to the cycle time @xmath21 . this relaxation leads to an additional heat flow which accounts for the change in the internal energy during the relaxation process . one can readily obtain the internal energy at @xmath37 as @xmath38 while after the relaxation it is @xmath1 . hence , the average internal energy change in the first step is @xmath39 now using the first law , the average heat absorption from the hot bath for the first step is @xmath40 hence efficiency of the engine for the underdamped case in the quasistatic limit is given by @xmath41 here we would like to emphasize that @xmath42 is defined ignoring fluctuations and the subscript q denotes the quasistatic limit . we will show later that fluctuations play an important role even in the quasistatic regime . work done during the cycle @xmath43 and heat absorbed during the first step @xmath44 are fluctuating quantities . stochastic efficiency is defined as @xmath45 @xcite and hence its average @xmath46 is not the same as @xmath47 which is given in eq.([uq - eff ] ) for quasistatic limit . this will be discussed in detail in subsequent sections . in our notation , the thermodynamic quantities are denoted by capital letters only for quasistatic limit , whereas , small letters are used to denote those quantities for finite time cycles . according to our convention negative work done on the system implies extraction of work ; while , negative heat means that heat enters into the system . it is important to note from eq.([eq - q ] ) that in quasistatic limit heat flows from the bath to the system provided @xmath48 and similarly from eq.([eq - w ] ) work can be extracted from the system if @xmath49 therefore , in quasistatic regime our model system operates in three different modes of operation depending on the ratio of the temperatures of the thermal baths . first , when @xmath50 is maintained , work can be extracted and heat is absorbed from hot bath and it acts as an engine . second , when @xmath51 is set , heat is absorbed from the bath but we can not extract work . and finally when we have @xmath52 neither heat is absorbed nor the work is extracted . in this case work done on the system heats up the hot bath . therefore , there is a particular regime in parameter space where the system act as an engine . this is in contrast to the carnot engine which works for arbitrary temperature difference between two baths . the above mentioned condition is only valid in the quasistatic limit . for finite time cycle the operational condition for heat engine depends on cycle time apart from @xmath1 and @xmath0 , which will be shown in our simulation . our exact expression of @xmath53 and @xmath54 are in complete agreement with our numerical results in the quasistatic limit . thus these analytical calculations act as a check for our numerical simulation . for finite - cycle - time we study our system numerically . when the langevin system is driven periodically it is known that after initial transients , the system will settle down to a tpss . the joint probability distribution @xmath55 of position and velocity of the particle is periodic in time , i.e. , @xmath56 . for numerical simulations we evolve our system with a time periodic protocol ( as shown in fig.[fig - pro ] ) . we have used heun s method for integrating the basic langevin equation @xcite with time step @xmath57 . we make sure that the system is in the tpss by going beyond the initial transient regime . we then consider at least @xmath58 cycles of operations and physical quantities are averaged over all these cycles . for rest of the paper we keep m , a , @xmath13 fixed at @xmath59 , @xmath60 , @xmath61 . we now make use of the concepts of stochastic energetics @xcite to calculate work , heat and internal energy for a given trajectory . the thermodynamic work done on the particle during first part of the cycle , in each computational step @xmath62 , is given by @xmath63 with @xmath64 and @xmath65 . now , @xmath66 where @xmath67 . the internal energy is a thermodynamic state function and hence its change during the isothermal process is given by @xmath68 . the heat absorption by the bath is @xmath69 using the first law of thermodynamics . the second step which is adiabatic is instantaneous and hence the particle does not get any chance to evolve . thus work done is only instantaneous change in internal energy , i.e. , @xmath70x^2(\tau/2)$ ] . similarly , for step three , work done is given by @xmath71 and @xmath72 ; internal energy change @xmath73 ; and heat delivered to the cold bath is @xmath74 . for the last adiabatic process , work done on the particle is @xmath75x^2(\tau)$ ] . the total work done on the system in a cycle is @xmath76 . it should be noted that each @xmath77 ( i=1,2,3,4 ) is a fluctuating quantity and their values depend on a particular phase space trajectory . and @xmath21 but for fixed @xmath78 . , width=283 ] in fig.([uf - phase ] ) , we have shown the phase diagram of the operation of our system . here we have varied @xmath1 and cycle time @xmath21 keeping @xmath0 fixed at 0.1 . there are three distinct regimes . the system acts as an engine when @xmath79 and @xmath80 . the angular bracket @xmath81 indicates average over several realizations . in the other two regimes the system ceases to work as a heat engine altogether ( @xmath82 ) . for @xmath83 we have two distinct domains with @xmath80 and @xmath84 . the latter implies work is done on the system which heats up the hot bath . in the large cycle time limit numerical results are consistent with our analytical predictions made in last section . we re - emphasize that the system works as a heat engine provided there is a minimal difference between @xmath1 and @xmath0 which depends on cycle time @xmath21 and other physical parameters . from the phase diagram it is apparent that , as we decrease @xmath21 for fixed @xmath1 , there exists a lower bound below which the system does not perform as an engine , as it only consumes work . , @xmath85 and @xmath86 with cycle time @xmath21 . , width=283 ] in fig.([uf - work ] ) , we have plotted @xmath87 , @xmath85 and @xmath86 with respect to cycle time @xmath21 . we have fixed @xmath88 and @xmath78 for all subsequent figures . starting from zero , @xmath87 initially increases and reaches a peak value . then it starts decreasing and finally saturates to a negative value ( -0.214 ) , which is close to our theoretical result ( from eq.([eq - w ] ) ) . the work can be extracted in the region where it becomes negative . as we increase cycle time , @xmath85 changes dramatically . it has a positive region sandwiched between two negative regions . when @xmath89 heat is released to the hot bath while work is done on the particle . in the quasistatic limit it saturates at the theoretical value -0.523(from eq.([eq - q ] ) ) . in contrast to @xmath85 , @xmath86 is always positive , i.e. , heat is always released to the cold bath . internal energy being a state function , @xmath90 is zero over a cycle in tpss and hence @xmath91 . using the saturation value of @xmath87 and @xmath85 we immediately get @xmath86 to be equal to 0.310 which is close to our numerical result . we now study the nature of stochastic efficiency @xmath92 and engine power @xmath93 as a function of cycle time . the engine is in tpss where probability distributions of system variables are periodic in time . however , for a given realization of a cycle , state of the system ( position and velocity ) does not come back to its initial state . thus for each cycle thermodynamic quantities will depend on the particular microscopic trajectory and hence @xmath43 , @xmath44 , @xmath94 , @xmath92 and @xmath95 are all fluctuating quantities from cycle to cycle . the average efficiency is defined as @xmath96 . due to fluctuation in @xmath43 and @xmath44 , it is to be noted that @xmath97 . fluctuation theorems @xcite put stringent condition on @xmath98 which is bounded by the carnot efficiency i.e. , @xmath99 . however , no such bound exist for @xmath100 @xcite . and @xmath101 with cycle time @xmath21.the doted blue line denotes the quasistatic limit for @xmath101 . , width=283 ] the first law for any microscopic realization of cycle can be written as @xmath102 the change in the internal energy @xmath103 is unbounded . it is zero only on the average . similarly @xmath104 and @xmath43 take values in the range ( @xmath105 ) but are constrained by first law . hence it is not surprising that @xmath92 can take values between @xmath106 to @xmath107 . in fig.([uf - eff ] ) we have plotted efficiencies @xmath100 and @xmath101 as a function of cycle time . initially for small @xmath21 , our system does nt work as an engine . due to large dissipation work can not be extracted ( @xmath108 ) . in this regime , efficiency is negative . on further increasing @xmath21 , efficiency becomes positive and it monotonically increases . for large @xmath21 , @xmath100 and @xmath101 the saturation value for @xmath101 is 0.41 which can be obtained analytically in quasistatic regime . in general @xmath109 . we find both @xmath100 and @xmath101 are less than the carnot efficiency @xmath110 . with cycle time @xmath21 . , width=283 ] in fig.([uf - pow ] ) , average power @xmath111 is plotted as a function of @xmath21 . there is a negative region for low cycle time . beyond the critical value of @xmath112 , power becomes positive and exhibits a peak and finally tends to zero in the large @xmath21 limit . the efficiencies @xmath100 and @xmath101 at maximum power are given by 0.16 and 0.25 respectively . both of these values are less than @xmath113 . for different cycle times ( @xmath21=0.7 , 7.0 , 70.0 ) . , width=283 ] as mentioned earlier , physical quantities @xmath44 , @xmath43 and @xmath92 are strongly fluctuating variables . to study these fluctuations we focus on probability distribution of these quantities @xmath114 , @xmath115 and @xmath116 . in figs.([uf - wdist ] ) , ( [ uf - q1dist ] ) and ( [ uf - effdist ] ) we have plotted them for three different time periods . for @xmath117 , distribution of @xmath43 and @xmath44 are sharply peaked around zero with @xmath118 , @xmath119 . as we increase the cycle time @xmath115 and @xmath120 become broad , asymmetric and shift towards negative side . for large negative value of arguments the distributions exhibit long tail . for large positive values of @xmath43 and @xmath44 the distribution falls off exponentially or faster @xcite . the trajectory responsible for positive values are atypical and sometimes referred to as transient second law violating trajectories @xcite . strong fluctuations in heat and work persist even in the quasistatic limit(@xmath121 ) . these fluctuations in work are mainly attributed to two adiabatic processes , while fluctuations of @xmath44 result from relaxation process when the system , in contact with low temperature bath , is brought in direct contact with high temperature reservoir . for different cycle times . , width=283 ] for different cycle times . , width=283 ] for @xmath117 , @xmath100 is negative ( -0.26 ) . the distribution @xmath116 is asymmetric and there is a broad shoulder on the negative side . as we increase @xmath21 , distribution shifts towards positive side . it is not surprising to see the finite weight for values @xmath122 and @xmath123 @xcite . as we increase the cycle time the standard deviation of @xmath92 ( @xmath124 ) , becomes smaller . however , it remains larger compared to mean values . for example , @xmath125 and corresponding @xmath126 at @xmath127 and @xmath128 whereas @xmath129 for @xmath130 . we would like to emphasize that mean is dominated by fluctuations even in the quasistatic regime . any physical quantity with relative variance larger than one , is referred to as non - self averaging quantity . for such cases mean ceases to be a good physical variable and one has to resort to the analysis for full probability distribution . this is one of our main result . non - self averaging quantities arises mainly in physics of quenched disordered systems . note that , @xmath92 becomes positive if both @xmath43 and @xmath44 are positive or both of them are negative . @xmath92 becomes negative when @xmath43 and @xmath44 have opposite signs . in order to have a better understanding of our system we have plotted the joint distributions of @xmath43 and @xmath44 for different @xmath21 in fig.([tau0.7 ] ) . for a given cycle , the system acts as an engine when both @xmath43 and @xmath44 are negative i.e , in the third quadrant of the plot . using our numerical results we have calculated the ratio of the total number of realizations falling in the third quadrant to the total number of realizations . these fractions for @xmath117 , 7.0 and 70.0 are calculated to be 0.226 , 0.583 and 0.858 , respectively . it is clear from this that for large cycle times the reliability of the system working as an engine increases . though we observe that even in quasistatic regime there are realizations for which the system does not act as an engine this is due to strong fluctuations in work and heat as discussed earlier . in tpss the joint probability density @xmath55 is periodic in time : @xmath131= @xmath55 . for simplicity we write @xmath132 . from the definition of stochastic entropy @xcite of the system @xmath133 , the change in the system entropy for a trajectory over a cycle is given by @xmath134 where @xmath135 and @xmath136 are the initial and final phase space points for a particular realization of the cycle . to calculate @xmath137 we evaluate @xmath138 at the initial point of the cycle which also coincide at the end point @xmath139 . in fig.([u-3d ] ) we have plotted joint phase space distribution at tpss for three different values of @xmath117 , @xmath140 and 70.0 . we see that for @xmath117 and @xmath127 phase space distributions are not symmetric and there exist strong correlation between @xmath6 and @xmath7 which was ignored in the earlier literature @xcite . only in the large @xmath21 limit the distribution becomes symmetric ( see fig.([u-3d]c ) ) . the cross - correlation between position and velocity disappears and the distribution @xmath141 becomes uncorrelated gaussian in the quasistatic limit . due to correlation , the width of the distribution become larger as we decrease cycle time @xmath21 . recently , frhe in tpss has been derived @xcite . it extends the total entropy production fluctuation theorem of seifert @xcite applied to heat engine . the total entropy production @xmath142 over a cycle is a stochastic variable and in our present case is given by @xmath143 using the first law ( eq.[first - law ] ) @xmath144 the second law which is valid on average , can be stated as @xmath145 . in tpss , @xmath146 , which implies @xmath147 . thus the second law puts the constraint on efficiency which is defined as @xmath98 . it should be noted that this constraint is valid for any finite time cycle in tpss , unlike the carnot which is valid for macroscopic engines in the quasistatic regime . however , it does not put any constraint on the average efficiency ( @xmath148 ) . the fluctuation theorem for heat engine replaces the inequality relation of the second law by the equality relation , namely @xcite , = 7.0 . , width=283 ] = 7.0,width=283 ] . [ uf - ent ] @xmath149 eq.([ift ] ) is frhe in tpss . by calculating all the relevant stochastic variables @xmath43 , @xmath44 , @xmath137 , @xmath103 over all trajectories for finite @xmath21 we have verified eq.([ift ] ) in the tpss . we have obtained the value to be 0.96 for @xmath127 , which is well within our numerical accuracy . we would like to emphasize that , in eq.([ift ] ) , four stochastic variables appear in the exponent . small changes in these values affect the exponential function by a large amount . given this fact , our observed value of @xmath150 is quite satisfactory . for the same parameter value @xmath127 , in fig.([uf - u ] ) we have plotted the probability distribution , @xmath151 , as a function of @xmath103 . in fig.([uf - ent ] ) , we have plotted the probability distribution of change of system entropy @xmath152 and total entropy @xmath153 as a function of their arguments . it is clear that as @xmath154 and @xmath155 are state functions , @xmath151 and @xmath152 are symmetric with zero mean . however , the distribution @xmath153 is asymmetric with a long tail for positive large @xmath142 . there is also a finite weight towards negative @xmath142 . this contribution arises due to transient second law violating periodic cycles @xcite . however , @xmath156 remains positive as demanded by the second law . till now we concentrated on symmetric cycle , i.e. , equal contact times of the system with hot and cold bath . naturally , the question arises what will happen if the cycle is time asymmetric . to the best of our knowledge this question has not been addressed in earlier literature . if the contact time of one bath is different from that of the other , it can affect work output , heat dissipation to each bath , power and efficiency . however , in the quasistatic limit there should not be any effect of this asymmetry . this is clear from fig.([uw - new ] ) that the average work , for three different asymmetric cycles , asymptotically approach each other in the quasistatic limit . in the non - quasistatic limit , work extracted by the engine for asymmetric cycles is small compared to symmetric cycle . from fig.([ueff - new ] ) it is seen that @xmath100 is lower for asymmetrical cycles . the inset shows even in quasistatic limit @xmath109 for @xmath157 . we have verified separately that asymmetry also decreases the power . thus asymmetry in the cycle degrades the performance characteristics of the engine . vs @xmath21 for symmetric as well as asymmetrical cycles . here @xmath158 and @xmath159 are contact times of the particle with hot and cold bath respectively . thus , @xmath160 for our case . , width=283 ] vs @xmath21 for symmetric as well as asymmetrical cycles . inset : comparision of @xmath100 with @xmath101 for @xmath157.,width=283 ] we now briefly compare the nature of power and efficiency of our system when the confining potential is different . we have taken the confining potential @xmath161 with n=2,4,6 . for n=4,6 the confining potentials are referred to as hard potential . the equilibrium distributions for hard potentials are no longer gaussian and hence in the quasistatic limit , the average work , heat dissipation etc . , will be different from those for the harmonic potential . in fig.([uh - eff ] ) and ( [ uh - pow ] ) we have plotted @xmath100 and @xmath111 as a function of cycle time for different potentials . average efficiency @xmath100 for large @xmath21 decreases as potential becomes harder and thereby degrading the performance . @xmath100 saturates at the higher value of @xmath21 ( not shown in the figure ) . from fig.([uh - pow ] ) we observe that harder the potential smaller will be the critical time @xmath21 above which the system acts as an engine . for large cycle time the power decreases as the potential becomes harder . however , we see clearly that there are three values of efficiencies @xmath100 and @xmath101 at maximum power 0.16 , 0.10 , 0.08 and 0.25 , 0.16 , 0.13 for n=2,4,6 respectively . it is apparent that the efficiency at maximum power is model dependent and decreases as the potential becomes harder . even the saturation value is different and it is lower for harder potential . clearly , these two figures indicate that operational characteristics of our system are model dependent . thus we do not expect any universal relation involving only the average efficiency at maximum power and temperatures of the reservoirs . with @xmath21 for different types of potentials.,width=283 ] with @xmath21 for different types of potentials . , width=283 ] so far we have studied our system in detail in the underdamped regime which is a general case . from now on we restrict to the overdamped regime and highlight some qualitative differences . in the overdamped limit , dynamics of the system follows langevin eq.([u - lan1 ] ) , where inertial effects are ignored . this approximation is valid when the time steps of the observation are much larger than @xmath162 . the internal energy of the system is given only in terms of potential energy . for this case equilibrium distribution of a particle in a static harmonic potential is given by @xmath163 from which one can easily obtain the free energy . the analytical calculation for average thermodynamic quantities in quasistatic limit are similar to the underdamped case . the total average work done on the particle during the entire cycle is given by as @xmath164 interestingly , the expression for @xmath53 remains the same as for the case of the inertial system discussed earlier and the system extracts work provided @xmath165 . using same arguments similar to the underdamped case and keeping in mind only the fact that there is only one phase space variable , namely position , the average internal energy change in the overdamped limit in the first step can be expressed as @xmath166 using the first law the average heat absorption from the hot bath during the first step is @xmath167 and @xmath21 but for fixed @xmath78 . , width=283 ] the expression for efficiency in the overdamped case is @xmath168 which is different from the earlier case . in quasistatic limit , from eq.([o - q1 ] ) heat flows from the bath to the system provided @xmath169 . this ratio @xmath170 differs from that obtained for the underdamped case . from eqs.([o - wtot ] ) and ( [ o - q1 ] ) , the system acts as an engine for the same condition ( @xmath171 ) as for the underdamped case . a finite temperature difference between hot and cold bath is required so that the system can act as a heat engine . analysis for finite time cycle is carried out by numerical methods as discussed earlier . for a better understanding in the overdamped regime , all the parameters have been kept same as in the underdamped case . in fig.([o - phase ] ) , we have plotted the phase diagram for the overdamped case keeping @xmath0 fixed at 0.1 . for large @xmath21 ( quasistatic limit ) we observe , from phase diagram , that the system operates as a heat engine provided @xmath1 is greater than a critical value . this critical value is close to the theoretical value of 0.142 obtained from the bounds determined in quasistatic calculation . the phase diagram shows a qualitative difference from the underdamped phase diagram ( fig.([uf - phase ] ) ) . the system always acts as an engine in @xmath172 limit provided we are above a critical value of @xmath1 , which is not the case for the underdamped engine . this is clear from fig.([of - work ] ) , where we have plotted average work done on the system @xmath87 and average heat released to each bath with @xmath85 and @xmath86 as a function of @xmath21 . note that the observed anomalous part for @xmath87 and @xmath85 in the underdamped case for small @xmath21 regime is absent in this regime . the quantities @xmath87 , @xmath85 and @xmath86 show monotonic behavior and saturate at large cycle time to their analytical limits -0.214 , -0.324 and 0.110 , respectively . unlike the underdamped case here , @xmath87 and @xmath85 are always negative . , @xmath85 and @xmath86 with cycle time @xmath21 . , width=283 ] and @xmath101 with cycle time @xmath21 . the doted blue line denotes the quasistatic limit for @xmath101 . , width=283 ] with cycle time @xmath21 . , width=283 ] in fig.([of - eff ] ) we have plotted the average of efficiency @xmath173 and @xmath101 as a function of @xmath21 . both the efficiencies increase monotonically from zero and saturate for large @xmath21 . the saturation value of @xmath101 is close to the theoretically predicted value of 0.660 . the saturation value of @xmath100 is found numerically to be 0.571 which is much less than the corresponding value of @xmath101 . both these values are less than the carnot value @xmath110 . it is clear that @xmath174 due to the strong correlation between fluctuating variables @xmath43 and @xmath44 for all @xmath21 . from fig.([of - pow ] ) , we see that power exhibits a sharp peak at @xmath175 . corresponding efficiencies @xmath100 and @xmath101 at maximum power are equal to 0.11 and 0.51 which are less than the c - a result ( @xmath113 ) . for different cycle times in the overdamped case . , width=283 ] for different cycle times in the overdamped case . , width=283 ] for different cycle times in the overdamped case . , width=283 ] to study the nature of fluctuations in the overdamped regime we have plotted the distribution @xmath115 , @xmath114 and @xmath116 in figs.([ow - dist ] ) , ( [ oq1-dist ] ) and ( [ oeff - dist ] ) respectively . the qualitative nature of the distributions of @xmath115 and @xmath114 remain the same for different values of @xmath21 as in the underdamped case . the fluctuations are smaller compared to the underdamped case . the distribution @xmath116 shows a double peak behavior for @xmath117 with @xmath176 and standard deviation @xmath177 . for @xmath127 , @xmath178 and @xmath179 . for @xmath121 , @xmath180 and @xmath181 . we observe that even in the quasistatic regime , fluctuations of @xmath92 dominate over the mean value . thus @xmath92 is a non self averaging quantity . we have also seen that the fraction of the realizations for which the system acts as an engine , increases with cycle time @xmath21 . numerical values for these fractions are 0.488 , 0.817 and 0.861 , for @xmath117 , 7.0 and 70.0 , respectively . hence , finite fraction of realization does not act as an engine even in quasistatic limit . similar to the underdamped case , the reliability of the system to act as an engine increases with @xmath21 . finally , we discuss the performance characteristics of our system in the overdamped regime using an experimental protocol @xcite . the experimental protocol consists of only two steps , in which the two adiabatic steps of fig.([fig - pro ] ) are absent . in the quasistatic regime the system acts as an engine for any temperature difference and there is no bound on @xmath1 unlike our four step protocol @xcite . this suggests that the phase diagram will depend on the nature of protocol as well as on system parameters and is not unique . as discussed earlier , most of the work fluctuations specially in quasistatic regime arise from two adiabatic steps . in the absence of these two steps , we have observed in our simulation that in the quasistatic regime , work distribution @xmath115 is sharply peaked like a delta function at @xmath182 ( analytical result ) @xcite . however , fluctuations in @xmath44 persist even in the quasistatic regime as a result of the relaxation process that follows when the system , in contact with the cold bath , is brought in direct contact with high temperature reservoir . the distribution of stochastic efficiency @xmath116 exhibits a qualitative differences . it has almost zero weight for @xmath183 in large @xmath21 limit and shows a broad double peak feature which is confined in the region @xmath184 . beyond @xmath185 a long tail is observed . for @xmath121 we have numerically calculated @xmath186 and @xmath187 . even for this protocol we notice that fluctuations dominates over the mean value . the details of these results will be published elsewhere @xcite . we summarise our results in this section . we have carried out an extensive analysis of a single particle stochastic heat engine by manipulating a brownian particle in a harmonic trap with a periodically time dependent stiffness constant as a protocol . the cycle consists of two isothermal steps and two adiabatic steps similar to that of carnot engine . the proposed model is studied taking into account both the inertial and overdamped langevin equations . thermodynamic quantities , defined over microscopic phase space trajectory of our system , fluctuate from one cycle of operation to another . their magnitude depends on the trajectory of the particle during the cycle . this is done by using the methods of stochastic energetics . average value of thermodynamic quantities and their distribution functions have been calculated numerically in tpss . analytical results of average thermodynamic quantities have been obtained in the quasistatic regime . these results are consistent with the corresponding numerical results . we have reported several new results which were not addressed in earlier literature . the full phase diagram for operation of a system is given in both inertial and high friction regime . they differ from each other qualitatively . in both cases it is also shown that system acts as an engine provided the temperature difference is greater than a critical value ( unlike carnot engine ) . this critical value depends on system parameters and is consistent with analytical results in quasistatic limit . moreover , for fixed bath temperatures and system parameters there should be a critical cycle time above which the system acts as an engine . the mean of the stochastic efficiency is dominated by its fluctuations ( @xmath188 ) even in quasistatic regime , making the efficiency a non - self averaging quantity . for such cases mean ceases to be a good physical variable and one has to resort to the analysis for full probability distribution . this is one of our main result . we have also shown that @xmath189 . our analysis of model dependence of finite cycle time clearly rules out any simple universal relation ( e.g. , @xmath190 ) between efficiency at maximum power and temperature of the baths . time asymmetric periodic protocol makes engine less efficient . only in the quasistatic regime time asymmetry does not play any role . for given cycle time there are several realizations which do not work as a heat engine . these are referred to as transient second law violating trajectories . number of these realizations decreases as we increase @xmath21 . the fractions of realisations following second law with corresponding @xmath21 are reported earlier sections both in underdamped and overdamped regimes . thus for large cycle time the reliability of the system working as an engine increases . persistence of these realizations even in quasistatic regime can be attributed to the fluctuation of heat and work distributions . fluctuations in work are mainly attributed to two adiabatic processes connecting two isotherms , while fluctuations of @xmath44 result from the relaxation of the system , when brought in direct contact with high temperature reservoir from low temperature bath . . we have shown that in tpss @xmath55 exhibit strong correlation between variables @xmath6 and @xmath7 in small cycle time limit . however , it becomes uncorrelated as we approach quasistatic limit . for analytical simplicity it had been generally assumed in earlier literature that there is no correlation between @xmath6 and @xmath7 in @xmath55 ( see for example @xcite ) . in the inertial regime we have also verified the recently proposed fluctuation theorems for heat engines in a tpss . our results are amenable to experimental verifications . a.m.j . thanks dst , india for financial support and . thanks mpipks , germany for partial support . 10 h. b. callen , thermodynamics and an introduction to thermostatistics ( john wiley & sons , 2006 ) . f. l. curzon and b. ahlborn , am . @xmath191 , 22 ( 1975 ) . u. seifert , rep . @xmath192 , 126001 ( 2012 ) and references therein . t. schmiedl and u. seifert , . c. van den broeck , . m. esposito , r. kawai , k. lindenberg and c. van den broeck , . s. sheng and z. c. tu , arxiv : 1404.7008 . v. blickle and c. bechinger , nature phys . * 8 * , 143 ( 2012 ) . c. bustamante , j. liphardt and f. ritort , phys . today @xmath193 , 43 ( 2005 ) . k. sekimoto , stochastic energetics ( springer , 2010 ) . k. sekimoto , prog . @xmath194 , 17 ( 1998 ) . d. dan and a. m. jayannavar , physica a. @xmath195 , , 404 ( 2005 ) . s. saikia , r. roy and a. m. jayannavar . a. @xmath196 , , 367 ( 2007 ) . p. jop , a. petrosyan and s. ciiberto . . j. hoppenau , m. niemann , and a. engel , . z. c. tu , v. holubec , arxiv : 1404.2030 . g. verley , t . willaert , c. van den broeck and m. esposito , arxiv:1404.3095 . n. a. sinitsyn . s. lahiri , s. rana , a. m. jayannavar , . m. campisi , . h. risken , the fokker - planck equation : methods of solution and applications ( springer 1996 ) . w. t. coffey , yu . p. kalmykov and j. t. waldorn , the langevin equatin ( world scientific 2004 ) . r. mannela , in : j.a . freund and t. poschel ( eds ) , stochastic process in physics , chemistry and biology , lecture notes in physics , vol . 557 springer - verlag , berlin(2000 ) p353 . e. boksenbojm , b. wynants and c. jarzynski , physica a. @xmath197 , 4406 ( 2010 ) . f. ritort , poincare seminar @xmath198 , 193 ( 2003 ) . g. m. wang , e. m. sevick , e. mittag , d. j. searles , and d. j. evans , . m. sahoo , s. lahiri , a. m. jayannavar , . u. seifert , . u. seifert , . s. lahiri and a. m. jayannavar a. saha , s. lahiri and a. m. jayannavar , .	we have performed an extensive analysis of a single particle stochastic heat engine constructed by manipulating a brownian particle in a time dependent harmonic potential . the cycle consists of two isothermal steps at different temperatures and two adiabatic steps similar to that of a carnot engine . the engine shows qualitative differences in inertial and overdamped regimes . all the thermodynamic quantities , including efficiency , exhibit strong fluctuations in a time periodic steady state . the fluctuations of stochastic efficiency dominate over the mean values even in the quasistatic regime . interestingly , our system acts as an engine provided the temperature difference between the two reservoirs is greater than a finite critical value which in turn depends on the cycle time and other system parameters . this is supported by our analytical results carried out in the quasistatic regime . our system works more reliably as an engine for large cycle times . by studying various model systems we observe that the operational characteristics are model dependent . our results clearly rules out any universal relation between efficiency at maximum power and temperature of the baths . we have also verified fluctuation relations for heat engines in time periodic steady state . [ 2 ] [ 3]phys . rev . lett . * # 1 * , # 2 ( # 3 ) [ 3]phys . rev . e * # 1 * , # 2 ( # 3 ) [ 3]phys . rev . a * # 1 * , # 2 ( # 3 ) [ 3]j . stat . mech . * # 1 * , # 2 ( # 3 ) [ 3]eur . phys . j. b * # 1 * , # 2 ( # 3 ) [ 3]rev . mod . phys . * # 1 * , # 2 ( # 3 ) [ 3]europhys . lett . * # 1 * , # 2 ( # 3 ) [ 3]j . stat . phys . * # 1 * , # 2 ( # 3 ) [ 3]prog . theor . phys . suppl . * # 1 * , # 2 ( # 3 ) [ 3]physics today * # 1 * , # 2 ( # 3 ) [ 3]adv . phys . * # 1 * , # 2 ( # 3 ) [ 3]j . phys . condens . matter * # 1 * , # 2 ( # 3 ) [ 3]j . phys . a * # 1 * , # 2 ( # 3 )
two - higgs doublet models ( 2hdm ) @xcite are one of the most popular extensions of the standard model ( sm ) , even without invoking supersymmetry , for which more than one scalar doublet is a necessary condition . in a 2hdm with two doublets @xmath1 and @xmath2 , there are five physical scalars : the two @xmath3-even neutrals @xmath4 and @xmath5 , the @xmath3-odd neutral @xmath6 , property if the scalar potential violates @xmath3 . for 2hdm with spontaneous @xmath3 violation , see , _ e.g. _ , ref.@xcite . ] and two charged scalars @xmath7 . while a generic 2hdm may contain flavor - changing neutral current ( fcnc ) interactions ( see , _ e.g. _ , ref . @xcite for such type of 2hdms ) , one usually invokes some discrete symmetry to banish such fcnc at the tree - level , based on the glashow - weinberg - paschos ( gwp ) @xcite theorem that there is no tree - level fcnc if all right - handed fermions of a given electric charge couple to only one of the doublets . it turns out that there are four 2hdms that satisfy the gwp criterion when a discrete @xmath0 symmetry is applied on the lagrangian . they are : 1 . type i , for which all fermions couple with @xmath2 and none with @xmath1 ; 2 . type ii , for which up - type quarks couple to @xmath2 , down - type quarks and charged leptons couple to @xmath1 ; 3 . type y ( sometimes called type iii or flipped ) , for which up - type quarks and charged leptons couple to @xmath2 and down - type quarks couple to @xmath1 ; 4 . type x ( sometimes called type iv or lepton - specific ) , for which all charged leptons couple to @xmath1 and all quarks couple to @xmath2 . among them , type ii 2hdm has been most widely investigated @xcite because the scalar sector of minimal supersymmetry is a type ii 2hdm . the tightest constraint on any 2hdm comes from the fact that the observed scalar resonance at the large hadron collider ( lhc ) can be identified with the sm higgs boson with @xmath8 gev @xcite . thus , the yukawa and gauge couplings of the 2hdm must be so aligned as to make the lighter @xmath3-even mass eigenstate @xmath4 to almost coincide with the sm higgs boson . this is known as the _ alignment limit _ , and while the allowed parameter space may vary from one 2hdm to the other , the qualitative results are quite similar @xcite . there are other constraints , like the oblique parameters or the lower limit on the @xmath7 mass coming from @xmath9 decay rate @xcite , but such constraints are in general not equally valid for all 2hdms . constraint , @xmath10 gev , can be evaded in type i and lepton - specific 2hdm . ] the theoretical constraints include , just like any other extensions of the sm , the vacuum stability , validity of the perturbative nature of the couplings , and constraints coming from the requirement of unitarity of scattering amplitudes @xcite . in this paper , we will focus upon the scalar potential of the 2hdms . this is much more complicated , even at the tree - level , compared to the sm scalar potential , if both the @xmath3-even neutral scalars are allowed to have nonzero vacuum expectation values ( vev ) . as has been shown in refs . @xcite , the 2hdm scalar potential , even with the softly broken @xmath0-symmetry , may allow more than one minima . a similar analysis was very recently done for @xmath0-breaking models @xcite . in refs . @xcite , the authors considered the scalar potential of a @xmath0-conserving 2hdm at the tree - level , and showed that it can allow multiple normal non - equivalent stationary points ( at most two of them can be minima ) . however , a charge - violating or @xmath3-violating minimum can not coexist with a normal minimum , _ i.e. _ , where the @xmath3-even neutral fields get the vev . this can be put in a more succinct way : minima with different natures can not co - exist in 2hdm @xcite . also , the data from lhc all but rules out those points where the second minimum is deeper than the electroweak ( ew ) minimum , _ i.e. _ , the minimum where the universe lives . in this paper , we consider both @xmath0-conserving and @xmath0-violating 2hdm scalar potentials for our analysis . by @xmath0-breaking 2hdm , we mean those with dimension-4 operators violating @xmath0 , like @xmath11 or @xmath12 . there can be soft @xmath0-violating terms like @xmath13 or @xmath14 in the potential ; if there are no dimension-4 @xmath0-breaking terms , we call those models @xmath0-conserving , although , strictly speaking , they are not . we show that the conclusions drawn about the nature of the minima of the potential in the context of @xmath0-conserving 2hdms remain valid for @xmath0-violating 2hdms too . to check the robustness of the tree - level results , we further perform a one - loop analysis of the 2hdm potential , and choose only those models that show a double minima . a nice review of the one - loop corrections in the context of @xmath0-symmetric 2hdm and scale invariant 2hdm can be found in refs . it is a common knowledge that the one - loop corrections @xcite can be significant only in the flat direction of the potential . for the sm , this is easy to obtain @xcite ; so is for the inert doublet models @xcite where one of the vevs is zero . for a generic 2hdm , this is a cumbersome task , but can be done , in principle , following the prescriptions of gildener and weinberg @xcite , and the ray in the potential space along which the tree - level potential is zero can be found . however , if @xmath15 , where @xmath16 and @xmath17 are the vevs of the @xmath3-even neutral components of @xmath1 and @xmath2 respectively ( so that @xmath18 ) , the potential along the @xmath1 direction is _ almost _ flat , so it is instructive to show the variations of the potential perpendicular to this direction , _ i.e. _ , along @xmath2 . another important point is the setting of the regularization scale @xmath19 for the one - loop corrections . variation of @xmath19 is equivalent to the variations of the tree - level quartic couplings @xmath20 , as can be seen from a renormalization group argument . as the nature of the potential can best be described by these quartic couplings , we would like to forward a prescription of choosing @xmath19 for the 2hdm : choose it so that the position of the ew minimum remains unaltered . as we will see , this keeps the position of the second minimum too almost unaltered . of course , the depths of the potential at the two minima will change . with such a prescription for choosing @xmath19 , the conclusions about the stability of the ew minimum that were drawn from a tree - level analysis remain unaltered . the paper is arranged as follows . in section ii we briefly review 2hdms , with softly broken @xmath0 symmetry and without @xmath0 symmetry , and list the constraints and the minimization conditions on the potential . section iii introduces the one - loop corrected effective potential and modified minimization conditions for type ii 2hdm with and without @xmath0 symmetry . our results , for both tree - level and one - loop analysis , are shown in section iv . section v summarizes the paper . some calculation details and relevant expressions are relegated to the appendix . to start with , let us focus on the most canonical 2hdms , where the @xmath0 symmetry is broken only softly by a dimension-2 operator , and all the couplings are real . the notations used here essentially follow those in ref . later on , we will introduce both dimension-4 @xmath0-breaking operators as well as complex parameters in the scalar potential . let us denote the two doublets , both with hypercharge @xmath21 , by @xmath1 and @xmath2 , which can be written more explicitly as _ a= _ a^+ ( _ a + i_a ) , a = 1,2 . we further assume the vevs to be aligned towards the direction of the cp - even neutral field , so that @xmath22 , @xmath23 , and we conventionally denote @xmath24 . invoking a @xmath0 symmetry @xmath25 , @xmath26 , so that there is no tree - level flavor - changing neutral current ( fcnc ) , one may write v ( _ 1 , _ 2)&= & m_11 ^ 2 _ 1^_1 + m_22 ^ 2 _ 2^_2 - m_12 ^ 2 ( _ 1^_2 + _ 2^_1 ) + & & + 12_1(_1^_1)^2 + 12_2(_2^_2)^2 + _ 3 ( _ 1^_1)(_2^_2 ) + & & + _ 4 ( _ 1^_2)(_2^_1)+ 12_5 . [ vtreecon ] here @xmath27 , @xmath28 , @xmath29 , @xmath30 , @xmath31 , @xmath32 , @xmath33 and @xmath34 are all real and @xmath29 softly breaks the @xmath0 symmetry . as mentioned before , to differentiate from the potential that breaks the @xmath0 symmetry with dimension-4 operators , such models will henceforth be called the @xmath0-conserving or @xmath0-symmetric models , even though it is broken softly by @xmath29 . the two cp - even neutral states @xmath35 and @xmath36 are in general not mass eigenstates . the corresponding mass matrix can be diagonalized through a rotation by an angle @xmath37 , and the mass eigenstates are h = _ 2- _ 1 , h = _ 2 + _ 1 , where @xmath38 is the lighter ( heavier ) eigenstate . note that if @xmath39 is an odd ( even ) multiple of @xmath40 , @xmath41 becomes identical with the sm higgs boson , with a vev of @xmath42 gev . for example , the @xmath43 ( @xmath44 ) coupling is just the sm coupling times @xmath45 ( @xmath46 ) , where @xmath47 is any weak gauge boson , @xmath48 or @xmath49 . the limit where @xmath4 behaves as the sm higgs boson is known as the alignment limit . the lhc data strongly favours the alignment limit and this sets a nontrivial constraint on the parameter space . the allowed parameter space , of course , depends on what type of 2hdm is chosen . we refer the reader to ref . @xcite for a study of the alignment limit in type i and type ii 2hdms , and to ref . @xcite for a typical example of constraints coming from 2hdms with tree - level fcnc . the most generic yukawa interactions for these four models can be written as @xcite , _ y = -_j=1 ^ 2 , where @xmath50 , @xmath51 , @xmath52 , @xmath53 , @xmath54 and @xmath55 are generic doublet quarks , doublet leptons , singlet down - type and singlet up - type quarks , and singlet charged leptons respectively . @xmath56 , @xmath57 , @xmath58 are @xmath59 complex matrices , containing yukawa couplings for the down , up , and leptonic sectors respectively . in our analysis we will consider only top , bottom , and @xmath60 yukawa couplings to be nonzero . the masses of the charged higgs , @xmath7 , and the pseudoscalar , @xmath6 , can be written as m_h^^2 & = & m_12 ^ 2 v^2 - 12 ( _ 4 + _ 5)v^2 , + m_a^2 & = & m_12 ^ 2 v^2 - _ 5 v^2 . [ masssq - z2 ] the couplings can be expressed in terms of masses of the physical states and the mixing angles @xmath37 and @xmath61 as @xcite _ 1 & = & ( c_^2 m_h^2 + s_^2 m_h^2 - ) , + _ 2 & = & ( s_^2 m_h^2 + c_^2 m_h^2 - ) , + _ 3 & = & + ( m_h^2 - m_h^2 ) - , + _ 4 & = & ( m_a^2 - 2m_h^^2 ) + , + _ 5 & = & - . where @xmath62 is the mass of @xmath38 , and @xmath63 and @xmath64 are generic shorthand notations for @xmath65 and @xmath66 respectively . the requirement that the scalar potential always remains bounded from below leads to the following stability conditions for 2hdms with @xmath0 symmetry @xcite , _ 1,_2 0 , _ 3- , _ 3 + _ 4 - _ 5- . [ stab - z2 ] note that @xmath32 , @xmath33 , and @xmath34 can in principle be negative ; however , if the @xmath0 symmetry is exact , this may lead to tachyonic masses for @xmath7 and @xmath6 . the most general renormalizable scalar potential of 2hdm can be written as @xcite v ( _ 1 , _ 2 ) & = & m_11 ^ 2 _ 1^_1 + m_22 ^ 2 _ 2^_2 - + & & + 12_1(_1^_1)^2 + 12_2(_2^_2)^2 + _ 3 ( _ 1^_1)(_2^_2 ) + _ 4 ( _ 1^_2)(_2^_1 ) + & & + , [ vtreebreak ] where the model parameters @xmath27 , @xmath28 , @xmath30 , @xmath31 , @xmath32 , @xmath33 are real and @xmath29 , @xmath34 , @xmath68 and @xmath69 can in principle be complex , and `` h.c . '' stands for hermitian conjugation . we will follow the same convention as the previous subsection and consider two different cases , namely , @xmath29 , @xmath34 , @xmath68 and @xmath69 are ( i ) real , and ( ii ) complex . for this case , the masses of @xmath7 and @xmath6 can be written in an analogous way of eq . ( [ masssq - z2 ] ) , using the minimization conditions of the potential : m_h^^2 & = & m_12 ^ 2 v^2 - 12 ( _ 4 + _ 5)v^2 - _ 6 v^2 - _ 7 v^2 , + m_a^2 & = & m_12 ^ 2 v^2 - _ 5 v^2 - _ 6 v^2 - _ 7 v^2 . [ massq - noz2-real ] the requirement that the scalar potential always remains bounded from below leads to the same set of equations as in ( [ stab - z2 ] ) , with an extra condition : & & _ 1,_2 0 , _ 3- , _ 3 + _ 4 - _ 5- , + & & 2_6 + _ 712(_1 + _ 2 ) + _ 3 + _ 4 + _ 5 . [ stab - noz2 - 1 ] the mixing angle between the cp - even neutral states is given by @xmath70 , where a & = & m_12 ^ 2 - ( _ 3 + _ 4 + _ 5 ) v_1 v_2 - 32 _ 6 v_1 ^ 2 - 32 _ 7 v_2 ^ 2 + b & = & - 12 _ 1 v_1 ^ 2 + 12 _ 2 v_2 ^ 2 + 12 m_12 ^ 2 ( v_1/v_2 - v_2/v_1 ) + 34(_7-_6 ) v_1 v_2 + _ 7 v_2 ^ 3 - _ 6 v_1 ^ 3 . the scalar potential of the 2hdm shows a much more complicated structure than that of the sm . the potential can have multiple non - equivalent normal stationary points ( only two of them can be minima ) @xcite , depending on the parameters . these minima can be all normal ( where only @xmath35 and @xmath36 get nonzero vev ) , charge - breaking ( where at least one of the charged fields @xmath71 gets a nonzero vev ) or cp violating ( where the two vevs have a nontrivial phase between them ) . it has been shown in refs . @xcite , that for the canonical @xmath0-symmetric 2hdm , ( i ) existence of a normal minimum rules out a charge breaking or cp violating minimum , those extrema can only be saddle points at best ; ( ii ) there can be more than one normal minima , only one of which corresponds to the standard model ( sm ) , which we call the ew minimum . if the ew minimum is the global one , we are in a stable situation ; if it is not , the universe may tunnel down to the deeper minimum if the tunneling time is less than or of the order of the lifetime of the universe . it has also been shown that the lhc data effectively rules out the parameter space where the ew minimum is shallower . we would like to extend this result to @xmath0-breaking 2hdm with both real and complex parameters , and also investigate the nature of the potential when one - loop corrections are taken into account . in a normal minimum , we may write _ 1_n = 0 v_1 , _ 2_n = 0 v_2 . in certain situations , several non - equivalent normal extrema are allowed by the minimization conditions of the potentials and at most two of them can be local minima . thus , there can be a second minimum with vevs @xmath72 , with @xmath73 gev , where exactly same symmetries are broken . obviously , even with the same parameters of the potential , the masses of all sm particles are going to change from the ew vacuum configuration . at the ew vacuum , the potential , as follows from eq.([vtreebreak ] ) , can be written as v_0 & = & 12 m_11 ^ 2 v_1 ^ 2 + 12 m_22 ^ 2 v_2 ^ 2 - m_12 ^ 2 v_1v_2 + 18 _ 1 v_1 ^ 4 + 18 _ 2 v_2 ^ 4 + 14 ( _ 3 + _ 4 + _ 5)v_1 ^ 2 v_2 ^ 2 + & & + 12 _ 6 v_1 ^ 3 v_2 + 12 _ 7 v_2 ^ 3 v_1 , [ sm - vac - pot ] and at the other minimum , we get the corresponding @xmath74 by replacing ( @xmath16,@xmath17 ) in eq . ( [ sm - vac - pot ] ) by @xmath72 . if @xmath75 , the ew ( second ) vacuum is stable and is the global minimum . the minimization conditions of the potential are easy to obtain : f_1(v_1,v_2 ) & & m_11 ^ 2v_1 - m_12 ^ 2v_2 + 12 _ 1v_1 ^ 3 + 12 ( _ 3 + _ 4 + _ 5)v_1v_2 ^ 2 + 32 _ 6 v_1 ^ 2 v_2 + 12 _ 7 v_2 ^ 3 = 0 , + f_2(v_1,v_2 ) & & m_22 ^ 2v_2 - m_12 ^ 2v_1 + 12 _ 2v_2 ^ 3 + 12 ( _ 3 + _ 4 + _ 5)v_2v_1 ^ 2 + 12 _ 6 v_1 ^ 3 + 32 _ 7 v_2 ^ 2 v_1 = 0 . [ potmin-1 ] with & > & 0 , + & > & 0 . if the symmetry breaks in such a way that one of the charged fields , say @xmath76 , gets a non - zero vev @xmath77 too , such that _ 1_cb = 0 v_1 , _ 2_cb = v_2 , the minimization conditions of the potential are & & m_11 ^ 2v_1 - m_12 ^ 2v_2 + 12 _ 1v_1 ^ 3 + 12 _ 3 v_1 ( v_2 ^ 2 + ^2 ) + 12 ( _ 4 + _ 5)v_1v_2 ^ 2 + 32 _ 6 v_1 ^ 2 v_2 + 12 _ 7(^2 + v_2 ^ 2)v_2 = 0 , + & & m_22 ^ 2v_2 - m_12 ^ 2v_1 + 12 _ 2 ( v_2 ^ 2+^2)v_2 + 12 ( _ 3 + _ 4 + _ 5)v_2v_1 ^ 2 + 12 _ 6 v_1 ^ 3 + 12 _ 7 ( ^2 + 3 v_2 ^ 2)v_1 = 0 , + & & m_22 ^ 2 + 12 _ 2 ( ^2 + v_2 ^ 2 ) + 12 _ 3 v_1 ^ 2 + _ 7 v_1 v_2 = 0 , [ potmin - cb ] and m_11 ^ 2 + 32 _ 1 v_1 ^ 2 + 12 _ 3(^2+v_2 ^ 2 ) + 12 ( _ 4 + _ 5)v_2 ^ 2 + 3 _ 6 v_1 v_2 & > & 0 , + m_22 ^ 2 + 12 _ 2 ( 3 v_2 ^ 2 + ^2 ) + 12 ( _ 3 + _ 4 + _ 5)v_1 ^ 2 + 3 _ 7 v_1 v_2 & > & 0 , + m_22 ^ 2 + 12 _ 2 ( 3 ^2 + v_2 ^ 2 ) + 14 _ 3 v_1 ^ 2 + _ 7 v_1 v_2 & > & 0 . [ potmin - cb2 ] at the cp violating extremum , one of the cp - odd neutral fields acquire a nonzero vev @xmath78 . in other words , the vevs develop a relative phase . thus , _ 1 = _ 1^+ ( ( _ 1 + v_1 ) + i(_1 + ) ) , _ 2= _ 2^+ ( ( _ 2 + v_2 ) + i_2 ) . we have similar minimization conditions for the potential : & & m_11 ^ 2v_1 - m_12 ^ 2v_2 + 12 _ 1(v_1 ^ 2 + ^2 ) v_1 + 12 ( _ 3 + _ 4 + _ 5)v_1v_2 ^ 2 + 12 _ 6(3v_1 ^ 2 + ^2)v_2 + 12 _ 7 v_2 ^ 3= 0 , + & & m_22 ^ 2v_2 - m_12 ^ 2v_1 + 12 _ 2 v_2 ^ 3 + 12 ( _ 3 + _ 4 ) v_2(v_1 ^ 2 + ^2 ) + 12 _ 5 ( v_1 ^ 2 -^2 ) v_2 + 12 _ 6 ( v_1 ^ 2 + ^2 ) v_1 + & & + 32 _ 7 v_1 v_2 ^ 2 = 0 , + & & m_11 ^ 2 + 12 _ 1 ( ^2 + v_1 ^ 2)+ 12 ( _ 3 + _ 4)v_2 ^ 2 - 12 _ 5 v_2 ^ 2 + _ 6 v_1 v_2 = 0 , and m_11 ^ 2 + 12 _ 1(3 v_1 ^ 2 + ^2 ) + 12 ( _ 3 + _ 4 + _ 5)v_2 ^ 2 + 3 _ 6 v_1 v_2 & > & 0 , + m_22 ^ 2 + 32 _ 2 v_2 ^ 2 + 12 ( _ 3 + _ 4 ) ( v_1 ^ 2 + ^2 ) + 12 _ 5 ( v_1 ^ 2 -^2 ) + 3 _ 7 v_1 v_2 & > & 0 , + m_11 ^ 2 + 12 _ 1 ( 3v_1 ^ 2 + ^2 ) + 12 ( _ 3 + _ 4 - _ 5)v_2 ^ 2 + _ 6 v_1 v_2 & > & 0 . next we consider the 2hdms with @xmath29 , @xmath34 , @xmath68 and @xmath69 complex , and use m_12 ^ 2 = m_12 ^ 2r + i m_12 ^ 2i , _ 5 = _ 51 + i _ 52 , _ 6= _ 61 + i _ 62 , _ 7 = _ 71 + i _ 72 . most of the relevant expressions are identical with the real parameter case discussed before , with the @xmath79 replacing @xmath80 respectively . while this applies to the expressions of physical mass eigenstates , there is an extra condition so that the charged goldstone boson is still massless : -m_12 ^ 2i + 12 _ 52 v_1 v_2 + 12 _ 62 v_1 ^ 2 + 12 _ 72 v_2 ^ 2 = 0 . [ zero - gold ] this shows that only three out of the four phases are independent . the conditions for the stability of the potential are & & _ 1,_2 0 , _ 3- , _ 3 + _ 4 - - , + & & 2_61 + _ 7112(_1 + _ 2 ) + _ 3 + _ 4 + _ 51 , 2_62 + _ 7212(_1 + _ 2 ) + _ 3 + _ 4 - _ 51 . similarly , the minimization conditions for the normal and charge - breaking minima are obtained from the corresponding expressions for the real parameter case by the substitutions mentioned before . for the charge - breaking minima , the imaginary parts of the potential parameters play a role . the minimization conditions are m_11 ^ 2v_1 - m_12 ^ 2r v_2 + 12 _ 1(v_1 ^ 2 + ^2 ) v_1 + 12 ( _ 3 + _ 4 + _ 51)v_1v_2 ^ 2 + 12 _ 52v_2 ^ 2 & & + + 12 _ 61(3v_1 ^ 2 + ^2)v_2 + _ 62 v_1 v_2 + 12 _ 71 v_2 ^ 3 & = & 0 , + m_22 ^ 2v_2 - m_12 ^ 2rv_1 - m_12 ^ 2i + 12 _ 2 v_2 ^ 3 + 12 ( _ 3 + _ 4 ) v_2(v_1 ^ 2 + ^2 ) + 12 _ 51 ( v_1 ^ 2 -^2 ) v_2 & & + + _ 52 v_1 v_2 + 12 _ 61 ( v_1 ^ 2 + ^2 ) v_1 + 12 _ 62 ( v_1 ^ 2 + ^2 ) + 32 _ 71 v_1 v_2 ^ 2 + 32 _ 72v_2 ^ 2 & = & 0 , + m_11 ^ 2 -m_12 ^ 2i v_2 + 12 _ 1 ( ^2 + v_1 ^ 2 ) + 12 ( _ 3 + _ 4)v_2 ^ 2 - 12 _ 51 v_2 ^ 2 + 12_52v_1 v_2 ^ 2 & & + + _ 61 v_1 v_2 + 12 _ 62(3 ^ 2 + v_1 ^ 2 ) v_2 + 12 _ 72v_2 ^ 3 & = & 0 , and m_11 ^ 2 + 12 _ 1(3 v_1 ^ 2 + ^2 ) + 12 ( _ 3 + _ 4 + _ 51)v_2 ^ 2 + 3 _ 61 v_1 v_2 + _ 62 v_2 & > & 0 , + m_22 ^ 2 + 32 _ 2 v_2 ^ 2 + 12 ( _ 3 + _ 4 ) ( v_1 ^ 2 + ^2 ) + 12 _ 51 ( v_1 ^ 2 -^2)+ _ 52v_1 + 3 _ 71 v_1 v_2 + 3 _ 72 v_2 & > & 0 , + m_11 ^ 2 + 12 _ 1 ( 3v_1 ^ 2 + ^2 ) + 12 ( _ 3 + _ 4 - _ 51)v_2 ^ 2 + _ 61 v_1 v_2 + 3_62 v_2 & > & 0 . the tree - level analysis will be performed using these expressions and the corresponding solutions of the simultaneous minimization equations . the form of the one - loop correction to the scalar potential is quite standard , and is given by m_i m_i(_c1,_c2 ) , v_1 = _ i = b , f n_i m_i^4 ( - c_i ) , [ 1loop - generic ] where the sum runs over all bosonic ( the physical scalars , the unphysical goldstone bosons , @xmath48 and @xmath49 ) and fermionic ( for our case , only @xmath81 , @xmath82 and @xmath60 ) degrees of freedom . here @xmath19 is the regularization scale . the masses are field - dependent quantities , being functions of the background fields @xmath83 and @xmath84 . they can be thought as the positions of the minima in the field space . in the limit @xmath85 , @xmath86 , the field - dependent masses are equal to the physical masses . the constants @xmath87 and @xmath88 are given by & & n_h = n_h = n_a = n_g^0 = 1 , n_h^ = n_g^ = 2 , + & & n_w = 6 , n_z = 3 , n_t = n_b = -12 , n_= -4 , + & & c_h = c_h = c_a = c_h^ = c_g^0 = c_g^ = c_t = c_b = c_= 32 , c_w = c_z = 56 . the full potential can be written as v = v(_1,_2 ) + v_1 [ v_eff ] where @xmath89 is the tree - level potential as given in eq . ( [ vtreecon ] ) or ( [ vtreebreak ] ) . at this point , let us again note that the masses in eq . ( [ 1loop - generic ] ) are functions of background fields @xmath83 and @xmath84 . the positions of the minima are functions of not only @xmath83 and @xmath84 but also the regularization scale @xmath19 . to have any idea of the nature of the potential after one - loop correction , one has to fix @xmath19 by some prescription . the renormalization group improved one - loop potential clearly shows that the variation in @xmath19 is equivalent to the redefinition of the coupling parameters of the theory . one popular way is to fix it in such a way that the one - loop corrections are minimum , hoping that this will minimize the higher - order corrections too . to be physically more transparent , we try a different approach : we fix @xmath19 in such a way that the position of the ew minimum remains ( almost ) unchanged with respect to the tree - level position . this will keep the one - loop corrected field - dependent masses to be at the same values of the tree - level masses at the ew vacuum . only the depth of the potential changes by the one - loop correction . thus , our prescription is to tune @xmath19 in such a way that @xmath85 and @xmath86 ( so that , for example , the goldstone bosons are still massless ) . as we will show later , if one tunes @xmath19 so that the position of the ew minimum is unchanged , it will keep the position of the second minimum almost unchanged too . as a concrete example , let us now discuss the type ii 2hdm ; as the yukawa couplings enter the picture , one needs to specify the type of 2hdm under consideration . we have checked that the qualitative features remain unchanged in all other 2hdms . to keep the discussion as much general as possible , let us focus on the one - loop correction to the 2hdm ( without @xmath0 symmetry ) tree level potential . we can obtain the same for @xmath0 symmetric 2hdm by putting @xmath68 and @xmath69 to be equal to zero , and by making @xmath34 and @xmath29 real . using the definitions of @xmath90 from eq . ( [ potmin-1 ] ) , we can write the modified minimization condition for one - loop corrected potential @xmath47 as f_1(v_1,v_2 ) + . _ _ c1=v_1 = 0 , f_2(v_1,v_2 ) + . _ _ c2=v_2 = 0 , [ mod - min ] where , .__c1=v_1 & = & [ mod - min-1 ] and .__c2=v_2 & = & , [ mod - min-2 ] with @xmath91 and @xmath92 , @xmath93 and @xmath94 being the @xmath95 and @xmath96 gauge couplings . yukawa couplings for @xmath81 , @xmath82 , and @xmath60 are denoted by @xmath97 , @xmath98 and @xmath99 respectively . the @xmath100-functions have been defined in appendix a. note that these expressions are valid only if we tune @xmath19 to keep the position of the ew minimum of the one - loop corrected effective potential @xmath47 , defined in eq . ( [ v_eff ] ) , unchanged with respect to its position in the tree level potential . one also needs to check the second derivatives to ensure that the extremum is a local minimum : . _ _ c1=v_1 > 0 , . _ _ c2=v_2 > 0 . [ mod - min-3 ] the expressions for second derivatives are also given in appendix a. note the absence of the goldstone bosons in eqs . ( [ mod - min],[mod - min-1],[mod - min-2 ] ) because of our choice of @xmath19 which keeps them massless . from a random scan over the parameter space spanning over @xmath101 different choices of model parameters , we generate a number of models for which the following conditions are satisfied . our analysis includes the canonical 2hdm with @xmath0 symmetry , and 2hdm without @xmath0 with both real and complex parameters . * the potential has to be stable at all scales before it either blows up ( due to one or more couplings hitting the landau pole ) or becomes unbounded from below . * the dimensionless couplings must remain perturbative over the entire range of validity of the theory , except maybe at the very end where they approach the landau pole . * there should be one minimum of the scalar potential for which @xmath102 gev . this we will call the ew vacuum . this , in fact , acts as the tightest constraint and rules out the largest chunk of randomly generated models . we focus only on those models that allow another local minimum apart from the ew vacuum . both tree - level minimization conditions , for @xmath16 and @xmath17 , should be satisfied in both the vacua . * other constraints like that on the charged higgs boson mass coming from @xmath9 are satisfied . we use @xmath10 gev . while not all constraints are valid for all 2hdms , we focus only on the type ii 2hdm . * the 125 gev resonance found at the large hadron collider must have properties close to that of the sm higgs boson . in other words , the alignment limit should be maintained . we have kept @xmath103 to be between @xmath104 and @xmath105 , noting that this is a rather conservative limit . the ranges for our scan is as follows : 0.0 _ 1 , _ 2 1.0 , -1.0 _ 3 , _ 4 , _ 51.0 , m^2_12 410 ^ 6 gev^2 , 1.0 50.0 , where the parameters are taken to be at the ew scale . apart from this , for the @xmath0-violating cases , we have taken -1.0 _ 6 , _ 7 1.0 ( real couplings ) , -1.0 _ 51 , _ 52 , _ 61 , _ 62 , _ 71 , _ 72 1.0 ( complex couplings ) . also , for this case the scan is made on @xmath106 , as the phase in @xmath29 is fixed by eq . ( [ zero - gold ] ) . we find a few common characteristics for these models that allow two minima . they are : + ( i ) the ew vacuum is always deeper than the other vacuum . this happens mostly because of the imposition of the experimental constraints . thus , even with the introduction of @xmath0-breaking parameters , there is no chance of tunneling to the other minimum , at least at the tree - level . this reinforces the conclusions obtained by the authors of ref . + ( ii ) if both the vacua are normal , there is no other minimum that breaks charge conservation or @xmath3 . this , again , is in tune of what the authors of refs . @xcite found . we would now like to perform the same analysis on the one - loop corrected potential on those models that satisfy the initial constraints and show the presence of a double minima at the tree - level . the regularization scale @xmath19 is so chosen as to make @xmath85 and @xmath86 . this is , of course , a rather restrictive choice , but keeps all the masses as well as @xmath107 unchanged at the ew minimum even after the one - loop corrections are implemented . in a generic 2hdm where both cp - even neutral fields can get nonzero vev , this is a tricky job , and mathematically much more complicated than the cases of only sm , or sm extended by a gauge singlet scalar , or the inert doublet models where one vev is always zero . the complication is further enhanced by the fact that at the tree - level , there are two minima of the potential . what we do is the following . we fix @xmath85 and tune @xmath19 in such a way that @xmath84 coincides with @xmath17 . this keeps the position of the ew minimum invariant but changes the depth . we could have done this the other way round too , namely , keeping @xmath86 and adjusting @xmath19 to make @xmath83 coincide with the tree level value ; however , we prefer the first approach as the contribution of @xmath17 is larger in @xmath108 for @xmath109 . we then use the same @xmath19 but fix the classical minimum for @xmath35 at @xmath110 . it so happens that @xmath111 , in all cases , lies close to @xmath112 ; the coincidence is not as exact as the ew vacuum , but this being the shallower secondary minimum , we will not be so much bothered about its exact position . thus , we compare @xmath113 with @xmath114 . we will also show the effect of varying the scale @xmath19 on the classical field values . what we observe is that the conclusions drawn from a tree - level analysis is more or less unchanged ; the deeper minimum remains deeper . this is not entirely unexpected , as the one - loop corrections are only a small effect that can not overcome the difference in depths of these two minima . for @xmath115 , @xmath35 direction is almost flat , so we show our results in the constant-@xmath35 direction , varying the potential @xmath47 with @xmath36 . our results are discussed in more detail later for some typical benchmark points , but for both the cases that we consider ( @xmath0 conserving and breaking ) , we never found the second minimum becoming deeper than the ew one after the one - loop corrections . thus , if the ew vacuum is the deeper one at tree - level , it remains so after the radiative corrections ; there is no chance of developing a deeper vacuum and tunnelling into it . in fact , if @xmath19 is tuned in such a way that the ew vacuum is not shifted from its tree - level position , _ it always gets deeper by the one - loop corrections _ ; @xmath116 at the ew vacuum is always negative . in table [ tab : tab - bench ] , we show three benchmark points for the @xmath0-conserving type - ii 2hdm , characterized by small , medium , and large values of @xmath117 respectively . all these models show the existence of a second and shallower minimum compared to the ew one . these three models , namely , @xmath118 , @xmath119 , and @xmath120 are valid up to @xmath121 gev , @xmath122 gev , and the planck scale respectively ; for the first two models , at least one of the couplings become nonperturbative at the validity scale and the model soon hits the landau pole thereafter . obviously , the comparatively low validity range for the first benchmark can be ascribed to the relatively large quadratic couplings to start with at the ew scale . the evolutions are checked with one - loop renormalization group equations for the type ii 2hdm @xcite . .benchmark points for @xmath0-conserving type - ii 2hdm . [ cols="^,^,^,^ " , ] ( l ) and @xmath111 ( r ) with @xmath19 for @xmath123 ( blue ) , @xmath124 ( red ) , and @xmath125 ( golden).,title="fig:",width=264,height=188 ] ( l ) and @xmath111 ( r ) with @xmath19 for @xmath123 ( blue ) , @xmath124 ( red ) , and @xmath125 ( golden).,title="fig:",width=264,height=188 ] just like the @xmath0-conserving case , we choose three benchmarks for the case where the potential violates the @xmath0 and also involves complex parameters . these benchmarks are shown in table [ tab : tab - bench2 ] . the flatness of the potential in the @xmath35 direction is again most manifest for @xmath125 , the benchmark with lowest @xmath16 . at the same time , the ew minimum always gets deeper with the one - loop corrected potential , and there is no qualitative change from the tree - level result . because of the comparatively large values of the couplings , all the three models tend to hit the landau pole much before the planck scale , namely , at 45.8 tev , 4.8 tev , and 216.7 tev . this is expected because @xmath124 starts with larger values of the quartic couplings at the ew scale . in general , for the double - minima case when @xmath0 is broken , most of the couplings at the ew scale have to be large to start with and thus such class of models are not stable beyond a few hundreds of tev , which may be contrasted with the @xmath0-conserving double - minima models . in fig . [ fig : varz2asymm ] , we show how @xmath84 and @xmath111 changes with @xmath19 for the three benchmarks . in this paper we have tried to investigate the nature of the potential of type ii 2hdm , breaking the @xmath0 symmetry either softly or through dimension-4 operators . there are some known results for the @xmath0-symmetric 2hdm at the tree level . our goal was to investigate how far these conslusions are reliable if one ( i ) breaks the @xmath0 symmetry at the tree level with operators with real or complex couplings , ( ii ) does a one - loop correction on the potential . for the first part , we find that the introduction of @xmath0 breaking does not change the conclusions qualitatively : the scalar potential can accommodate at most two local minima , both of which have to be normal . if there is a normal minimum , there can not be a charge - breaking or @xmath3 violating minimum of the potential . the lhc data highly disfavours those models where the second minimum is deeper than the ew minimum , possibly making the ew vacuum an unstable or metastable one . for one - loop corrections , we use a regularization scale that keeps the position of the ew minimum invariant , changing only its depth . when we focus on models with two minima , this prescription keeps the position of the second minima almost unchanged too . the one - loop corrections can not change the relative depths of the minima , _ i.e. _ , the ew minimum still remains deeper after the correction , ruling out the possibility of a metastable vacuum . the conclusions are identical for @xmath0 symmetric and @xmath0 breaking 2hdm . while the conclusions were drawn for type ii 2hdm , the results are qualitatively the same for other 2hdms too , as the only change comes from the yukawa couplings that enter the one - loop corrections . the stability of the 2hdm at higher energy scales is a complex issue because of more fields and couplings . one needs the stability conditions to be valid at all scales and the couplings to remain perturbative for calculability ( or , at least , not blow up ) . a general tendency that can intuitively be deduced from the rg equations is that higher values of quartic couplings at the ew scale tend to pull down the range of validity of the theory , which means that some other ultraviolet complete theory takes hold beyond that range . however , a large part of the parameter space is still compatible with the stability up to the planck scale for the @xmath0-conserving class of models . for the @xmath0-violating class , the existence of two minima generally forces some of the quartic couplings to be large at the ew scale and hence such models cease to be valid beyond a few hundreds of tev at the most ; for smaller couplings , one gets the single - minimum models . * acknowledgements * the authors acknowledge nabarun chakrabarty and dipankar das for several helpful discussions . i.c . acknowledges the council for scientific and industrial research , government of india , for a research fellowship . a.k . acknowledges the department of science and technology , government of india , and the council for scientific and industrial research , government of india , for support through research grants we use the following shorthand notations : f(,i ) = . ( ) _ _ ci = v_i , g(,i ) = _ _ ci = v_i . f(h,1 ) & = & 14(2_1 v_1 + ( v_1 ^ 2-v_2 ^ 2)++ - ) + & & -(_1 v_1 ^ 2 - _ 2 v_2 ^ 2 - ( v_1 ^ 2-v_2 ^ 2)+ -+ ) + & & ( _ 1 v_1 - + ( v_1 ^ 2- v_2 ^ 2)+++ ) + & & -(-m_12 ^ 2r + _ 345 v_1 v_2++ ) ( _ 345 v_2 + 3_61v_1 ) , + f(h,2 ) & = & 14(2_2 v_2 + ( v_2 ^ 2-v_1 ^ 2)+- + ) + & & -(_1 v_1 ^ 2 - _ 2 v_2 ^ 2 - ( v_1 ^ 2-v_2 ^ 2)+ -+ ) + & & ( -_2 v_2 + + ( v_1 ^ 2- v_2 ^ 2)+-- ) + & & -(-m_12 ^ 2r + _ 345 v_1 v_2++)(_345 v_1 + 3_71v_2 ) , + f(h,1 ) & = & 14(2_1 v_1 + ( v_1 ^ 2-v_2 ^ 2)++ - ) + & & + ( _ 1 v_1 ^ 2 - _ 2 v_2 ^ 2 - ( v_1 ^ 2-v_2 ^ 2)+ -+ ) + & & ( _ 1 v_1 - + ( v_1 ^ 2- v_2 ^ 2)+++ ) + & & + ( -m_12 ^ 2r + _ 345 v_1 v_2++)(_345 v_2 + 3_61v_1 ) , + f(h,2 ) & = & 14(2_2 v_2 + ( v_2 ^ 2-v_1 ^ 2)+- + ) + & & + ( _ 1 v_1 ^ 2 - _ 2 v_2 ^ 2 - ( v_1 ^ 2-v_2 ^ 2)+ -+ ) + & & ( -_2 v_2 + + ( v_1 ^ 2- v_2 ^ 2)+-- ) + & & + ( -m_12 ^ 2r + _ 345 v_1 v_2++)(_345 v_1 + 3_71v_2 ) , + where _ 67&= & _ 61 + _ 71 , _ 345 = _ 3 + _ 4 + _ 51 , + a & = & ( + -+- ) ^2 + & & + ( -m_12 ^ 2r+_345v_1v_2++)^2 . . _ _ c1=v_1 & = & [ 4 m_a^2 ( - _ 51- + - ) ( - 1 ) . + & & + 8 ( ( 1-)- v_1 _ 51 - -- + ) ^2 + & & + 8 m_h^^2 ( - 12_45- + - ) ( - 1 ) + & & + 4 ( ( 1-)-v_1 _ 45 - --_71v_2+)^2 + & & -6 m_w^2 g_2 ^ 2 ( 13 - ) + g_2 ^ 4 v_1 ^ 2(2 + 3 ) - 3 m_z^2 g^2 ( 13 - ) + & & + ( 2 + 3 ) + 24 m_b^2 y_b^2 ( 1 - ) - 24 v_1 ^ 2 y_b^4 + & & + 8 m_^2 y_^2 ( 1 - ) - 8 v_1 ^ 2 y_^4 + 8 ^2 + & & . + 4 m_h^2 g(h,1 ) ( -1 ) + 8 ^2 + 4 m_h^2 g(h,1 ) ( -1 ) ] , + .__c2=v_2&= & . + g. c. branco , p. m. ferreira , l. lavoura , m. n. rebelo , m. sher and j. p. silva , `` theory and phenomenology of two - higgs - doublet models , '' phys . rept . * 516 * , 1 ( 2012 ) [ arxiv:1106.0034 [ hep - ph ] ] . for a recent review , see g. bhattacharyya and d. das , `` scalar sector of two - higgs - doublet models : a mini - review , '' arxiv:1507.06424 [ hep - ph ] . g. c. branco and m. n. rebelo , `` the higgs mass in a model with two scalar doublets and spontaneous cp violation , '' phys . b * 160 * , 117 ( 1985 ) . g. c. branco , w. grimus and l. lavoura , `` relating the scalar flavor changing neutral couplings to the ckm matrix , '' phys . lett . b * 380 * , 119 ( 1996 ) [ hep - ph/9601383 ] . f. j. botella , g. c. branco , a. carmona , m. nebot , l. pedro and m. n. rebelo , `` physical constraints on a class of two - higgs doublet models with fcnc at tree level , '' arxiv:1401.6147 [ hep - ph ] . g. bhattacharyya , d. das and a. kundu , `` feasibility of light scalars in a class of two - higgs - doublet models and their decay signatures , '' phys . d * 89 * , 095029 ( 2014 ) [ arxiv:1402.0364 [ hep - ph ] ] . s. l. glashow and s. weinberg , `` natural conservation laws for neutral currents , '' phys . d * 15 * , 1958 ( 1977 ) . + e. a. paschos , `` diagonal neutral currents , '' phys . d * 15 * , 1966 ( 1977 ) . h. s. cheon and s. k. kang , `` constraining parameter space in type - ii two - higgs doublet model in light of a 126 gev higgs boson , '' jhep * 1309 * , 085 ( 2013 ) [ arxiv:1207.1083 [ hep - ph ] ] . b. coleppa , f. kling and s. su , `` constraining type ii 2hdm in light of lhc higgs searches , '' jhep * 1401 * , 161 ( 2014 ) [ arxiv:1305.0002 [ hep - ph ] ] . o. eberhardt , u. nierste and m. wiebusch , `` status of the two - higgs - doublet model of type ii , '' jhep * 1307 * , 118 ( 2013 ) [ arxiv:1305.1649 [ hep - ph ] ] . o. eberhardt , `` fitting the two - higgs - doublet model of type ii , '' arxiv:1405.3181 [ hep - ph ] . n. chakrabarty , u. k. dey and b. mukhopadhyaya , `` high - scale validity of a two - higgs doublet scenario : a study including lhc data , '' jhep * 1412 * , 166 ( 2014 ) [ arxiv:1407.2145 [ hep - ph ] ] . g. aad _ et al . _ [ atlas and cms collaborations ] , `` combined measurement of the higgs boson mass in @xmath126 collisions at @xmath127 and 8 tev with the atlas and cms experiments , '' phys . * 114 * , 191803 ( 2015 ) [ arxiv:1503.07589 [ hep - ex ] ] . d. chowdhury and o. eberhardt , `` global fits of the two - loop renormalized two - higgs - doublet model with soft @xmath0 breaking , '' arxiv:1503.08216 [ hep - ph ] . j. bernon , j. f. gunion , h. e. haber , y. jiang and s. kraml , `` scrutinizing the alignment limit in two - higgs - doublet models . part 1 : @xmath128 gev , '' arxiv:1507.00933 [ hep - ph ] . d. das and i. saha , `` search for a stable alignment limit in two - higgs - doublet models , '' phys . d * 91 * , no . 9 , 095024 ( 2015 ) [ arxiv:1503.02135 [ hep - ph ] ] . f. mahmoudi and o. stal , `` flavor constraints on the two - higgs - doublet model with general yukawa couplings , '' phys . d * 81 * , 035016 ( 2010 ) [ arxiv:0907.1791 [ hep - ph ] ] . o. deschamps , s. descotes - genon , s. monteil , v. niess , s. tjampens and v. tisserand , `` the two higgs doublet of type ii facing flavour physics data , '' phys . d * 82 * , 073012 ( 2010 ) [ arxiv:0907.5135 [ hep - ph ] ] . b. gorczyca and m. krawczyk , `` tree - level unitarity constraints for the sm - like 2hdm , '' arxiv:1112.5086 [ hep - ph ] . a. barroso , p. m. ferreira , i. p. ivanov and r. santos , `` metastability bounds on the two higgs doublet model , '' jhep * 1306 * , 045 ( 2013 ) [ arxiv:1303.5098 [ hep - ph ] ] . a. barroso , p. m. ferreira , i. ivanov and r. santos , `` tree - level metastability bounds in two - higgs doublet models , '' arxiv:1305.1235 [ hep - ph ] . p. m. ferreira , r. santos and a. barroso , `` stability of the tree - level vacuum in two higgs doublet models against charge or cp spontaneous violation , '' phys . b * 603 * , 219 ( 2004 ) [ phys . b * 629 * , 114 ( 2005 ) ] [ hep - ph/0406231 ] . i. p. ivanov , `` minkowski space structure of the higgs potential in 2hdm , '' phys . d * 75 * , 035001 ( 2007 ) [ phys . d * 76 * , 039902 ( 2007 ) ] [ hep - ph/0609018 ] . i. p. ivanov , `` minkowski space structure of the higgs potential in 2hdm . ii . minima , symmetries , and topology , '' phys . d * 77 * , 015017 ( 2008 ) [ arxiv:0710.3490 [ hep - ph ] ] . i. p. ivanov and j. p. silva , `` tree - level metastability bounds for the most general two higgs doublet model , '' arxiv:1507.05100 [ hep - ph ] . m. sher , `` electroweak higgs potentials and vacuum stability , '' phys . rept . * 179 * , 273 ( 1989 ) . m. sher , `` the coleman - weinberg phase transition in extended higgs models , '' phys . d * 54 * , 7071 ( 1996 ) [ hep - ph/9607337 ] . j. s. lee and a. pilaftsis , `` radiative corrections to scalar masses and mixing in a scale invariant two higgs doublet model , '' phys . d * 86 * , 035004 ( 2012 ) [ arxiv:1201.4891 [ hep - ph ] ] . s. r. coleman and e. j. weinberg , `` radiative corrections as the origin of spontaneous symmetry breaking , '' phys . d * 7 * , 1888 ( 1973 ) . m. quiros , `` finite temperature field theory and phase transitions , '' hep - ph/9901312 . l. fromme , s. j. huber and m. seniuch , jhep * 0611 * , 038 ( 2006 ) [ hep - ph/0605242 ] ; + a. tranberg and b. wu , jhep * 1301 * , 046 ( 2013 ) [ arxiv:1210.1779 [ hep - ph ] ] . n. khan and s. rakshit , `` constraints on inert dark matter from metastability of electroweak vacuum , '' arxiv:1503.03085 [ hep - ph ] . b. swiezewska , `` inert scalars and vacuum metastability around the electroweak scale , '' jhep * 1507 * , 118 ( 2015 ) [ arxiv:1503.07078 [ hep - ph ] ] . e. gildener and s. weinberg , `` symmetry breaking and scalar bosons , '' phys . d * 13 * , 3333 ( 1976 ) . s. p. martin , `` taming the goldstone contributions to the effective potential , '' phys . d * 90 * , no . 1 , 016013 ( 2014 ) [ arxiv:1406.2355 [ hep - ph ] ] . i. chakraborty and a. kundu , `` two - higgs doublet models confront the naturalness problem , '' phys . d * 90 * , no . 11 , 115017 ( 2014 ) [ arxiv:1404.3038 [ hep - ph ] ] .	we perform a detailed analysis of the two - higgs doublet model ( 2hdm ) potential . at the tree - level , the potential may accommodate more than one minima , one of them being the electroweak ( ew ) minimum where the universe lives . the parameter space allowed after the data from the large hadron collider ( lhc ) came in almost excludes those cases where the ew vacuum is shallower than the second minimum . we extend the analysis by including terms in the 2hdm potential that break the @xmath0 symmetry of the potential by dimension-4 operators and show that the conclusions remain unchanged . furthermore , a one - loop analysis of the potential is performed for both cases , namely , where the @xmath0 symmetry of the potential is broken by dimension-2 or dimension-4 operators . for quantitative analysis , we show our results for the type - ii 2hdm , qualitative results remaining the same for other 2hdms . we find that the nature of the vacua from the tree - level analysis does not change ; the ew vacuum still remains deeper . * on the scalar potential of two - higgs doublet models * indrani chakraborty and anirban kundu + _ department of physics , university of calcutta , + 92 acharya prafulla chandra road , kolkata 700009 , india _ pacs no . : 12.60.fr , 14.80.ec
this study is part a series to quantify the degree and evolution of substructure within galaxies . here we introduce the methods and use the lmc as a test case . the distribution of young stars in the lmc was first studied with objective algorithms by feitzinger & braunsfurth ( 1984 ) , who found that the distribution was not random but had distinct patterns on all of the spatial scales probed ( 60 pc to 8 kpc ) . other studies have come to similar conclusions , that young stars are distributed hierarchically within galaxies ( e.g. efremov 1984 , battinelli , efremov , & magnier 1996 , elmegreen & salzer 1999 , elmegreen & elmegreen 2003 , elmegreen et al . 2006 , bastian et al . 2005 , 2007 , snchez et al . this pattern observed in the young stellar distribution is thought to originate from the distribution and fragmentation of gas within galaxies ( e.g. elmegreen & efremov 1996 ; elmegreen & falgarone 1996 ) . it has been noted that many nearby star forming clusters also have hierarchical structure seemingly dictated by the structure of the dense gas of natal molecular clouds ( e.g. lada & lada 2003 ) . using statistical techniques , gutermuth et al . ( 2005 ) studied three galactic clusters of varying degrees of embeddedness and demonstrated that the least embedded cluster was also the least dense and the least substructured of the three . that result hinted at the idea that the youngest clusters are substructured , but that dynamical interactions and ejection of the structured gas contributes to the evolution and eventual erasure of that substructure in approximately the cluster formation timescale of a few myr ( palla & stahler 2000 ) . by having an accurate model of the spatial evolution of stellar structures we can approach a number of fundamental questions about star - formation , including what is the percentage of stars born in `` clusters '' and whether this depends on environmental conditions . using automated algorithms on infrared spitzer surveys of star - forming sites within the galaxy ( e.g. allen et al . 2007 ) , such constraints are now becoming possible . in addition to constraining models of star - formation on galaxy and cluster scales , the spatial distribution of stars can have important implications on a variety of other astrophysical problems . for example , the mixture of gas and stars can affect the temperature structure within hii regions which in turn can influence metallicity determinations based on emission line spectra ( ercolano , bastian , stasinska 2007 ) . thus , a small dependence on clustering properties of young stars on galactocentric distance ( i.e. if there is a slight preference to form more concentrated stellar groups in the central parts of galaxies compared to the outer regions as may be expected if gas pressure plays a dominant role in cluster formation ) could artificially enhance the metallicity gradient observed in spiral galaxies . previous works on stellar structures have mainly concentrated on the youngest stellar objects ( ob stars ) , while the evolution of these objects have remained largely unstudied . efremov ( 1995 ) studied the distribution of cepheids , whose ages are significantly higher than ob stars , and found evidence of huge stellar complexes which he suggested may be made up of dissolved aged associations . zaritsky et al . ( 2004 ) showed the stark contrast between the spatial distribution of young and old stellar populations in the lmc using two colour - magnitude cuts through the magellanic clouds photometric survey ( mcps ) . the goal of this paper is to quantify the timescale over which stellar structures within the lmc evolve . to this end , we use two methods to estimate the evolution and analyse each method s strengths and weaknesses . our primary dataset comes from the mcps where we make cuts in colour and magnitude space in order to select differing stellar populations ( quantifying the age through comparison with stellar isochrones with a given initial mass function ) which we compliment with catalogues of stellar clusters ( hunter et al . 2003 ) and embedded young stellar objects ( whitney et al . 2008 ) . the paper is organised in the following way . in [ sec : obs ] we introduce the observations and the methods used , while in [ sec : size ] we compare the properties ( size and luminosity distributions ) of the young star - forming groups between the lmc and m33 . in [ sec : subsamples ] we extract twelve samples of stars of varying mean ages in order to compare their clustering properties , which is carried out in [ sec : spatial ] . in [ sec : comparison ] we discuss the implications of our results and present a series of simple models to aid in the interpretation , and in [ sec : conclusions ] we present our conclusions . our primary set of data comes from the magellanic cloud photometry survey of zaritsky et al . this dataset consists of u , b , v , and i band imaging of essentially the entire optical extent of large magellanic cloud . the original list consists of over 24 million sources , far too many for our algorithm , so we restrict our study to various cuts in colour and magnitude space . the method that we employ to first study the distribution of star - forming sites is the same as was presented in bastian et al . ( 2007 ) for the study of the spatial distribution of ob stars in m33 . we refer the reader to that work for details and only provide a summary of the method here . using the photometry of individual stars we make colour and magnitude cuts in order to select young stars ( see [ sec : size ] for details of the cuts ) . we then construct a minimum spanning tree ( mst ) of the spatial positions of stars which pass our criteria . the tree is then fractured ( fmst ) by applying a breaking distance , d@xmath6 , in which all edges of the tree , longer than d@xmath6 , are broken ( i.e. removed ) , creating independent groups . in order to avoid a selection bias which has hampered previous studies , we apply multiple breaking distances ( 20 in all ) from pc to kpc scales . throughout this paper , unless otherwise noted , we have used a value of five for n@xmath7 , which is the minimum number of sources used to define a group . in [ sec : subsamples ] we extend our study by not only looking at the youngest stars , but also the distribution of stars of various ages . throughout this study we have assumed a distance to the lmc of 50 kpc ( e.g. marconi & clementini 2005 ) . for comparison with models ( both fractal and uniform distributions ) we will assume that the lmc is largely 2d - i.e. that it is a disk - like face on galaxy . this assumption will be justified in [ sec : q ] . as a first step towards understanding the spatial distribution of star - forming regions in the lmc we perform a similar study as was carried out on m33 ( bastian et al . we have used cuts in colour and magnitude space ( namely @xmath8 and @xmath9 - uncorrected for extinction ) in order to select young ( massive ) stars , resulting in 5093 stars which pass these criteria . the sample is then analysed using the fractured minimum spanning tree method , selecting 20 breaking distances between the effective resolution limit of 1.5 pc and 1 kpc ( with equal steps in logarithmic space of 0.1485 ) . we refer the reader to bastian et al . ( 2007 ) for details on the method . this method produces a catalogue of groups and their properties , namely size and luminosity , for each breaking distance applied . we also combined these catalogues into a total sample from which we removed groups found by multiple breaking radii ( i.e. duplicate groups that had the exact same position and radius ) only retaining the groups found by the smallest breaking radius . we note that in the individual catalogues we did not remove duplicates ( i.e. groups that were found by other d@xmath6 ) , as each breaking distance was treated seperately . we first look at the size distribution of the groups found . in bastian et al . ( 2007 ) we found that for m33 there was not a characteristic size for star - forming regions , contrary to what is often quoted in the literature . we showed that previous claims of a characteristic size of @xmath10 pc for ob - associations ( e.g. bresolin et al . 1998 ) was due to the selection of a single breaking distance or scale in which to carry out the analysis . we carried out a similar analysis for the lmc , and come to similar conclusions . figure [ fig : sizefunc ] shows the cumulative size distribution of groups found with various breaking distances and for the total sample . the dashed ( red ) lines show the best fitting log - normal function to each distribution , which provides a good fit for each breaking distance . the total sample displays an excess tail at large radii ( @xmath11 pc ) relative to the best fitting analytical fit , however it is well fit up until this point . this break is similar , although at slightly smaller scales ( 400 pc vs. 1 kpc ) to that seen in the autocorrelation function of young stars in the lmc ( odekon 2008 , her fig . it is also similar to the scale where a break in the hi distribution is observed , @xmath12 pc ( padoan et al . 2001 ; elmegreen et al . 2001 ; kim & park 2007 ) . this feature , seen in the hi and in the young stars , has been suggested to relate to the scale height of the galaxy , where structures begin to change from largely 3d to 2d objects . we note that odekon ( 2008 ) also found a break in the autocorrelation function for young stars in m33 , whereas no similar scale was identified in the study of bastian et al . ( 2007 ) . we confirm the results found for m33 , that the turn - over in the size - distribution depends on the breaking distance used in an almost one - to - one way , suggesting that previous reports of a characteristic size in star - forming regions were due to the selection of a single breaking distance . hence , we confirm that the lmc does not display evidence for a characteristic scale in the star - formation process . similar results were found for the spiral galaxy ngc 628 ( elmegreen et al . 2006 ) . the luminosity of each group was found by summing the v - band luminosity of all the stars found to be part of the group . for simplicity , we just concentrate on the total sample ( i.e. all of the derived groups with multiple detections removed ) , although we note that the results for each breaking distance were similar . the cumulative luminosity distribution is shown in fig . [ fig : lumfunc ] , along with a power - law fit to the data of the form @xmath13 , shown as a solid ( red ) line . we find that the distribution is well fit by such an analytic form , with @xmath14 . this is in excellent agreement with what was found for the group luminosity function in m33 ( bastian et al . it is also in good agreement with the study of groups of ob stars in the smc based on the mst method by oey , king & parker ( 2004 ) . the implications of this are discussed in detail in bastian et al . we note that the luminosity function may not directly relate to the mass function of the groups if there is a substantial age spread in the groups or if the initial mass function ( imf ) of stars is dependent on environment ( i.e. since we only sample the upper end of the stellar imf we are insensitive to any fluctuations below our sample limits ) . , where the best fitting @xmath15 and error are given in the panel.,width=321 ] the above sections dealt solely with young ob stars , however , using different colour and magnitude cuts in the cmds of the lmc stars from the catalogue of zaritsky et al . ( 2004 ) , we can select populations of different mean ages . this is done by taking colour - magnitude regions ( corrected for foreground extinction of 0.41 , 0.32 , 0.25 , 0.15 mag for u , b , v , and i , respectively ) along the main sequence and red - giant branches of the cmd . in order to minimize the effect of larger contamination in the outskirts of the galaxy , and to assure complete geometrical coverage , we limited our selection to the inner three degrees ( in radius ) of the lmc . for selection of the age boxes we follow the same procedure as gieles , bastian , & ercolano ( 2008 , hereafter gbe08 ) . the adopted regions are shown in figure [ fig : cmd ] , where the sizes of the boxes were chosen such that they enclose 2000 stars within their boundaries . in order to determine the average age of each of the boxes , we constructed a synthetic population of stars , assuming a constant star - formation rate and a random sampling from a salpeter initial mass function ( 1955 ) . we used the padova isochrones ( girardi et al . 2002 and references therein , z=0.008 ) to assign a magnitude and colour for each model star , and assume a distance to the star of 50 kpc . we then find the number of stars at each age which fall into our colour - magnitude regions , and take the mean age of the stars which are contained within the designated region . two examples ( boxes 2 and 6 ) are shown in fig . [ fig : age - windows ] . the selection cuts and mean ages are shown in table [ table : cuts ] . we note that using the median age of each box instead of the mean does not significantly change our conclusions . since we are using a cmd to select stars , the primary parameter that we are selecting is in fact stellar mass and not stellar age . we implicitly assume that high and low mass stars ( above a few solar masses ) have the same spatial distribution . once the boxes have been selected we can then look at the spatial positions of the stars for each box . six examples are shown in fig . [ fig : spatial ] where the box number and mean age are given in each panel . it is clear that the younger populations contain considerably more substructure than the older populations . the lmc bar is clearly visable in the older boxes , namely boxes 7 , 9 , and 12 in fig . [ fig : spatial ] , and is clearly dominating the spatial distribution in box 12 . in the next section we will quantify these differences . cartwright & whitworth ( 2004 ) have developed a formalism to quantify the structure of young star clusters . their method uses the normalised mean mst branch length , @xmath16 , and the normalised distance between all sources within a region , @xmath17 . the ratio between these two quantities , @xmath16/@xmath17= @xmath2 , is able to distinguish between a power - law ( centrally concentrated ) profile and a profile with sub - structure ( i.e. a fractal distribution ) . additionally , this parameter can quantify the index of the power - law or the degree of sub - substructure , which , if one assumes is due to a fractal nature , its fractal dimension can also be estimated . assuming a three dimensional structure , if @xmath2 is less than 0.79 then the region is fractal , larger than 0.79 refers to a power - law structure , and a value of 0.79 implies a random distribution ( a 3d fractal of dimension 3 is a random distribution ) . however , if the distribution is two dimensional , as is approximately true if one is looking at a disk - like galaxy face on ( like the lmc ) , then a random distribution has a @xmath2 value of 0.72 . while this method was developed and has been applied to young embedded clusters ( cartwright & whitworth 2004 , schmeja & klessen 2006 , kumar & schmeja 2007 ) , it has also recently been applied to full stellar populations within the smc ( gbe08 ) , finding that the q parameter increases with the mean stellar population age . during our analysis , we found an unexpected relation between the measured @xmath2 value and the elongation parameter introduced by schmeja & klessen ( 2006 ) , in that more elongated clusters have a lower @xmath2-value . we note here , that while this does not affect our results ( each box has an elongation less than 1.2 ) it may be a serious problem for young embedded clusters which often have a filamentary structure ( e.g. gutermuth et al . 2008 ) . this issue and a proposed solution are dealt with in more detail in appendix a. additionally , in appendix b we discuss the effects of differential extinction on the @xmath2 parameter . in the bottom panel of fig . [ fig : panel ] we show the measured @xmath2-value for each of our 12 boxes ( filled circles ) . the @xmath2-value is clearly increasing with age , beginning at a value of @xmath18 . in order to measure the age at which the spatial distribution becomes uniform , we fit the @xmath2-value as a function of the logarithm of age ( shown as a dashed line in the bottom panel of fig . [ fig : panel ] and calculate when this function crosses that of a flat distribution ( i.e. when the @xmath2-value becomes constant ) . this happens at @xmath19 myr , where the error was calculated by the uncertainty in the value of the flat region of the diagram , with a value of 0.73 . this is similar to what is expected for a 2d slightly centrally concentrated stellar distribution ( a 2d power - law distribution with index @xmath20 , decreasing outward ; cartwright & whitworth 2004 ) , as is our reference distribution ( see [ sec : tpcf ] ) . this indicates that the spatial distribution of the stars becomes more homogeneous as the populations age . however , box 12 has a much larger @xmath2-value , which we attribute to the lmc bar , a strongly centrally concentrated 3d structure in the galaxy . errors were estimated by generating artificial populations with the same number of sources , deriving their @xmath2-value , and taking the standard deviation . the second technique that we used to determine the timescale for structures to be erased in the lmc is the two - point correlation function ( tpcf ) . this method determines the distance between all possible pairs of stars ( d@xmath21 ) , shown as a histogram , which is then normalized to that of a reference distribution , i.e. @xmath22 / @xmath23 . here , @xmath24 is taken as taken to be a slightly centrally concentrated powerlaw with index 0.3 , which is a good representation of the inner three degrees of the lmc for stars with v @xmath25 20 . this is similar to that done by gomez et al . ( 1993 ) who used the tpcf to study the distribution of pre - main sequence stars in taurus , although these authors used a random distribution as their reference distribution . we are then left with a histogram for each age box of the number of connections with a certain distance , relative to the reference distribution . for each of the age boxes we fit the distribution ( in a logarithm of distance vs. logarithm of the number ) with a linear relation to get the slope and zero - point . if the distributions are becoming more similar to the reference distribution , then the slope and zero - points should tend towards values of 0 ( see appendix [ appendix : contamination ] for a demonstration of this ) . the results for all age boxes are shown in the top and middle panels of fig . [ fig : panel ] . similar to the @xmath2-parameter method , we find that the distribution for the boxes is approaching the reference distribution as the mean stellar age increases , for both the zero - point and slope of the distributions . the fact that the values do not go precisely to zero is most plausibly due to a slight mis - characterization of the reference distribution . errors on each of the points were calculated by randomly selecting 1100 stars in the age box and repeating the analysis . as can be seen by comparing all three panels in fig . [ fig : panel ] , the three measurements of the disappearance of substructure agree very well . for the two measurements provided by the tpcf , we fit the zero - point and slope as a function of the logarithm of age ( shown as dashed lines ) , and find that this crosses the flat portion of the diagrams at @xmath26 myr , where the error indicates the uncertainty in defining the flat region of the diagram . myr , suggesting that most substructure has been erased by this age . the results of the @xmath2-parameter are shown in the bottom panel , which largely agree with the results of the tpcf , i.e. that structure is erased in @xmath27 myr . additionally , we show the evolution of @xmath2 for the cluster sample of hunter et al . ( 2003).,width=321 ] the three measurements of the erasure of substructure within the galaxy ( the @xmath2-parameter , and the slope and zero - point of the tpcf ) are all in close agreement . _ these results show that stars are formed with a large amount of substructure in the lmc and that , as this population evolves , it becomes more uniform , becoming statistically indistinguishable from the background reference distribution within @xmath5 myr . _ through the @xmath2-parameter , and comparing with artificially generated distributions , as was done in cartwright & whitworth ( 2004 ) , we estimate that stars are born with a fractal distribution , with a 2d fractal dimension of @xmath4 . such evolution of stellar structures has also been seen in ngc 1313 ( pellerin et al . 2007 ) and ngc 4449 ( annibali et al . 2007 ) using hst - acs data . both studies found that massive stars ( ob stars ) contain more structure than older stellar populations . pellerin et al . ( 2007 ) explain this evolution of structure as due to infant mortality , i.e. that young stars are often formed in dense clusters which then pop ( kroupa 2002 ) when the residual gas left over from the star - formation process leaves the cluster ( e.g. bastian & goodwin 2006 ) . due to our ability to quantify the timescale and degree of clustering we can rule out this scenario , which we will return to in [ sec : interpretations ] . in such statistical analyses as presented here , contaminating sources may potentially seriously affect the results . in order to estimate the magnitude of these effects we have run a series of simulations where we generate artificial distributions with known parameters , and substitute random ( or reference ) distributions with a given percentage . the results of these experiments are shown and discussed in appendix [ appendix : contamination ] . in particular , we note that young stars will contribute to all age boxes ( with the exception of box 12 ) , and since they display a larger degree of substructure , some degree of substructure is expected to be present in all boxes . however , the contamination is expected to be low ( much less than 30% ) , hence the statistical methods should not be affected by them , although they can be seen by eye in fig . [ fig : spatial ] . in order to further understand the evolution of stellar structures within the lmc we take advantage of two recent surveys . the first is that of whitney et al . ( 2008 ) who used the sage ( meixner et al . 2006 ) multi - band spitzer survey of the lmc to select high probability young stellar objects ( ysos ) . these populations are expected to trace the evolution of forming stars / clusters from the parent clouds , through the embedded phase to the point where they can be detected at optical wavelengths . the second survey was that of hunter et al . ( 2003 ) who catalogued the cluster population of the lmc , and measured spatial positions , magnitudes , and ages for over 900 clusters . constraining the yso sample to the inner three degrees of the lmc ( as was done for the optically selected samples ) we find @xmath28 . this is significantly higher than that found for the young optically selected ( ob star ) sample ( @xmath29 ) . since agb stars and ysos have similar colours even in the mid - ir ( e.g. whitney et al . 2008 ) it is possible that this difference is due to contamination of the yso sample . if it is due to contamination ( see appendix [ appendix : contamination ] ) then this implies that the yso sample has approximately @xmath30 contamination from stars which have a similar distribution to that of our reference distribution . this may be the case if agb stars are the main contamination . the alternative interpretation , is that the yso sample is made up largely of lower - mass stars than our ob star sample ( whitney et al . 2008 ) , and that the distribution of star formation is mass dependent . such extreme spatial dependence on mass appears to be at odds with observations of young galactic star - forming regions ( as indicated by the similarity of the stellar imf in different environments ) . finally , we note that the lower spatial resolution of the yso sample may act as to raise the measured @xmath2-value . for the cluster distribution , we again limit the sample to the inner three degrees in order to draw useful comparisons with our optically selected samples , and additionally provide a magnitude cut in the data of @xmath31 in order to avoid incompleteness effects in the central region . we then break the cluster sample up into 8 age bins , with each bin containing an equal number of clusters , namely 69 . the resulting distribution of @xmath2-parameters vs. cluster age is shown in the bottom panel of fig . [ fig : panel ] . as seen in the stellar distributions , the @xmath2-parameter of the cluster populations increases with mean age , and becomes flat between @xmath32 myr . the larger scatter in the cluster points relative to the stellar boxes is due primarily to the lower number of points used to calculate @xmath2 ( @xmath33 vs. @xmath34 ) . overall , the cluster population loses structure in a similar way to the stellar population , and if anything , the structure removal is faster for the clusters . we will return to this point in [ sec : interpretations ] . it has been shown that the majority of clusters that form become disrupted within the first few 10s of myr of their lives ( e.g. lada & lada 2003 ) . the mechanism believed to be responsible for this is the rapid removal of gas , due to stellar winds and/or supernovae , from the natal cluster . this gas removal leaves the cluster in an unbound state due to the removal of a large fraction of the gravitational potential of the cluster , that leaves the remaining stars in a super - virial state . if the star - formation efficiency is low ( @xmath35 ) the entire cluster may become unbound and expand rapidly . these types of clusters have been referred to as popping clusters ( kroupa 2002 ) and evidence of their expanding nature has been seen in their low light - to - mass ratios for a given age ( goodwin & bastian 2006 ) . the clusters are predicted to expand with a velocity close to their velocity dispersion and may have a severe effect on the observed galactic structure . the standard scenario for the popping clusters is that the stars and gas are in , or close to , virial equilibrium when the natal gas is removed . thus , we can use the virial equation to find the relation between a cluster s mass ( gas+stars ) and velocity dispersion ( e.g. spitzer 1987 ) . due to the relatively low star - formation rate in the lmc we do not expect a large population of massive clusters . assuming a stochastically sampled power - law mass function of the clusters ( gas+stars ) with index @xmath3 , lower mass limits of @xmath36 , and cluster radii of 1 pc , we can set up a simple model of the evolution of the structure of a galaxy due to infant mortality . we start with a spatial distribution of clusters with the same @xmath2-value as the youngest observed clusters . at each of these cluster centres we place a cluster of mass m@xmath37 and radius 1 pc we then assign a number of stars to each cluster , directly proportional to its mass , with a gaussian spatial distribution ( stochastically sampled ) . the velocity dispersion of each cluster ( gas+stars ) is derived through the virial equation , assuming a given star - formation efficiency ( sfe , c.f . goodwin & bastian 2006 ) . the cluster expansion is mimicked by increasing r@xmath37 , with r@xmath37@xmath38@xmath39 , where @xmath40 is the time since the cluster began expanding and @xmath39 is the cluster velocity dispersion . at each timestep , we select 1000 stars at random and calculate their @xmath2-value . only stars within the inner 3 degrees ( the simulations were set up relative to the observed lmc distribution ) were selected . the results of these simulations are shown in fig . [ fig : evo ] , where the top , middle and bottom panels represent simulations where the mass function of the clusters is sampled down to 100 , 10 , & 1 m@xmath41 respectively . since we have used the same number of clusters in each simulation , the horizontal shift in the simulations is related to the average velocity dispersion of the clusters , which decreases with decreasing m@xmath42 . if clusters start more compact ( as suggested by observations of young clusters , e.g. mackey et al . 2007 , bastian et al . 2008 ) , this would shift the curves to the left . the cluster mass function is thought to be continuous down to single stars ( e.g. oey et al . 2004 ) , arguing for the applicability of the simulations with m@xmath42 = 1 m@xmath41 . this suggests that the erasure of structures is not being driven by popping clusters. however , a caveat to these simulations is that we have assumed that all clusters pop in the same manner , regardless of mass . this , however , may not be true , as the number of o - stars ( which drive the gas removal , and hence the popping ) in the cluster may influence the expansion properties ( e.g. baumgardt & kroupa 2008 ) . additionally , we have not taken the motion of the clusters through the galaxy into account which also destroys substructure ( see [ sec : galdyn ] ) . we can test the effect of infant mortality another way , namely by looking at the evolution of the @xmath2-value for stars and that for catalogued star clusters . if infant mortality was a dominant cause of structural evolution within the galaxy , then we would expect the @xmath2-values of the stars to increase more rapidly with age than that of clusters . this is because the star clusters which survive the infant mortality phase will not have their positions changed , whereas the stars that formed in clusters that popped will be expanding away from their natal positions . in other words , the stars will have two components to their velocity , namely expansion and motions through the galaxy , whereas clusters will only have one component , the velocity dispersion of the galaxy . for the cluster catalogue we use the hunter et al . ( 2003 ) survey presented in [ sec : comparison ] . in the bottom panel ( open circles ) of fig . [ fig : panel ] we show the measured @xmath2-value for each of the cluster sub - samples . in general we see the same behavior as found for the stellar boxes , namely that the @xmath2 value increases with mean age , beginning with a highly fractal , or substructured distribution and becoming flat near the same value as that found for the stellar boxes . the evolution between the stars and clusters is quite similar , in fact the clusters seem to evolve slightly faster ( although this does not appear to be statistically significant given the small number of clusters per age bin ) than the stars . hence we conclude that the same mechanism is erasing the substructure seen in the stellar and cluster distributions , which eliminates the possibility that popping clusters are the dominant cause , although this effect is likely taking place within the galaxy . we conclude that the dominating effect of structure evolution in the lmc is general galactic dynamics . it is the velocity dispersion of the galaxy which eliminates substructures and pushes the stellar distributions toward a random , or background , distribution . these results are consistent with the findings of the distribution of cepheids in the galaxy ( efremov 1995 ) and the lmc ( marconi et al . for the galaxy , efremov ( 1995 ) found evidence for the presence of large cepheid clouds with sizes of a few hundred parsecs , which , considering the age of cepheids ( a few 10s to 100s of myr ) is what is expected from our results . more directly , marconi et al . ( 2004 , their figs . 4 & 5 ) found that young cepheids ( @xmath43 myr ) display much more substructure in their spatial distribution than older cepheids ( @xmath44 myr ) whose distribution is becoming closer to random , in agreement to that found by elmegreen & efremov ( 1996 ) . finally , we note that this is a similar timescale ( @xmath45 myr ) to that found by efremov & elmegreen ( 1998 ) for the average separation between pairs of star clusters in the lmc to become constant , i.e. for the cluster population to become statistically uniform . the lmc has a velocity dispersion of @xmath46 km / s ( carrera et al . 2008 ) , if we adopt a radius of the lmc of @xmath47 kpc this translates to a crossing time ( @xmath48 ) of @xmath49 myr . this is remarkably similar to the timescale derived for the elimination of structure within the lmc . from this , we conjecture that structure will be erased on a crossing - time , independent of spatial scale . this may explain why young lower - mass clusters often contain large amounts of substructure ( e.g. gutermuth et al . 2005 ) , while many of the massive young clusters ( such as orion ) have much smoother central regions , where the crossing time is significantly lower than the outer regions ( see elmegreen 2008 ) . in gbe08 we show the same quantitative relationship between the stars and clusters in the smc and reach a similar conclusion , namely that popping clusters are not driving the evolution of substructure within the galaxy . in the smc , substructure is removed in @xmath50 myr , which is again similar to the galactic crossing time . note that the erasure of structure in the popping cluster scenario ( e.g. fig . [ fig : evo ] ) is mainly dependent on the average internal velocity dispersion @xmath51 of the clusters . since @xmath51 is independent of the number of clusters in a population ( it only depends on the mass function index and lower mass limit of clusters - which we assume to be invariant ) then all galaxies should have structure erases in approximately the same time . however , in the timescale of structure erasure is twice as fast as in the lmc . an interesting difference between the smc and lmc stellar and cluster distributions is the @xmath2-values which they attain when the become flat . for the smc the @xmath2-value of both clusters and stars becomes flat at @xmath52 whereas for the lmc , both distributions become flat at @xmath53 . this difference shows the ability of the method to quantify the structure of distributions , as the smc is largely three dimensional whereas the lmc is mostly disk - like , or two dimensional ( both slightly centrally concentrated ) . the expected @xmath2-values for homogeneous 2d and 3d distributions are 0.71 and 0.79 , respectively ( cartwright & whitworth 2004 ) , therefore explaining the observed difference between the @xmath2-values in the lmc and smc . we have presented an analysis of the spatial distribution of stellar structures within the lmc . by using the zaritsky et al . ( 2004 ) magellanic cloud photometric survey and sampling different parts of the stellar colour - magnitude diagram , we have selected stellar populations of varying mean age . this is supplemented with other datasets , namely the whitney et al . ( 2008 ) young stellar object ( yso ) infrared sample and the star cluster catalogue of hunter et al . ( 2003 ) . based on the analysis of the distribution of ob stars , we confirm our earlier results on m33 , namely that the star - forming groups are well approximated by a power - law luminosity function with index of @xmath3 and that there is no preferred size in the distribution of star - forming sites ( e.g. that of the ob association ) . instead , we find a continuous distribution of sizes , and find that previous reports of a characteristic size were due to resolution effects , along with choosing , apriori , a single size - scale to probe . we have used two methods to quantify the amount of substructure within a galaxy , the q method ( carthwright & whitworth 2004 ) , and the two - point correlation function ( tpcf ) . we found that stars are formed in a highly structured way ( approximating the distribution as a 2d fractal we find the fractal dimension to be @xmath4 ) and that this distribution evolves rapidly , becoming indistinguishable from the background or reference distribution after @xmath54 myr . we also found that the distribution of stellar clusters follows that of the stars of similar age , adding confidence in our age dating method as well as our structural analysis . thus , our findings suggest that stars are formed in a highly substructured distribution and that this distribution evolves rapidly , with small scale structures dispersing first , and that all of the original structure is removed in @xmath5 myr . this is approximately the crossing time of the galaxy and as such , we suggest that structures in galaxies , presumably regardless of spatial scale , are eliminated in a crossing time . similar analysis done on the smc ( gbe08 ) finds that substructure is erased on @xmath55 myr , which is comparable to the crossing time of that galaxy . by comparing the observations to simple models , we conclude that infant mortality ( or popping star clusters ) have a negligible effect on the stellar distribution in the lmc and that general galactic dynamics is the major driver of structural change . it is worth noting that our results are not necessarily in contradiction with the infant mortality scenario , but they show that this mechanism does not explain the observed evolution of stellar structure in galaxies . only a small fraction of stars are born in stellar clusters which will survive long enough to be able to be identified as clusters in optical surveys ( e.g. lada & lada 2003 , gieles & bastian 2008 , bastian 2008 ) . whether this is due to a high fraction of stars born in embedded clusters which quickly disrupt or due to a large fraction of stars forming outside clusters ( or more likely , some combination of these two extremes ) is still an open question at the moment . we gratefully thank simon goodwin for discussions and help with generating fractal distributions . we also thank barbara whitney for providing her list of yso candidates before publication and anil seth for interesting discussions . the referee , jason harris , is thanked for helpful comments and suggestions on the text , techniques , and results presented in this paper . nb gratefully acknowledges the support of the harvard - smithsonian center for astrophysics , where a large part of this work was carried out . be was partially supported by _ grants go6 - 7008x and go6 - 7009x . 99 allen , l. , et al . 2007 , protostars and planets v , 361 annibali , f. , aloisi , a. , mack , j. , tosi , m. , van der marel , r. p. , angeretti , l. , leitherer , c. , & sirianni , m. 2008 , aj , 135 , 1900 bastian , n. , gieles , m. , efremov , y. n. , & lamers , h. j. g. l. m. 2005 , a&a , 443 , 79 bastian , n. , & goodwin , s. p. 2006 , mnras , 369 , l9 bastian , n. , ercolano , b. , gieles , m. , rosolowsky , e. , scheepmaker , r. a. , gutermuth , r. , & efremov , y. 2007 , mnras , 379 , 1302 bastian , n. , gieles , m. , goodwin , s. p. , trancho , g. , smith , l. j. , konstantopoulos , i. , & efremov , y. 2008 , mnras , 389 , 223 bastian , n , 2008 , mnras , 390 , 759 battinelli , p. , efremov , y. , & magnier , e. a. 1996 , a&a , 314 , 51 baumgardt , h. , & kroupa , p. 2007 , mnras , 380 , 1589 bresolin , f. , et al . 1998 , aj , 116 , 119 carrera , r. , gallart , c. , hardy , e. , aparicio , a. , & zinn , r. 2008 , aj , 135 , 836 cartwrigth , a. & whitworth , a.p . 2004 , mnras , 348 , 589 efremov , yu.n . 1984 , vestnik ( messenger ) of the ussr acad . sci . , # 12 , 56 efremov , yu.n . 1995 , aj , 110 , 2757 efremov , yu.n . , & elmegreen , b.g . 1998 , mnras , 299 , 588 elmegreen , b. g. , & efremov , y. n. 1996 , apj , 466 , 802 elmegreen , b. g. , & falgarone , e. 1996 , apj , 471 , 816 elmegreen , b. g. , & elmegreen , d. m. 2001 , aj , 121 , 1507 elmegreen , b. g. , kim , s. , & staveley - smith , l. 2001 , apj , 548 , 749 elmegreen , b. g. , elmegreen , d. m. , chandar , r. , whitmore , b. , & regan , m. 2006 , apj , 644 , 879 elmegreen , b. g. 2008 , apj , 672 , 1006 elmegreen , d. m. , & salzer , j. j. 1999 , aj , 117 , 764 ercolano , b. , bastian , n. , & stasiska , g. 2007 , mnras , 379 , 945 gieles , m. , & bastian , n. 2008 , a&a , 482 , 165 gieles , m. , bastian , n. , ercolano , e. 2008 , mnras , in press ( arxiv:0809.2295 ) girardi , l. , bertelli , g. , bressan , a. , chiosi , c. , groenewegen , m. a. t. , marigo , p. , salasnich , b. , & weiss , a. 2002 , a&a , 391 , 195 gomez , m. , hartmann , l. , kenyon , s. j. , & hewett , r. 1993 , aj , 105 , 1927 goodwin , s. p. , & whitworth , a. p. 2004 , a&a , 413 , 929 goodwin , s. p. , & bastian , n. 2006 , mnras , 373 , 752 gutermuth , r. a. , megeath , s. t. , pipher , j. l. , williams , j. p. , allen , l. e. , myers , p. c. , & raines , s. n. 2005 , apj , 632 , 397 gutermuth , r. a. , et al . 2008 , apj , 674 , 336 hunter , d. a. , elmegreen , b. g. , dupuy , t. j. , & mortonson , m. 2003 , aj , 126 , 1836 kim , s. , & park , c. 2007 , apj , 663 , 244 kroupa , p. 2002 , mnras , 330 , 707 kumar , m. s. n. , & schmeja , s. 2007 , a&a , 471 , l33 mackey , a. d. , wilkinson , m. i. , davies , m. b. , & gilmore , g. f. 2007 , mnras , 379 , l40 marconi , m. , & clementini , g. 2005 , aj , 129 , 2257 marconi , m. , bono , g. , caputo , f. , cassisi , s. , pietrukowicz , p. , pietrzynski , g. , & gieren , w. 2006 , memorie della societa astronomica italiana , 77 , 67 meixner , m. , et al . 2006 , aj , 132 , 2268 lada , c. j. , & lada , e. a. 2003 , araa , 41 , 57 massey , p. , olsen , k. a. g. , hodge , p. w. , jacoby , g. h. , mcneill , r. t. , smith , r. c. , & strong , s. b. 2007 , aj , 133 , 2393 odekon , m.c . 2008 , apj , 681 , 1248 oey , m. s. , king , n. l. , & parker , j. w. 2004 , aj , 127 , 1632 padoan , p. , kim , s. , goodman , a. , & staveley - smith , l. 2001 , apjl , 555 , l33 palla , f. , & stahler , s. w. 2000 , apj , 540 , 255 pellerin , a. , meyer , m. , harris , j. , & calzetti , d. 2007 , apjl , 658 , l87 salpeter , e. e. 1955 , apj , 121 , 161 snchez , n. , alfaro , e. j. , elias , f. , delgado , a. j. , & cabrera - cao , j. 2007 , apj , 667 , 213 schmeja , s. , & klessen , r. s. 2006 , a&a , 449 , 151 spitzer , l. 1987 , `` dynamical evolution of globular clusters '' , princeton , nj , princeton university press , 1987 , 191 p. , whitney , b. a. , et al . 2008 , aj , 136 , 18 zaritsky , d. , harris , j. , thompson , i. b. , & grebel , e. k. 2004 , aj , 128 , 1606 in this section we investigate the influence of the elongation on the calculated q parameter . for consistency , we define the elongation in the same way as schmeja & klessen ( 2006 ) , namely as the ratio of the derived radius when approximating the region as a circle to the radius when using the convex hull method . schmeja & klessen ( 2006 ) give the relation between the elongation and the major to minor axis ratio of an ellipse . in order to test the effect we first create 3d spherical models of varying fractal dimensions , uniform distribution and various centrally concentrated profiles , similar to what was done in cartwright & whitworth ( 2004 ) . we then multiply one of the three principle axes by a factor and measure the q value and elongation along the three projected axes . the results for two fractal models ( fractal dimensions of 2.0 and 2.5 ) , a uniform distribution and a centrally concentrated distribution with power - law radial index of -2 are shown in fig . [ fig : q - elongation ] . it is clear that increasing the elongation decreases the measured q value . we have also fit each relation with a linear slope shown as a dashed line . the error bars represent the standard deviation of three realisations of the simulations ( the fractal models begin with a random seed ) . while the slope of the change in q as a function of the elongation is somewhat dependent on the starting q value ( i.e. what type of distribution is used ) the overall trend is quite similar for all profiles considered . as such , we use the average slope of the distributions in the left panel of fig . [ fig : q - elongation ] in order to provide a correction function . for this we find @xmath56 . the right panel of fig . [ fig : q - elongation ] shows the resulting values of the simulations ( shown in the left panel ) after this correction has been applied . here the dashed lines represent the input q value ( i.e. that determined for the profile where no elongation has been applied ) . while the given correction appears to correct well the simulations we note that large errors can be expected for systems with high elongations ( @xmath57 , i.e. major / minor axis ratios greater than 2 ) . an additional affect which may influence the measured @xmath2 parameter is that of differential extinction . this can be realized conceptually by imagining a uniform stellar population and putting a highly sub - structured distribution of gas / dust in the line of sight . if the opacity is high enough , this will remove whole sections of the stellar distribution from sight , creating the appearance of a non - uniform stellar population . here , we we attempt to quantify this affect . for the simulations we make 3d fractal distributions of point sources with dimension 2.0 following the method presented in goodwin and whitworth ( 2004 ) . for each simulation we create 20,000 gas particles and 1,000 stellar particles independently . we then chose a line of sight and counted the number of gas particles between the observer and each star within some projected tolerance , i.e. we find the surface density of gas between the observer and each star . stars which were found to have gas surface densities above a certain critical value were removed from the dataset . finally , we calculate q for the original ( non - extincted ) and extincted datasets and compare their values . three projections of one of the simulations are shown in fig . [ fig : q - sims - figures ] where the contours represent the integrated column density of gas along the line of sight , red circles are stars which have been removed from the original dataset ( i.e. are found to be extincted ) , and green circles represent stars which were not extincted ( i.e. the red and green circles together make up the initial stellar distribution ) . the main result is that extinction from clouds tends to lower the value of q , where the change is related to the number of stars removed ( i.e. the opacity of the clouds ) . typical changes in the derived q value , @xmath58 , are lowered by 0.04 to 0.08 when 20% or 50% of the sources are extincted out of the sample , respectively . this makes the distribution appear to have a lower fractal dimension than the actual distribution . we have tested the effect of varying the number of particles used and the input fractal dimensions of the distribution and found that both effects are minimal . finally , we test the effect of culling the stars randomly and find that this does not change the value of q ( @xmath59 ) . since an unphysically large number of stars must be extincted out of our sample before q is affected , we conclude that extinction does not significantly affect our results . due to a variety of effects , some contamination may be expected to enter observational datasets . this could be due to field stars in the galactic halo or young ( lower - mass ) stars which have colours and magnitudes which fall in the cmd boxes used to selected older age stars . in order to estimate the strength of these effects we have constructed the following experiments . we start with a 2d projection of a 3d fractal model , chosen to have a similar q - value as our youngest stellar age - box . we then replace a fraction of the model stars , whose positions are chosen according to the probability function of a background distribution . for the background we chose three distributions ( all 2d distributions ) ; a stochastic distribution , and two centrally concentrated distributions , both power - laws , with indices 0.3 and 1.0 . we used the prescription in cartwright & whitworth ( 2004 ) to construct the background distributions . the centrally concentrated distribution with index 0.3 was chosen as this matches the projected radial profile of the lmc , when restricted to the inner three degrees in radius . the results of the simulations are shown in fig . [ fig : contamination - q ] . in each case , we see the q - value increase rapidly with increasing contamination . in particular , the q - value in the centrally concentrated case ( with power - law index of 0.3 ) becomes largely flat for contamination fractions above 0.6 . this explains why small amounts of fractal substructure in the older age boxes ( due to young stars which enter our selection box ) do not significantly affect the results , as a large fraction of these stars would need to be present . figure [ fig : contamination - tpcf ] shows how the two - point correlation function analysis is affected by the same contamination . as seen in the q - distribution , the slope and zero - point of the tpcf smoothly transform from the value expected for a fractal distribution , to that expected for the background or reference distributions . thus we conclude that while younger stars ( which display more substructure ) will contribute to all age boxes , this effect will not significantly alter our conclusions , as the expected contamination is low , their distribution will not significantly change the measured q distribution or alter the tpcf .	we present an analysis of the spatial distribution of various stellar populations within the large magellanic cloud . we combine mid - infrared selected young stellar objects , optically selected samples with mean ages between @xmath0 and @xmath1 myr , and existing stellar cluster catalogues to investigate how stellar structures form and evolve within the lmc . for the analysis we use fractured minimum spanning trees , the statistical @xmath2 parameter , and the two - point correlation function . restricting our analysis to young massive ( ob ) stars we confirm our results obtained for m33 , namely that the luminosity function of the groups is well described by a power - law with index @xmath3 , and that there is no characteristic length - scale of star - forming regions . we find that stars in the lmc are born with a large amount of substructure , consistent with a 2d fractal distribution with dimension @xmath4 and evolve towards a uniform distribution on a timescale of @xmath5 myr . this is comparable to the crossing time of the galaxy and we suggest that stellar structure , regardless of spatial scale , will be eliminated in a crossing time . this may explain the smooth distribution of stars in massive / dense young clusters in the galaxy , while other , less massive , clusters still display large amounts of structure at similar ages . by comparing the stellar and star cluster distributions and evolving timescales , we show that infant mortality of clusters ( or popping clusters ) have a negligible influence on galactic structure . finally , we quantify the influence of the elongation , differential extinction , and contamination of a population on the measured @xmath2 value . [ firstpage ] galaxies : star clusters galaxies : magellanic clouds galaxies : structure
the cosmic microwave background ( cmb ) anisotropies detected by the cobe satellite ( bennet et al . 1996 ) confirmed the notion that the present structure in the universe originated from small density fluctuations at early times . the gravitational collapse of overdense regions could explain the present - day abundance of bound objects , such as galaxies or x - ray clusters , under the appropriate extrapolation of the detected large - scale anisotropies to smaller scales ( e.g. , baugh et al . 1997 ; pen 1996 ) . recent deep observations with the hubble space telescope ( steidel et al . 1996 ; madau et al . 1996 ) and ground - based telescopes , such as keck ( lowenthal et al . 1996 ) , have constrained considerably the evolution of galaxies and their stellar content at @xmath5 . however , in the bottom - up hierarchy of the popular cold dark matter ( cdm ) cosmologies , galaxies were assembled out of building blocks of smaller mass . the elementary building blocks , i.e. the first gaseous objects to form , had a total mass of order the jeans mass ( @xmath15 ) , below which gas pressure opposed gravity and prevented collapse ( haiman & loeb 1997 ; ostriker & gnedin 1997 ) . in the standard cdm cosmology , these _ basic _ building blocks formed at @xmath16@xmath17 ( see fig the _ first light _ from stars and quasars ended the `` dark ages '' of the universe and initiated a `` renaissance of enlightenment '' in the otherwise fading glow of the big bang . it is easy to see why the mere conversion of trace amounts of gas into stars or black holes at this early epoch could have had a dramatic effect on the ionization state and temperature of the rest of the gas in the universe . nuclear burning releases @xmath18 ev per hydrogen atom , and thin - disk accretion onto a schwarzschild black hole releases ten times more energy ; however , the ionization of hydrogen requires only 13.6 ev . it is therefore sufficient to convert a minimum fraction of @xmath19 of the baryonic mass into stars or black holes in order to ionize the rest of the universe . ( the actual required fraction is higher because some of the emitted photons are below 13.6 ev and because each hydrogen atom recombines multiple times at high redshifts ) . the free electrons produced in this process erase the microwave background anisotropies on angular scales below the size of the horizon at the reionization epoch ( @xmath20 ) , by an amplitude comparable to their optical depth to thomson scattering . a variety of cdm models that are all consistent with both the cobe anisotropies ( @xmath21 ) and the abundance of objects today ( @xmath22 ) differ appreciably in their initial amplitude of density fluctuations on small scales . the reionization history of the universe is determined by the collapse redshift of the smallest objects ( @xmath23 ) and is therefore ideally suited to discriminate among these models . we have not yet observed the first transition that the universe made between its nearly smooth initial state and its clumpy present state . if i had been asked to envision an instrument that would directly probe this transition epoch , i would have probably sketched something like the next generation space telescope ( ngst ) . there are three reasons that make this choice natural : ( i ) at @xmath24 , the rest frame optical uv emission of star clusters or quasars is redshifted to infrared wavelengths in the observer frame ; ( ii ) the first sources are faint because of their high redshift and low mass . their detection , therefore , requires a sensitive infrared telescope with a large collecting area , in space . the last requirement is necessary to avoid atmospheric background / opacity and to reduce the zodiacal background through an extended solar orbit ; and ( iii ) high - redshift objects are likely to be denser and more compact than their low - redshift counterparts . this simply follows from the increase in the mean density of the universe at early times . in the bottom - up hierarchy of structure formation , this tendency is also enhanced by the decrease of the nonlinear mass scale with redshift . the physical size of an object of mass @xmath25 should scale roughly as @xmath26 , where @xmath27 is its formation redshift . high angular resolution is therefore required for imaging star clusters at @xmath14 . the detection of the transition epoch could provide invaluable information about the ionization and star formation history of the universe . the study of stars at high - redshift is also linked to the `` archeology '' of old stars in the milky way halo . the metallicity - age trends which are identified in the milky - way galaxy ( mcwilliam et al . 1996 ) should be related to the various epochs of star formation in the early universe . so much for theoretical arguments . but is there direct evidence that structure started to form in the universe long before galaxies were assembled ? the answer is , yes . by now , a number of observational clues constitute a `` smoking gun '' for an early population of stars that had formed at @xmath28 : * recent spectroscopic observations of the forest at @xmath29 show evidence for a metallicity @xmath4 in absorbers with hi column densities as low as @xmath30 ( cowie 1996 ; songaila & cowie 1996 ; tytler et al . numerical simulations identify such absorbers with mildly overdense regions in the intergalactic medium , out of which nonlinear objects such as galaxies condense ( hellstein et al . 1997 , and references therein ) . indeed , damped absorbers , which are thought to be the progenitors of present day galaxies , show similar metallicities during the early phase of their formation , at @xmath31 ( lu , sargent , & barlow 1996 ) . the universality of this mean metal abundance in absorbers , spanning a range of six orders of magnitude in hi column density , indicates that an early phase of metal enrichment by stars occurred throughout the universe at @xmath28 . * the lack of complete absorption shortward of the resonance ( the so - called `` gunn - peterson effect '' ; gunn & peterson 1965 ) in quasar spectra at @xmath32 implies that the intergalactic gas was highly photoionized before that time . on the other hand , the bright quasar population has been observed to decline at @xmath29 , and so stellar sources might be necessary to keep the intergalactic gas in its inferred high - ionization state ( shapiro & giroux 1987 ; miralda - escud & ostriker 1990 ) . * microlensing surveys argue that @xmath33 ( instead of the previously known range of 0100% ! ) of the mass of the halo of the milky way galaxy might be in the form of massive compact halo objects ( machos ) with a mass @xmath34 ( alcock et al . it remains to be seen whether future microlensing studies confirm these early reports . a stellar population that forms before galaxies are assembled , would behave as collisionless matter and populate the halos of galaxies like cold dark matter ( cdm ) particles do . in this review i will describe the properties of the first stars and quasars and their effect on the ionization history of the universe . much of the quantitative details of this sketch can be found in a series of papers that i wrote in collaboration with my student , zoltan haiman . the underlying theme of these papers was to calibrate the net fraction of gas that is converted into stars based on the measured c / h ratio in the forest . this calibration fixes the reionization history for a given stellar initial mass function ( imf ) and a particular cdm cosmology . the various stages in the reionization history of the universe are illustrated in figure 2 . this sequence follows the collapse redshift history of baryonic objects shown in figure 1 , which was calculated with a spherically - symmetric code for gas and dark matter ( haiman , thoul , & loeb 1996 ) . for objects with baryonic masses @xmath35 , gravity dominates and results in the characteristic bottom - up hierarchy of cdm cosmologies . at lower masses , gas pressure delays the collapse and results in a top - down hierarchy , reminiscent of hot dark matter cosmologies . the first objects to collapse are located at the `` knee '' that separates these regimes . such objects reach virial temperatures of several hundred degrees and could fragment into stars only through cooling by molecular hydrogen [ see haiman et al . ( 1996 ) or tegmark et al . ( 1997 ) , for details regarding the chemistry network leading to the formation of @xmath36 in a primordial gas ] . however , @xmath36 is fragile and could easily be photo - dissociated by photons with energies of @xmath37@xmath38ev , to which the universe is transparent even before it gets ionized . haiman , rees , & loeb ( 1996 ) showed that a uv flux of @xmath39 is capable of dissociating @xmath36 throughout the collapsed environments in the universe . this flux is lower by @xmath40 orders of magnitude than the minimum flux necessary to ionize the universe , which amounts to one uv photon per baryon . the inevitable conclusion is that soon after the first stars form , the formation of more stars due to @xmath41 cooling is suppressed . further fragmentation is possible only through atomic line cooling , which is effective in objects with high virial temperatures , @xmath42k . such objects correspond to a total mass @xmath43^{-3/2}$ ] . figure 2 illustrates this sequence of events by describing two classes of objects : those with @xmath44k ( small dots ) and those with @xmath45k ( large dots ) . in the first stage ( top panel ) , some low - mass objects collapse , form stars , and create ionized hydrogen ( hii ) bubbles around them . once the uv background between 11.213.6ev reaches some critical level , @xmath41 is photo - dissociated throughout the universe and the formation of new stars is delayed until objects with @xmath42k collapse . each massive source creates an hii region which expands into the intergalactic medium . initially the volume of the universe is dominated by neutral hydrogen ( hi ) . but as new sources appear exponentially fast ( due to the gaussian nature of the perturbations ) , numerous hii bubbles add up , overlap , and transform all the remaining hi into hii over a short period of time . this process resembles a first - order phase transition . since the characteristic separation between sources is eventually very much smaller than the hubble distance , the transition completes over a period of time which is much shorter than the hubble time , and can be regarded as sudden . because the potential wells of the first clusters are relatively shallow ( @xmath46@xmath47 ) , supernova driven winds are likely to have expelled the metal rich gas out of these systems and mixed it with the intergalactic medium . incomplete mixing could have led to the observed order - of - magnitude scatter in the c / h ratio along lines - of - sight to different quasars ( rauch , haehnelt , & steinmetz 1996 ; hellsten et al . it is an interesting coincidence that the supernova energy output associated with a metal enrichment of @xmath4 corresponds to @xmath10 ev per hydrogen atom , which is just above the binding energy of these early star clusters . supernova feedback in these objects could have therefore dictated the average metallicity observed in the forest . direct observations of these supernovae might be feasible in the future ( miralda - escud & rees 1997 ) . the measured c / h ratio can be used to calibrate the net fraction of gas that is converted into stars by a redshift @xmath48 . for a scalo ( 1986 ) imf of stars , this corresponds to converting @xmath49 of all the gas into stars of the gas in each object formed stars or whether a small fraction of all objects formed stars with a higher efficiency . we assume that supernova feedback limits star formation to a single starburst and sets the efficiency of star formation to @xmath50 in all objects . in the alternative scenario , ngst should see fewer but brighter sources . ] . haiman & loeb ( 1997 ) have used this fraction and an extension of the press schechter formalism to calculate the early star formation history in a variety of cdm cosmologies . we combined the collapse redshift information in figure 1 with the press - schechter theory and assumed that each object converts a fixed fraction of its gas into stars in a single starburst . we then used the spectral atlases of kurucz ( 1993 ) and the stellar evolutionary tracks of schaller et al . ( 1992 ) to find the emission spectrum of the stars as a function of redshift . the ionizing radiation from each star cluster was absorbed in part by the gas in it ( assumed to be distributed as a singular isothermal sphere at the virial temperature ) and then propagated into a smooth intergalactic medium ( igm ) , using the radiative transfer equations in an expanding universe . for a wide range of models and a scalo imf , we have found that the universe is reionized by a redshift @xmath920 , and that the resulting optical depth to electron scattering is @xmath2 ( cf . this result changes qualitatively only if the imf is tilted by a power - law index @xmath51 relative to the scalo shape ( a bias towards high mass stars is likely due to the lack of cooling by metals and the associated increase in the jeans mass ) . the sensitivity to model parameters other than the imf is only logarithmic because of the exponential dependence of the collapsed baryonic fraction on redshift for gaussian fluctuations . however , the formation time of cdm halos has a very weak dependence on mass scale at low masses , and it is not obvious whether the press - schechter approach is accurate in this limit . three - dimensional simulations ( gnedin & ostriker 1997 ) find reionization somewhat later , at @xmath52 , based on a significantly different star formation history than the our semi - analytic approach predicts . future microwave anisotropy satellites , such as the microwave anisotropy probe ( map ) or the planck surveyor , could detect a reionization optical depth as low as 2% or 0.5% respectively after processing their polarization data ( zaldarriaga , spergel , & seljak 1997 ) , and will therefore test these predictions . deep imaging with future infrared telescopes , such as the space infrared telescope facility ( sirtf ) or the next generation space telescope ( ngst ) , would be able to image pre galactic star clusters . for standard cdm with @xmath53 and a scalo imf , haiman & loeb ( 1997 ) predict that ngst should find @xmath54 star clusters per square arcminute at @xmath55 , given its planned detection threshold of 1 njy in the wavelength range of 13.5 @xmath56 m ( mather & stockman 1996 ) . the characteristic separation between such sources is a few arcseconds , much larger than their typical size of @xmath57 . the early epoch of star formation and metal enrichment is inevitably accompanied by the formation of dust in type ii supernova shells . this dust has two important observational signatures . first , the absorption of starlight energy and its re emission at long wavelengths distorts the spectrum of the cosmic microwave background ( cmb ) radiation ( wright et al . 1994 , and references therein ) . second , the opacity of the intergalactic medium to infrared sources at redshifts @xmath14 could be significant . for these redshifts , infrared in the observer frame corresponds to uv in the source frame a spectral regime in which dust absorption peaks . dust obscuration must therefore be considered when predicting the performance of future infrared telescopes such as ngst . loeb & haiman ( 1997a ) find that for a scalo imf and a uniform mix of metals and dust with the intergalactic gas , the dust distorts the microwave background spectrum by a @xmath58parameter in the range @xmath59@xmath60 , where @xmath61 is the average mass of dust produced per supernova . note that this range is not far below the current cobe limit of @xmath62 ( fixsen et al . the opacity of intergalactic dust to infrared sources at redshifts @xmath14 is significant , @xmath63@xmath64 , and could be detected with ngst . although dust suppresses the ly@xmath65 emission from early sources , the redshifts of star clusters at @xmath935 can be easily inferred from the lyman - limit break in their infrared spectrum between 13.5@xmath56 m ( see fig . as mentioned before , recent microlensing searches ( alcock et al . 1996 ) suggest that @xmath33 of the mass of the milky - way halo might be in the form of massive compact halo objects ( machos ) with a mass @xmath34 . a substantial macho population of dim stars would naturally be linked to the early universe . pre - galactic stars which originate in abundant low - mass systems are expected to behave as collisionless particles and populate the halos of galaxies , just as elementary cdm particles do . only stars which are born in massive rare ( high@xmath66 ) peaks are more likely to end up concentrated in galactic bulges due to the strong dynamical friction which acts on their parent systems . in general , dim stars could constitute the dark matter in galaxy clusters only if @xmath67 and the efficiency of converting gas into these stars is of order 80% . recent observations of the star formation history at @xmath68 might account for most of the luminous stellar population in galaxies today ( madau 1996 ) . however , it is important to remember that this population amounts only to a density parameter @xmath69 ( woods & loeb 1997 ) . therefore , if stellar machos make the dark matter with @xmath70 , then we have not observed yet @xmath71 of the star formation history in the universe . ngst could easily detect stellar machos during their earlier history when they were still bright due to hydrogen burning . for example , chabrier et al . ( 1996 ) and adams & laughlin ( 1996 ) have argued that an old population of white dwarfs could account for the microlensing data and still be consistent with metallicity , star counts , and infrared background constraints , as long as their imf was highly peaked around @xmath72 . but a @xmath73 star has a lifetime @xmath74 gyr during which it is more than an order of magnitude brighter than the sun . even if these stars formed at arbitrarily early times , the age of the universe is only @xmath75gyr at @xmath76 . therefore , the halos of galaxies at @xmath77 should be _ glowing _ in the infrared if dim white dwarfs constitute the dark matter in the present universe . existing hubble deep field observations ( guzmn et al . 1997 ) already place constraints on the extent of luminous halos around galaxies ( loeb & pinsonneault 1997 ) . however , optical observations probe the rest - frame uv and do not see the bulk of the stellar population . infrared observations with ngst would improve dramatically our ability to set constraints on unusual stellar populations in the halos of galaxies . most of the current theoretical modeling of star formation was developed to explain the complex properties of star forming regions in the interstellar medium ( ism ) of our galaxy . the reason for this bias is obvious : these are the environments for which data exists ! the situation resembles a search for a needle in a haystack , just because the haystack happens to be nearby . an alternative strategy is to invest ( nasa ) money in a long - range mission and search for the needle in simpler environments . i would like to argue that the formation process of the first generation of stars in the early universe should be easier to model than that of stars in the local ism . this follows from the fact that the initial conditions in cosmology are simpler and better - defined : ( i ) the universe started from a smooth initial state and a specified spectrum of fluctuations on top of it ; many details of this spectrum will be determined by forthcoming microwave anisotropy experiments ; ( ii ) the composition of the gas is well defined by standard big - bang nucleosynthesis ( schramm & turner 1996 ) , and the chemistry and cooling rates are simplified by the lack of heavy elements , dust , or cosmic rays ; and ( iii ) magnetic fields were probably negligible before stars formed ; most cosmological scenarios result in small field amplitudes ( sigl et al . 1997 , and references therein ) . due solely to lack of data , theoretical studies of the formation of the first stars are sparse in the literature . direct imaging and spectroscopy of the first stars by ngst could stimulate theoretical work on the simplest environment for star formation , namely the early universe . the era of classical cosmology will end if the map and planck satellites determine the basic cosmological parameters to a reasonable accuracy . since much of observational cosmology relies on starlight , cosmologists might then find it appropriate to shift their attention to the fundamental problem of star formation ; and there is no better place to start dealing with it than the early universe . extrapolation of the quasar luminosity function ( pei 1995 ) to redshifts @xmath28 predicts a noticeable tail of faint sources that could be detected with ngst out to the reionization epoch ( loeb & haiman 1997b ) . if these sources are brighter than the first star clusters , it would be easier to measure their absorption spectra and use them to determine the reionization redshift beyond which the gunn - peterson effect appears . bearing in mind that quasars can account for a considerable fraction of the uv background at lower redshifts ( meiksin & madau 1993 ) , it is quite possible that early _ low - luminosity _ quasars were in fact responsible for the reionization of the universe . quasars could be more effective than stars in ionizing the universe because : ( i ) their emission spectrum is harder ; ( ii ) the radiative efficiency of accretion flows could be more than an order of magnitude higher than that of a star ; ( iii ) quasars are brighter , and for a given recombination rate in their host system , the escape fraction of their ionizing photons is higher than for stars . first , let us address the black hole formation process . to form a black hole inside a given dark matter halo , the baryons must cool . for most objects , this is possible only due to atomic line cooling ( see 2 ) , at virial temperatures , @xmath42k , and baryonic masses @xmath78^{3/2}$ ] . after losing their thermal pressure , the cold baryons collapse to a thin disk on a dynamical time ( loeb & rasio 1994 ) . the basic question is then : which fraction of the cold baryons are able to sink to the very center of the potential well and form a massive black hole ? the main barrier to this process is angular momentum . the centrifugal force opposes radial infall and keeps the gas in typical disks at a distance which is 6 - 8 orders of magnitude larger than the schwarzschild radius of the system . however , eisenstein & loeb ( 1995 ) have demonstrated that a small fraction of all objects have a sufficiently low angular momentum so that the gas in them inevitably forms a compact semi - relativistic disk that evolves to a black hole on a short viscous time . these low - spin systems are born in special cosmological environments that exert unusually small tidal toques on them during their cosmological collapse . as long as the initial cooling time of the gas is short and its star formation efficiency is low , the gas would form this compact disk on a free - fall time . in typical systems the baryons dominate gravity inside the scale length of the disk . therefore , if the baryons in a low - spin system acquire a specific angular momentum @xmath79 which is only a tenth of the typical value , then the size of the resulting disk ( @xmath80 ) will be smaller by a factor of @xmath81 than average , and the rotation velocity of the disk will be a factor of @xmath10 larger than the velocity dispersion of the dark matter halo . for @xmath82k , the dark matter halo has a velocity dispersion of @xmath83 , and the low - spin disk would have a characteristic rotation velocity of @xmath84 ( sufficient to retain the gas against supernova - driven winds ) , a size @xmath85 pc , and a viscous evolution time shorter than the hubble time . if a low - spin object of this type is embedded inside an overdense region that will eventually become a galactic bulge , the black hole progenitor would eventually sink to the center of the bulge by dynamical friction and seed quasar activity . based on the phase - space volume accesible to low - spin systems ( @xmath86 ) , one would expect a fraction @xmath87 of all the mass in the universe to be included in such systems ( eisenstein & loeb 1995 ) . however , this is a conservative estimate . additional angular momentum loss due to dynamical friction of gaseous clumps in dark matter halos ( navarro , frenk , & white 1995 ) or bar instabilities in self - gravitating disks ( shlosman , begelman , & frank 1990 ) could only make the black hole formation process more prominent . the popular paradigm that all galaxies harbor black holes at their center ( haehnelt & rees 1993 ) simply _ postulates _ that in all massive systems , a small fraction of the gas ends up as a black hole , but does not explain quantitatively why this fraction obtains its particular small value . observations of galactic nuclei in the local universe imply black hole masses which are typically a fraction @xmath88 of the total baryonic mass of their host ( see summary of dynamical estimates in fig . 5 of kormendy et al . 1997 ; and the maximum ratio between a quasar luminosity and its host mass found by mcleod 1997 ) . this ratio is somewhat higher in low - mass galaxies , such as the compact ellipticals m32 and ngc 4486b . in particular , van der marel et al . ( 1997 ) infer a black hole mass of @xmath89 in m32 , which is a fraction @xmath90 of the stellar mass of the galaxy , @xmath91 , for the central mass - to - light ratio of @xmath92 ; while kormendy et al . ( 1997 ) infer a black hole mass of @xmath93 in ngc 4486b which is a fraction @xmath94 of the stellar mass . figure 1 implies that @xmath95 peaks with a baryonic mass @xmath96 and a velocity dispersion @xmath97 , similar to m32 or ngc 4486b , collapse at @xmath16 in standard cdm . with a @xmath50 star formation efficiency , such objects are @xmath10njy stellar sources for ngst . we assume that each such object forms a black hole with an efficiency , since the central density of the host galaxy grows as @xmath98 while its spin parameter has a negligible dependence on redshift . ] of @xmath99 , i.e. the black hole mass is @xmath100 . if the black hole shines at a fraction @xmath101 of the eddington limit with a characteristic quasar spectrum , then its infrared flux at @xmath16 would be @xmath102 njy . such a flux is comparable to the emission from the host star cluster and would easily be detected with ngst .- ray bursts could reach much higher luminosities in the first month@xmath103 of the burst ( groot et al . 1997 ; sahu et al . 1997 ) ; however , in the local universe such events are rarer by @xmath104 orders of magnitude than supernovae . ] among local galaxies , a fraction @xmath74@xmath105 of all galactic nuclei appear highly active ( huchra & burg 1992 ; maiolino & rieke 1995 ; ho , fillipenko , & sargent 1997 ; khler et al . this fraction can be explained in terms of the ratio between the eddington time and the current hubble time . a black hole which shines at a luminosity @xmath101 in eddington units would double its mass on a timescale , @xmath106 where @xmath107 is the radiative efficiency of the black hole in units of @xmath105 . since the black hole mass , @xmath108 , grows exponentially with the eddington @xmath109-folding time , the characteristic lifetime of the quasar activity is several times @xmath110 , namely @xmath111 of the current hubble time . but notice that for a fixed radiative efficiency , _ the quasar lifetime is independent of redshift _ while the hubble time is shorter at high redshift . for a flat universe , the ratio between @xmath110 and the hubble time @xmath112 is , @xmath113^{3/2 } \epsilon_{0.1}{\tilde l}^{-1}$ ] . therefore several tenths of all star clusters might show nuclear activity at @xmath16 . based on these numbers we infer that quasars would dominate the uv background during reionization if the fraction of gas converted to black holes is @xmath114 . we have suggested that the formation efficiency of black holes in low - mass galaxies at @xmath16 is similar to that found in galaxies in the local universe . it is however possible that the efficiency of black hole formation is substantially reduced in the shallow potential wells of the early population of sub - galactic halos ( haehnelt & rees 1993 ) . observations with ngst will determine whether this is the case or whether quasars indeed pre - dated massive galaxies ( loeb 1995 ) . i thank zoltan haiman , kim mcleod , jerry ostriker , and martin rees for useful discussions , and alexandra schmitz for assistance in the production of the figures . this work was supported in part by the nasa atp grant nag5 - 3085 and the harvard milton fund .	the transition between the nearly smooth initial state of the universe and its clumpy state today occurred during the epoch when the first stars and low - luminosity quasars formed . for cold dark matter cosmologies , the radiation produced by the first baryonic objects is expected to ionize the universe at @xmath0@xmath1 and consequently suppress by @xmath2 the amplitude of microwave anisotropies on angular scales @xmath3 . future microwave anisotropy satellites will be able to detect this signature . the production and mixing of metals by an early population of stars provides a natural explanation to the metallicity , @xmath4 , found in the intergalactic medium at redshifts @xmath5 . the next generation space telescope ( ngst ) will be able to image directly the `` first light '' from these stars . with its njy sensitivity , ngst is expected to detect @xmath6 star clusters per square arcminute at @xmath7 . the brightest sources , however , might be early quasars . the infrared flux from an eddington luminosity , @xmath8 , black hole at @xmath9 is @xmath10 njy at @xmath11 m , easily detectable with ngst . the time it takes a black hole with a radiative efficiency of @xmath2 to double its mass amounts to more than a tenth of the hubble time at @xmath9 , and so a fair fraction of all systems which harbor a central black hole at this redshift would appear active . the redshift of all sources can be determined from the lyman - limit break in their spectrum , which overlaps with the ngst wavelength regime , 1@xmath12 m , for @xmath13 . absorption spectra of the first generation of star clusters or quasars would reveal the reionization history of the universe . the intergalactic medium might show a significant opacity to infrared sources at @xmath14 due to dust produced by the first supernovae .
the discovery of the late - time cosmic acceleration poses one of the greatest challenges theoretical physics has ever faced . in principle , this phenomenon may be the result of unknown physical processes involving either modifications of gravitation theory or the existence of new fields in high energy physics . in the context of einstein s general theory of relativity this means that some sort of dark energy , constant or that varies slowly with time and space , dominates the current composition of the cosmos ( see , e.g. , @xcite for recent reviews ) . among the many possibilities to describe this exotic component , the simplest way is possibly by means of a cosmological constant @xmath0 , which acts on the einstein field equations as an isotropic and homogeneous source with @xmath4 . historically , @xmath0 has been identified with the energy density of the quantum vacuum , but any attempt to obtain such contribution in quantum field theories in curved space - time leads to results many orders of magnitude above the observed value ( @xmath5 @xmath6 ) , and even when an appropriate renormalization procedure is used , the resulting value usually disagrees with the actual one @xcite . this is the case , for example , of the vacuum density of free conformal fields in de sitter space - time , which is shown to be of the order of @xmath7 , where @xmath8 is the hubble parameter @xcite . although such a result can be used to construct non - singular scenarios in the high - energy limit @xcite , at late - times it leads to a very tiny @xmath0 . a possible way out of this problem is to consider the contribution of interacting fields in the low - energy limit , which is a more realistic situation . in this sense , some authors have argued that the qcd chiral phase transition originates a vacuum energy density scaling as @xmath9 , where @xmath10 mev is the energy scale of the qcd vacuum transition @xcite . by taking @xmath11 in the present universe , we obtain @xmath12 , which coincides with the current order of magnitude of the cosmological term . clearly , no theoretical result in this field is free of some assumptions , and a number of other proposals for a possible time dependence of @xmath0 can be found in the literature ( see , e.g. , @xcite ) . in particular , in the case of a vacuum density decaying linearly with @xmath8 , one can show that the corresponding cosmological model is indistinguishable from the @xmath0cdm model during the phase dominated by radiation , but it departs from the standard scenario during the matter epoch , in part owing to the production of dark matter associated to the vacuum decay @xcite . in a previous communication @xcite , we have tested the observational viability of this class of @xmath13cdm scenarios discussed above from a joint analysis involving type ia supernovae data ( supernova legacy survey collaboration @xcite ) along with the position of the first acoustic peak in the cmb spectrum of anisotropies ( @xmath14 ) from wmap @xcite , and the sdss observed distance to the baryonic acoustic oscillation ( bao ) peak at @xmath2 @xcite . the goal of the present contribution is to update the results of ref . @xcite by using three of the most recent sne ia data sets , namely , the updated compilation of union sample ( union2 ) @xcite , the nearby + sdss + essence + snls + hubble space telescope ( hst ) set of 288 sne ia discussed in ref . @xcite ( throughout this paper we refer to this set as sdss compilation ) and the constitution set of 397 sne ia @xcite . we consider two sub - samples of these two latter compilations that use salt2 @xcite and mlcs2k2 @xcite sne ia light - curve fitting methods . along with the sne ia data , and to improve the bounds on the free parameter of the model we use , besides the current value of @xmath14 and the bao peak at @xmath2 , the distance to the bao peak at @xmath1 , as recently discussed in ref . @xcite . for the sake of comparison we also perform the same analyses for the standard @xmath0cdm model . in order to verify the consistence between background and perturbative tests , we also perform a joint analysis including the observed matter power spectrum from the 2dfgrs . in the case of sdss ( mlcs2k2 ) and constitution ( mlcs2k2 17 ) data , we obtain a very good concordance with @xmath13 model . since mlcs2k2 depends weakly on standard cosmology as compared to salt2 , it is actually the most appropriate to test non - standard models , which suggests that the obtained concordance is robust . this paper is organized as follows . in sec . ii we discuss the basic equations and some features of the vacuum decay model with @xmath15 . observational tests involving sne ia , bao , cmb and lss data are performed in sec . iii . in sec . iv we present our main results . we also discuss quantitatively the influence of two sne ia light - curve fitting methods on parameter estimates using the current sdss and constitution compilation sets . we end by summarizing our main conclusion in sec . v. in units where @xmath16 , the friedmann equations for a spatially flat universe can be written as @xmath17 @xmath18 where @xmath19 and @xmath20 are the total energy density and pressure , respectively , and the second equation expresses the conservation of the total energy . by taking the cosmic fluid as composed by dust matter of density @xmath21 and a time - dependent cosmological term with @xmath22 , we have @xmath23 which means that the decay of the vacuum energy density is concomitant with matter production . in order to avoid problems with primordial nucleosynthesis and the spectrum of cmb , we must postulate that no baryonic matter or radiation is produced in this process . therefore , the present model may be considered as a particular case of models with interaction only in the dark sector ( see , e.g. , @xcite and refs . therein for other examples of scenarios with interaction in the dark sector ) . with the ansatz @xmath24 , where @xmath25 is a constant of the order of @xmath26 , the solution of the above equations is given by @xcite @xmath27^{\frac{2}{3}}\;,\ ] ] which coincides with the einstein - de sitter solution in the limit of early times and at late times tends asymptotically to the de sitter space - time . the corresponding hubble parameter as a function of the redshift is given by @xcite @xmath28,\ ] ] where @xmath29 and @xmath30 @xmath31 are , respectively , the current values of the matter density and hubble parameters . note that , similarly to the flat @xmath0cdm scenario , these are the only two parameters of the model to be adjusted by the data . note also that this class of @xmath0(t)cdm cosmologies does not contain the @xmath0cdm scenario as a particular case , which means that in principle they may be observationally distinguishable . in this section we investigate some observational constraints on both models from five different sne ia data sets along with the current value of @xmath14 from wmap , the distance to the bao peak at @xmath1 and @xmath2 and data of large scale structure ( lss ) distribution from 2dfgrs . we use in our analyses five of the most recent sne ia compilations available , namely , the union2 sample of ref . @xcite , the two compilations of the sdss collaborations discussed in ref . @xcite and two of the constitution compilations considered in @xcite . the union2 sample is an update of the original union compilation . it comprises 557 data points including recent large samples from other surveys and uses salt2 for sn ia light - curve fitting . the sdss compilation of 288 sne ia uses both salt2 and mlcs2k2 light - curve fitters and is distributed in redshift interval @xmath32 . this sample , along with the so - called cmb / bao ratio , has been used in ref . @xcite to perform a comparative analysis involving a number of nonstandard cosmological scenarios ( we refer the reader to this latter reference for more on the sdss compilation ) . the constitution compilations comprises 397 sne ia ( cfa3 + union ) calibrated with salt2 and two versions of the mlcs2k2 fitter : mlcs31 and mlcs17 . as discussed in @xcite , the former overestimates host - galaxy extinction , and here we will consider only the later version . the second observable we consider is the distance to a given observed baryonic acoustic oscillation , which is defined as @xmath33^{1/3}\;,\ ] ] where @xmath34 is the comoving angular - diameter distance . we will consider two different measurements of bao , at redshifts @xmath35 and @xmath36 , i.e. , @xmath37@xmath38where the value of @xmath39 at @xmath1 is obtained by combining the results of ref . @xcite with the value of @xmath40 obtained in ref . @xcite . an important aspect worth emphasizing at this point is that we can not use in our analysis the parameter @xmath41 , defined as @xmath42 @xcite , since in the present model the production of matter leads to a different dependence of the horizon at the time of matter - radiation equality on the present matter density @xcite . in all the cases , the observable quantity is the ratio @xmath43 , where @xmath44 is the sound horizon at the time of drag epoch . since @xmath44 is very weakly model - dependent ( provided the radiation epoch is the same as in the @xmath0cdm model ) , the distance @xmath39 can in fact be used in our analysis . as well known , it is also obtained from direct observations by using the @xmath0cdm as a fiducial model @xcite , but such a dependence is weak , and actually this particular observable does not impose strong limits on the model parameter @xcite . the most restrictive of the background tests we will perform in our analysis is the position of the first peak of the cmb spectrum of anisotropies . the simplest way to use it is certainly by considering the shift parameter , defined as @xmath45 , where @xmath46 is the redshift of the last - scattering surface . but again , this is not allowed in our analysis @xcite , since the angular scale @xmath47 is not the same as in the @xmath0cdm case , used in other analyses . ] . we , therefore , use directly the observed position of the peak , @xmath48 @xcite , which is related to the angular scale by @xcite @xmath49 with @xmath50 . here , @xmath51 is the radiation density at the time of last scatteringcdm case . ] . properly speaking , these relations were obtained for the @xmath0cdm case , but they depend only on pre - recombination physics and can be applied to the present model as a good approximation . nevertheless , owing to the process of matter production , the matter density at the time of last scattering is @xmath52 , i.e. , smaller than the @xmath0cdm value by a factor @xmath29 , what can be seen directly from the high-@xmath53 limit of the hubble function ( [ hz ] ) . therefore , in the present case we have @xcite @xmath54 leading to a relation between @xmath47 and @xmath14 different from the @xmath0cdm case . the theoretical expression for the acoustic scale is defined as the ratio @xmath55 with the sound velocity given by @xmath56 , where @xmath57 and @xmath58 stand for the present density parameters of baryons and photons , respectively . the hubble function used in ( [ la ] ) must take into account the radiation contribution @xmath59 and is rewritten as @xcite @xmath60 ^ 2 + \omega_r ( 1+z)^4 \right\}^{1/2}\;.\ ] ] except for scales near the horizon , vacuum perturbations are negligible ( see @xcite ) . therefore , we simplify our analysis by considering an unperturbed time - dependent cosmological term . the model prediction ( blue line ) together with the 2dfgrs data points @xcite are shown in the left panel of figure 6 . for comparison , the concordance @xmath0cdm fitting as provided by the bbks transfer function @xcite is also shown ( red line ) , as well as the result for the @xmath0cdm case of the same simplified analysis we have performed with @xmath0(t)cdm ( black line ) . in such analyses baryons are not included , which leads to an error of about @xmath61 as compared to a more exact study @xcite . in the case of @xmath0cdm , the best fit is given by @xmath62 , in agreement with @xcite . 0.1 in 0.1 in 0.1 in 0.1 in 0.1 in rrll + & & & + + union2(salt2 ) ........ & @xmath63 & 554.9 & 0.998 + union2(salt2) ...... &@xmath64 & 634.5 & 1.063 + sdss ( salt2) ....... &@xmath65 & 256.9 & 0.895 + sdss ( salt2)@xmath66 ....... &@xmath67 & 320.8 & 0.978 + sdss ( mlcs2k2 ) ....... & @xmath68 & 249.7 & 0.870 + sdss ( mlcs2k2)@xmath66 ........ &@xmath69 & 276.3 & 0.842 + constitution ( salt2 ) ....... & @xmath70 & 343.3 & 0.984 + constitution ( salt2)@xmath66 ....... & @xmath71 & 390.2 & 0.998 + constitution ( mlcs2k2 ) ....... & @xmath72 & 397.2 & 1.074 + constitution ( mlcs2k2)@xmath66 ....... & @xmath73 & 435.3 & 1.057 + rrll + & & & + + union2(salt2 ) ........ & @xmath74 & 552.7 & 0.994 + union2(salt2) ...... &@xmath75 & 613.1 & 1.027 + sdss ( salt2) ....... &@xmath76 & 256.0 & 0.892 + sdss ( salt2)@xmath66 ....... &@xmath77 & 305.8 & 0.932 + sdss ( mlcs2k2) ....... &@xmath78 & 249.1 & 0.868 + sdss ( mlcs2k2)@xmath66 ........ &@xmath79 & 403.8 & 1.231 + constitution ( salt2 ) ....... & @xmath80 & 341.7 & 0.979 + constitution ( salt2)@xmath66 ....... & @xmath81 & 515.8 & 1.319 + constitution ( mlcs2k2 ) ....... & @xmath82 & 397.4 & 1.074 + constitution ( mlcs2k2)@xmath66 ....... & @xmath83 & 570.2 & 1.384 + in order to derive the constraints on our model we maximaze the likelihood function @xmath84 , where @xmath85 takes into account all the data sets discussed above . in our analyses , the hubble parameter @xmath86 is considered as a nuisanse parameter so that we marginalize over it . the results are shown in figs . 1 - 5 , and tables i and ii summarize the main results of this paper for all sne data sets and joint analyses . in order to compare the two classes of models discussed in this paper we also show the values of @xmath87 and @xmath88 ( @xmath89 stands for degrees of freedom ) for each analysis performed . in the figures , the left and central panels show the superposition of the @xmath90 and @xmath91 confidence regions in the @xmath92 plane for the four tests we are considering , as well as the concordance ellipses of our joint analyses . in the case of sdss ( mlcs2k2 ) , the interval for the matter density in @xmath0(t)cdm is about @xmath61 higher than the flat @xmath0cdm value . both values , on the other hand , are higher than the usually accepted concordance value @xmath93 @xcite . we also note that in both cases the value of @xmath86 is relatively small ( @xmath94 ) compared to current estimates @xcite . nevertheless , when we marginalize over @xmath95 the matter density remains unchanged in both models . the right panels of the figures show the likelihood for @xmath29 after marginalization . in all the analyses the estimate of the matter density parameter for the @xmath0(t)cdm scenario is considerably larger than the corresponding value for the @xmath0cdm model . in particular , the estimates of @xmath29 for the vacuum decay model is up to @xmath96 off from the central value inferred by applying the virial theorem to cluster dynamics @xcite ( see also @xcite ) , but only @xmath97 off from the central value obtained by using other independent methods , such as the mean relative peculiar velocity measurements for pairs of galaxies @xcite . in terms of parameter estimation we note that very similar results are found for both union2 and sdss compilation with salt2 . an interesting aspect worth emphasizing is realized when a comparison between the analyses of figs . 2 and 3 is made . note that , although the only difference between these two analyses is the light - curve fitter used in the sne calibration , the estimated values of the cosmological parameter @xmath29 are considerably modified from analysis to analysis . for instance , by comparing only the sne ia bounds on @xmath29 for the standard @xmath0cdm model , we observe that the estimated values of the matter density parameter increases @xmath98 , from 0.27 to 0.40 ( see table ii ) . the same can be said about the constitution sample calibrated with salt2 and mlcs17 . similar conclusions are also drawn for the class of vacuum decay models studied here . in this case , estimates of the matter density parameter increase @xmath99 for the sdss sne ia data with mlcs2k2 light - curve fitter , in better accordance with the independent estimate based on large scale structure distribution @xcite . it is worth observing that these estimated values of @xmath29 obtained from sdss ( mlcs2k2 ) analyses for both models are relatively large , being @xmath96 off from the central value obtained using the cluster dynamics methods discussed earlier . in all the cases above we have considered spatially flat scenarios , usually predicted by inflation . for completness , we can relax this prior and include spatial curvature in both models . in the right panel of figure 6 we show the results for the sdss ( mlcs2k2 ) data set @xcite . we can see that the spatially flat universe is in accordance with data within @xmath90 confidence level in both models . although in good agreement with current observations , a residual @xmath0 term exacerbates the well known cosmological constant problem , requiring a natural explanation for its small , but nonzero , value . in this paper , we have discussed the observational viability of a class of alternative @xmath13cdm scenarios in which the vacuum energy density decays linearly with the hubble parameter @xmath8 . by considering the most recent sne ia compilations sets along with the current measurements of distance to the bao peaks at @xmath1 and @xmath2 and the position of the first acoustic peak of the cmb power spectrum , we show that in terms of @xmath3 statistics both models provide good fits to the data with the very same number of free parameters ( see tables i and ii ) . a recurrent aspect in all the analyses performed is the larger estimates of @xmath29 for @xmath13cdm relative to the values obtained in the @xmath0cdm context . when compared with some @xmath29 estimates from independent observational methods @xcite , these values may be off from @xmath97 up to @xmath96 . the same conclusion also arises in the @xmath0cdm context for the tests including sdss ( mlcs2k2 ) . we have also discussed quantitatively differences in parameter estimates between salt2 and mlcs2k2 light - curve fitters using the current sdss and constitution sne ia compilations . we have shown that for both models these differences may reach up to @xmath98 for the matter density parameter . these results reinforce the need and importance for a better understanding of unresolved systematic effects on sne ia calibration , as has been discussed by other authors ( see , e.g. , @xcite ) . finally , from a joint analysis which includes the observed matter power spectrum , it is clear that a good concordance is achived for the @xmath0(t)cdm model when we use the sdss ( mlcs2k2 ) and constitution ( mlcs17 ) compilations of sne ia . since the mlcs2k2 fitter depends weakly on standard cosmology , it is generally considered the most appropriate to test alternative models , which means that the concordance obtained here can be considered robust . actually , an inspection of panels ( c ) of figures 1 - 5 shows that , in the case of mlcs2k2 fitter , the @xmath13 model presents better concordance than the standard model . the authors acknowledge cnpq - brazil for the grants under which this work was carried out . jsa is grateful for the support of cnpq grants 304569/2007 - 0 and 481784/2008 - 0 . sc is thankful for the cnpq grant 305133/2008 - 0 . v. sahni and a. a. starobinsky , int . j. mod d9 * , 373 ( 2000 ) ; p. j. e. peebles and b. ratra rev . phys . * 75 * , 559 ( 2003 ) ; t. padmanabhan , phys . rept . * 380 * , 235 ( 2003 ) ; e. j. copeland , m. sami and s. tsujikawa , int . d15 * , 1753 ( 2006 ) ; j. s. alcaniz , braz . j. phys . * 36 * , 1109 ( 2006 ) [ arxiv : astro - ph/0608631 ] . l. h. ford , phys . rev . * d11 * , 3370 ( 1975 ) ; j. s. dowker and r. critchley , phys . rev . * d13 * , 3224 ( 1976 ) ; p. c. w. davies , phys . lett . * b68 * , 402 ( 1977 ) ; a. a. starobinsky , phys . lett . * b91 * , 99 ( 1980 ) . r. schutzhold , phys . 89 * , 081302 ( 2002 ) ; f. r. klinkhamer and g. e. volovik , phys . d * 79 * , 063527 ( 2009 ) ; f. r. urban and a. r. zhitnitsky , phys . rev . * d80 * , 063001 ( 2009 ) ; phys b688 * , 9 ( 2009 ) ; nucl . phys . * b835 * , 135 ( 2010 ) ; n. ohta , arxiv : 1010.1339 . see also s. banerjee _ et al . _ , phys . lett . * b611 * , 27 ( 2005 ) . m. @xmath100zer and m. o. taha , phys . lett . * b171 * , 363 ( 1986 ) ; nucl . b287 , 776 ( 1987 ) ; o. bertolami , nuovo cimento soc . ital . fis . * b93 * , 36 ( 1986 ) ; k. freese _ et al _ , nucl . phys . * b287 * , 797 ( 1987 ) ; w. chen and y - s . wu , phys . * d41 * , 695 ( 1990 ) ; m. s. berman , phys . rev . * 43 * , 1075 ( 1991 ) ; d. pavn , phys . rev . * d43 * , 375 ( 1991 ) ; j. c. carvalho , j. a. s. lima and i. waga , phys . rev . * d46 * 2404 ( 1992 ) ; a. i. arbab and a. m. m. abdel - rahman , phys . rev . * d50 * , 7725 ( 1994 ) ; j. a. s. lima and j. m. f. maia , phys . rev * d49 * , 5597 ( 1994 ) ; j. m. overduin and f. i. cooperstock , phys . rev . * d58 * , 043506 ( 1998 ) ; j. m. overduin , astrophys . j. * 517 * , l1 ( 1999 ) ; j. shapiro and j. sola , phys . lett * b475 * , 236 ( 1999 ) ; m. v. john and k.b . joseph , phys . rev . * d61 * , 087304 ( 2000 ) ; o. bertolami and p. j. martins , phys . rev . * d61 * , 064007 ( 2000 ) ; l. amendola , phys * d62 * 043511 ( 2000 ) ; l. amendola and d. tocchini - valentini , phys . rev . d * 64 * , 043509 ( 2001 ) ; r. g. vishwakarma , gen . * 33 * , 1973 ( 2001 ) ; a. s. al - rawaf , mod . a14 * , 633 ( 2001 ) ; m. k. mak , j. a. belinchon , and t. harko , ijmp * d11 * , 1265 ( 2002 ) ; i. l. shapiro and j. sola , jhep * 0202 * , 006 ( 2002 ) ; w. zimdahl and d. pavn , gen . . grav . * 35 * , 413 ( 2003 ) ; m. r. mbonye , ijmp * a18 * , 811 ( 2003 ) ; j. s. alcaniz and j. m. f. maia , phys . * d67 * , 043502 ( 2003 ) . [ astro - ph/0212510 ] ; i. l. shapiro , j. sola , c. espana - bonet , and p. ruiz - lapuente , phys . b574 * , 149 ( 2003 ) ; r. horvat , phys . rev . * d70 * , 087301 ( 2004 ) ; c. espana - bonet et al . jcap * 0402 * , 006 ( 2004 ) ; p. wang and x. meng , class . * 22 * , 283 ( 2005 ) ; s. carneiro and j. a. s. lima , ijmp * a20 * 2465 ( 2005 ) ; i. l. shapiro , j. sola , and h. stefancic , jcap * 0501 * , 012 ( 2005 ) ; e. elizalde , s. nojiri , s. d. odintsov , and p. wang , phys . rev . * d71 * , 103504 ( 2005 ) ; r. aldrovandi , j. p. beltrn almeida , and j. g. pereira , grav . & cosmol . * 11 * , 277 ( 2005 ) ; f. bauer , class . quant . grav . * 22 * , 3533 ( 2005 ) ; j. s. alcaniz and j. a. s. lima , phys . d * 72 * , 063516 ( 2005 ) ; j. d. barrow and t. clifton , phys . d * 73 * , 103520 ( 2006 ) ; b. wang , c. y. lin , and e. abdalla , phys . b637 * , 357 ( 2006 ) ; j. sola and h. stefancic , mod . * a21 * , 479 ( 2006 ) ; a. e. montenegro jr . and s. carneiro , class . . grav . * 24 * , 313 ( 2007 ) ; f. e. m. costa , j. s. alcaniz and j. m. f. maia , phys . d * 77 * , 083516 ( 2008 ) ; g. caldera - cabral , r. maartens and l. a. urena - lopez , phys . d * 79 * , 063518 ( 2009 ) ; j. f. jesus , r. c. santos , j. s. alcaniz and j. a. s. lima , phys . d * 78 * , 063514 ( 2008 ) ; q. wu , y. gong , a. wang and j. s. alcaniz , phys . b * 659 * , 34 ( 2008 ) ; f. e. m. costa , e. m. . barboza and j. s. alcaniz , phys . d * 79 * , 127302 ( 2009 ) ; f. e. m. costa and j. s. alcaniz , phys . d * 81 * , 043506 ( 2010 ) ; m. m. phillips , astrophys . j. * 413 * , l105 ( 1993 ) ; a. g. riess , w. h. press , and r. p. kirshner , astrophys . j. * 438 * , l17 ( 1995 ) ; s. jha , a. g. riess , and r. p. kirshner , astrophys . j. * 659 * , 122 ( 2007 ) . a. dekel , d. burstein and s. d. white , `` measuring omega '' , in critical dialogues in cosmology ( proceedings of the princeton 250th anniversary conference , june 1996 ) , ed . n. turok ( world scientific ) . n. a. bahcall and x. fan , publ . sci . * 95 * , 5956 ( 1998 ) .	we investigate the observational viability of a class of cosmological models in which the vacuum energy density decays linearly with the hubble parameter , resulting in a production of cold dark matter particles at late times . similarly to the flat @xmath0cdm case , there is only one free parameter to be adjusted by the data in this class of @xmath0(t)cdm scenarios , namely , the matter density parameter . to perform our analysis we use three of the most recent sne ia compilation sets ( union2 , sdss and constitution ) along with the current measurements of distance to the bao peaks at @xmath1 and @xmath2 and the position of the first acoustic peak of the cmb power spectrum . we show that in terms of @xmath3 statistics both models provide good fits to the data and similar results . a quantitative analysis discussing the differences in parameter estimation due to sne light - curve fitting methods ( salt2 and mlcs2k2 ) is studied using the current sdss and constitution sne ia compilations . a matter power spectrum analysis using the 2dfgrs is also performed , providing a very good concordance with the constraints from the sdss and constitution mlcs2k2 data .
among the so - called frank - kasper phases that can be geometrically described in terms of the basic coordination polyhedron with coordination number equal to 12 , 14 , 15 and 16 @xcite , the @xmath0-phase that can occur in transition - metal alloy systems , is known as the one without definite stoichiometric composition . consequently , it can exist in different composition ranges and , therefore , its physical properties can be tailored by changing its constitutional elements , and within a given constitution , by changing chemical composition . among about 50 examples of binary @xmath0-phases that have been reported so far in the literature , only two viz . @xmath0-fecr and @xmath0-fev are known to have well evidenced magnetic properties @xcite , though a knowledge of the @xmath0-phase magnetism remains not complete . in particular , neither values of the magnetic moments , @xmath5 , on particular lattice sites nor a coupling between them a lack of stoichiometry , leading to a huge number of different atomic configurations , combined with a weak magnetism i.e. small values of @xmath5 makes it very difficult for microscopic methods like mssbauer spectroscopy or nuclear magnetic resonance to uniquely determine spin - densities at the level of the sublattices . non - availability of big enough single - crystals of the @xmath0-phase prevents the neutron diffraction techniques to be used for that purpose . in these circumstances , performing theoretical calculations aimed at determining electronic and magnetic structures of the phase is highly needed and helpful . the present paper reports such calculations for the fe@xmath1v@xmath2 system where the @xmath0-phase exists in a wide range of composition ( @xmath3 ) , and its magnetic properties show a strong @xmath6-dependence . for example , the curie temperature may vary between 0 k , for @xmath7 , and @xmath8320 k , for @xmath9 @xcite . this gives a unique chance for testing the calculations which were already successfully applied for similar purpose in the @xmath0-fecr alloy system @xcite where the phase exists in a much narrower compositional and temperature ranges . electronic structure calculations for @xmath10fev compounds have been performed using the charge and spin self - consistent korringa - kohn - rostoker ( kkr ) method . the muffin - tin crystal potential has been employed within the local density approximation ( lda ) framework , where the von barth - hedin formula @xcite for the exchange - correlation part was used . the self - consistent crystal potentials were converged below 0.1 mry and charges below @xmath11 . the spin - polarized densities of states were computed employing the tetrahedron * k*-space integration technique . the local magnetic moments ( of fe and v ) inside each muffin - tin spheres were determined . accordingly , the fermi contact hyperfine fields ( @xmath12 ) resulted from the spin densities extrapolated to @xmath13 , @xmath14 , with the use of the well - known non - relativistic formula @xmath15 in koe @xcite . the kkr calculations allowed to determine core ( @xmath16 ) and valence ( @xmath17 ) terms , neglecting however dipolar and orbital contributions to the total hyperfine field . it is worth noting that electronic structure calculations have been done for ordered approximants of the disordered @xmath10fev alloys , and for that purpose the symmetry of the unit cell was lowered to a simple tetragonal one . in practice , the tetragonal unit cell ( space group p@xmath18/mnm ) and atomic positions were unchanged but variable occupancy made all 30 atomic positions crystallographically nonequivalent ( space group p@xmath19 ) being next occupied exclusively either by fe or v atom . calculations have been performed for two @xmath0-fev alloys with different compositions , namely fe@xmath20v@xmath21 and fe@xmath22v@xmath23 . they are equivalent to the formula fe@xmath1v@xmath24 with @xmath25 and @xmath26 , respectively . the chosen compositions correspond fairly well to border concentrations of the @xmath0-phase existance in the fe - v alloy system @xcite . fe and v atoms were distributed over five inequivalent sublattices of the tetragonal unit cell in close agreement with experimentally determined occupancies @xcite . actually , the used here atomic configurations in the unit cell remained strictly the same as those considered previously in electronic structure calculations of the @xmath0-fev phase in a paramagnetic state @xcite . in order to study magnetic properties from first principles calculations , two collinear models of a possible magnetic ordering were taken into account : ( i ) parallel alignment of local magnetic moments , called here ferromagnetic , fm , as derived directly from spin - polarized calculations , and ( ii ) antiparallel alignment , called here apm . in the latter , fe / v atoms in selected sublattices were initially coupled antiferromagnetically according to predictions of a symmetry analysis as given in details elsewhere @xcite . the magnetic moments and hyperfine fields , @xmath12 , corresponding to both fe and v atoms distributed over the five sublattices were a subject of a further analysis . their non - zero values were found for both fe and v atoms . as a rule , the magnetic moments ( and the hyperfine fields ) on v atoms were found to be much smaller , and , generally , polarized opposite , to those on fe ones . the calculated values of @xmath5 and @xmath27 for each sublattice were analyzed versus a number of fe atoms in the nearest neighborhood , @xmath4 . such approach combining the results obtained for the two border concentrations , shows that the two sets of the computed data are complementary , and they cover the expected range of @xmath4 for all sublattices . a detailed analysis of the @xmath28-dependence of the magnetic quantities for each of the sublattices shows that this relationship can be split , in most cases , into two regions showing a linear character . in the investigated @xmath0-fev system , an average magnetic moment per an atom , @xmath29 , on a given sublattice remains zero up to a critical value of @xmath30 , and for @xmath31 it varies linearly in accord with the equation @xmath32 a similiar dependence on @xmath4 was also obtained for the average @xmath27 , @xmath33 , for all sublattices , and both considered models of the magnetic ordering ( fm and apm ) . noteworthy , a similar discontinous @xmath28-dependence was already reported in the analysis of the atomic moment in _ bcc _ fe - v alloys @xcite , and also used in calculations of the hyperfine field distribution curves for nanocrystalline _ bcc _ fe - v alloys @xcite . the critical value of @xmath4 as found by these authors was rather small , @xmath34 , when accounting for the magnetic moments localized on fe - atoms . in the present case of the @xmath0-phase , @xmath35 values and slopes @xmath36 of the lines were found by a fitting the data calculated for fe and v atoms located on five various sublattices , separatelly . the obtained @xmath35-values in that way do not exceed 6 , as can be clearly seen in figs . [ fig1 ] and [ fig2 ] . it is worth noting that negative values of @xmath35 , as found in several cases , mean that the magnetic moments do not vanish for all @xmath4-values , as derived from eq . 1 . the calculated magnetic moments for both types of atoms located on all crystallographic positions combined with calculated probabilities of these position occupancies allowed to determine the average values of the magnetic moments , @xmath29 , for each sublattice and for any concentration of vanadium . in the same way , one could compute the average magnetic moment per atom ( per unit cell ) . in practice , such calculations were performed for four selected concentrations that corresponded to the samples on which the magnetic measurements were performed . a comparison between the calculated and experimentally determined quantities is displayed in fig . [ fig3 ] . it can be clearly seen that the values of @xmath29 calculated applying the fm model are systematically larger than those found experimentally , @xmath37 . the difference between the former and the latter changes from about @xmath38 per atom for @xmath39 to @xmath40 for @xmath41 . however , the overall dependence of @xmath29 on the vanadium content follows well that found experimentally . on the other hand , the @xmath29-values derived from the apm model are underestimated in comparison to those of @xmath37 but the agreement bewteen theoretical and experimental values markedly improves with increasing v content . for the vanadium richest sample , fe@xmath42v@xmath43 , the measured value of @xmath44 is practically the same as the one calculated with the apm model , @xmath45 . the relationship between the measured and calculated values of @xmath29 can be explained in terms of a total energy of the @xmath0-phase , @xmath46 , calculated for the two models of magnetic of ordering ( @xmath47 and @xmath48 ) . the energy difference @xmath49 , was determined using the electronic structure calculations for the two border concentrations of the @xmath0-phase within both fm and apm models . in the case of the @xmath0-fe@xmath20v@xmath21 alloy @xmath50-value ranges from @xmath51 to @xmath52 mry per atom , that would favour the fm model . for the @xmath0-fe@xmath22v@xmath23 alloy the @xmath50-value almost vanishes ( less than 0.05 mry ) . the total energy difference of the @xmath0-phase can be compared to kkr results obtained for fm and ( hypothetical ) afm ordering in a simple @xmath53-phase with the same fe - v composition . it yields the value of @xmath54 mry per atom ( also favoring fm ordering ) , being thus an order of magnitude larger than in the former case . hence , the small difference between total energies computed for the fm and apm structures may suggest that some regions of the sample have the antiparallel magnetic order ( apm ) , whereas in others the magnetic moments arrange mostly ferromagnetically ( fm ) . in other words , the magnetic properties of the @xmath0-fev can be thought of as a coexistence of two types of the ordering . the relative contribution of them should depend on the alloy composition . in view of the total energy analysis , the contribution of the apm - like regions should be increasing with the vanadium concentration , @xmath6 , since the @xmath50 value decreases with @xmath6 . the latter effect is also observed experimentally in the dependence @xmath55 - fig . [ fig3 ] . on the other hand , it should be remembered that both fm and apm models do not strictly refer to the ferromagnetic and antiferromagnetic ordering of fe / v magnetic moments in the @xmath0-phase . they should be rather treated as starting configurations , and finally converged magnetic moment values and their orderings may differ from the initial alignments . in consequence , magnetic moments on fe / v atoms may vary both in magnitude and in a mutual coupling , also due to the fact that the @xmath0-fev phase is a chemically highly disordered system . in particular , it is possible that the real magnetic ordering is not collinear . a necessary condition for the occurence of spin - canting i.e. competition of ferromagnetic and antiferromagnetic interactions , has been here , and also in the case of the @xmath56 @xcite theoretically suggested to exist . , versus average magnetic moment per atom , @xmath5 , in the @xmath0-fev phase , as measured and calculated in fm and apm models . ] another quantity pertinent to magnetic materials that has been calculated is a hyperfine field , @xmath12 . although its correlation with the magnetic moment is usually non - linear and , in some instances , even non - monotonous @xcite , in the case of the @xmath0-fev alloys @xmath27 and @xmath5 exhibit a perfect linear correlation @xcite . in addition , for the @xmath0-fev compounds the hyperfine fields had been measured both on @xmath57fe and @xmath58v @xcite , so they can be used as a proper data set for testing pertinent theoretical calculations . those obtained presently by means of the kkr calculations for the fm and apm models are shown in fig . [ fig4 ] and [ fig5 ] , respectively . as can be clearly seen one can distinguish two rangies for all sublattices and occupations , namely zero - field range ( @xmath59 ) for small values of @xmath4 , and proportionality range ( @xmath60 ) for higher @xmath4-values . in order to compare the calculated @xmath12-values with the experimental ones ( mssbauer and nmr data ) we consider only the absolute values , @xmath27 . concerning the fe - site hyperfine fields , we can make a comparison - see fig . [ fig6]a - only between their average values , @xmath61 , due to the fact that experimentally only such quantity is available from the @xmath57fe - site mssbauer spectroscopic measurements @xcite . as it follows from fig . [ fig6]a , the average hyperfine field on fe atoms as calculated in the frame of the fm model decreases linearly with @xmath6 from @xmath62 koe for @xmath39 to @xmath63 koe for @xmath64 . the corresponding @xmath61-values obtained by means of the apm model also decrease linearly with @xmath6 from @xmath65 koe to @xmath66 koe , but at a slightly lower rate . the measured values of @xmath67 show a non - linear behavior . for the lowest v concentrated sample ( @xmath39 ) its value lies between those found for the fm and apm models . for @xmath68 it is equal to the value calculated with the apm approach , while for higher @xmath6-values it is smaller than the calculated values . the observed quantitative agreement between the measured and the calculated values of the hyperfine fields is not fully satisfying what can have different reasons . first of all , we should bear in mind that in our present calculation only the fermi contact term was taken into account , while other possible contributions like dipolar and orbital ones , were neglected . but we do expect that the most important influence on the computed @xmath12-values could have a local ordering of magnetic moments which mostly modify its valence terms . it is well seen on the dependencies of @xmath16 and @xmath69 contributions vs. magnetic moments for fm model ( fig . [ fig7 ] ) . as expected , @xmath16 is perfectly linear vs. @xmath5 with the expected slope of about @xmath70 ( both for fe and v atoms whatever the sublattice ) , while for @xmath69 there is no apparent correlation with the magnetic moment , since it strongly depends on atomic and magnetic surroundings . in a real magnetic structure , valence contributions to @xmath12 may vary both in magnitude and in direction . the latter is not provided by the present calculations when using simplified fm and apm models . it is worth noting that for fe atoms the calculated @xmath12 are dominated by @xmath16 ( best seen on a , d and e sites in fig . [ fig7 ] ) , which are fully correlated with local magnetic moments . since the calculated hyperfine fields overestimate the experimental data , one should expect that the valence contributions , which are opposite in sign to the core contributions , may diminish this disagreement . hence , accounting for a more complex magnetic structure ( with different local magnetic moments directions that mostly affect @xmath69 ) could improve experiment versus theory agreement for the results on fe atoms . on the other hand the correlation between the calculated average fe - site hyperfine field and the corresponding magnetic moment ( per atom ) - see fig . [ fig8 ] - agrees quite well with the measured one : the slope of the experimentally found @xmath71 line has the average value of @xmath8 19 t/@xmath72 if the magnetic moment is calculated per an atom in the unit cell ( or 14.3 t/@xmath72 when the magnetic moment is determined per one fe - atom in the unit cell @xcite ) , while that of the theoretically determined lines amounts to @xmath814 t/@xmath72 for the fm ordering and to @xmath827 t/@xmath72 for the apm one . in other words , the experimentally found @xmath71 line lies between the two theoretically calculated ones . this can be taken as a further support for our approach that the magnetic ordering in the studied system can be treated as a combination of fm- and apm - ordered regions . concerning the @xmath58v fields , we have calculated the fields for four sublattices ( b , c , d , and e ) as we can compare these quantities with experimentally determined ones for each of the five sublattices @xcite . in addition , the average values of the v - site field , @xmath73 , were evaluated , too . a comparison between the theoretical and experimental data , displayed in fig . [ fig6]b , gives an evidence that the fields calculated for particular sites are underestimated by a factor of @xmath74 . in consequence , the average field , namely the calculated @xmath73 decreases with @xmath6 at a lower rate than the measured one . using the same arguments as for the fe - site hyperfine fields , we expect that @xmath69 are not sufficiently well evaluated within the fm and the apm models . [ fig7 ] shows that absolute values of @xmath69 are slightly larger than the corresponding ones of @xmath16 , resulting in relatively small total hyperfine fields on v. consequently , the calculated hyperfine fields became underetimated with respect to the experimental ones . fe mssbauer spectra recorded on the ( a ) @xmath0-fe@xmath75v@xmath76 and ( b ) @xmath0-fe@xmath77v@xmath78 sample at 4.2 k. dashed lines represent best fits to the data calculated within the fm and the apm models . ] the calculations reported in this paper can be also used in the analysis of mssbauer spectra . towards this end we assumed the mssbauer spectrum to be composed of five subspectra correspondig to the @xmath57fe atoms located on the five inequivalent lattice sites . the relative area of each subspectrum should be proportional to the number of fe atoms occupying a given site and it was assumed to be equal to that known from neutron diffraction experiments @xcite . each subspectrum represents a distribution of the hyperfine fields ( hfd ) with quadrupole splittings , @xmath79 , and relative isomer shift - values , @xmath80 , identical to those determined from room temperature measurements and presented in ref . . the hyperfine field distributions for each concentration were computed on the basis of probability distributions and calculated @xmath81 dependences obtained from fm and apm models . as can be clearly seen in fig . [ fig9 ] , the mssbauer spectrum recorded on the sample with @xmath39 could be cuccessfuly fitted assuming the hfd to be weighted average of hfd obtained from the fm model ( weigt 58% ) and the apm one ( 42% ) . for the sample with @xmath68 it was enough to take into accout the hfd calculated for the apm model , only . for more concentrated samples it was not possible to get any reasonably good fit on the basis of the calculated values . the electronic structure calculations of ordered approximants of the @xmath0-fev phase were done for border compositions of this phase , namely fe@xmath20v@xmath21 and fe@xmath22v@xmath23 . the magnetic properties , i.e. total and atomic magnetic moments , as well as fe and v hyperfine fields were calculated using two models for the magnetic moments alignment and compared to experimental values . it was found that both @xmath28 and @xmath27 are zero below @xmath35 , characteristic of each lattice site , and increase more or less linearly for higher @xmath4-values . the calculated @xmath29-values overestimate the experimental data when using the fm model , and underestimate them in the case of the apm one . this discrepancy was discussed in terms of the total energy difference as calculated for the fm and the apm models . the analysis of the calculated fermi contact terms of fe and v hyperfine fields showed a crucial role of valence contributions , being strongly dependent on local magnetic moments arrangements . the fact that only simple magnetic structures were accounted for the kkr calculations , did not allow to expect fully realistic determination of @xmath69 . the enhancement of the @xmath69 contribution which either lowers @xmath27 on fe atoms and increases this quantity on v atoms , would eventually lead to a better agreement with the experimental data . in spite of the aforementioned discrepancies between experimental and calculated values , the almost linear dependences of @xmath82 , determined in @xmath0-fev phase for the fm and the apm models , remains in satisfying agreement with measured one . the results reported in this study were partly obtained within the project supported by the ministry of science and higher education , warsaw ( grant no . n n202 228837 ) , by the european communities under the contract of association between euratom and ipplm within the framework of the european fusion development agreement as well as by the fp7-nmp-2010-large-4 collaborative project `` accmet '' ( no .	electronic structure korringa - kohn - rostoker calculations for the @xmath0-phase in fe@xmath1v@xmath2 were performed in compositional range of its occurance ( @xmath3 ) . fe and v magnetic moments and hyperfine fields were determined for five inequivalent lattice sites in two models of magnetic structure , namely ferromagnetic fm and so - called anti - parallel one , apm , dominated by antiferromagnetic coupling . the average magnetic moments calculated for fm state overestimate the experimental data , whereas the corresponding quantities computed for apm state , underestimate them . such a behavior remains in line with total energy values being similar for both models . the calculations showed that both average magnetic moments and hyperfine fields ( on fe and v atoms ) vary with a number of fe atoms in the nearest neighbor shell , @xmath4 , starting from critical values of @xmath4 , characteristic of each site . the calculated hyperfine fields for fe and v showed an important role of valence contributions , being strongly dependent on local magnetic moments arrangements . in the case of fe , the computed hyperfine fields are overestimated with respect to the mossbauer data , while the corresponding values for v are underestimated with respect to the nmr results . however , the linear correlation between average magnetic moment and hyperfine fields , observed experimentally in fev @xmath0-phase , can be well reproduced when combining theoretical results for the two above - mentioned magnetic structure models .
the search for evidence of the origin of galactic cosmic rays has been an active area of research for many decades . while little evidence exists about the sites at which very high energy nuclei are accelerated , the results of recent x - ray and gamma - ray observations indicate that at least some of the cosmic - ray electrons are accelerated in the shocks of supernova remnants @xcite . for example , sn 1006 is one remnant for which there is evidence that cosmic - ray electrons have been accelerated to energies as high as about 100 tev @xcite . in this paper , measurements of the x - ray and radio emission of the remnant are used to determine the parameters of the nonthermal electron spectrum . although there is no evidence to indicate that cosmic - ray nuclei are accelerated in sn 1006 , relativistic electrons and nuclei are expected to be accelerated in a similar manner @xcite . therefore , we estimate the parameters of the proton and helium spectra of the remnant . the results of this analysis show that the remnant is a significant source of galactic cosmic rays ( at least cosmic - ray electrons ) and provide some support for the idea that galactic cosmic rays are accelerated predominantly in the shocks of supernova remnants . section [ obs ] contains a description of some of the results of previous analyses of the x - ray data of sn 1006 . the data and analysis techniques used for the present work are described in section [ dat ] . section [ dis ] includes a discussion of the results of the present analysis . the conclusions are reviewed in section [ con ] . the first detection of x - ray emission from sn 1006 was reported by @xcite . the results of analyses of x - ray spectral data of the remnant have been reported by many authors @xcite . as early as 1979 , it was realized that the global x - ray spectrum of sn 1006 can not be fitted with a single thermal or nonthermal model @xcite . this realization is not surprising in retrospect because observations of the remnant with imaging x - ray devices show that the x - ray spectrum varies as a function of position in the remnant . the high - energy x - ray image of the remnant , like the radio image , is dominated by emission from the northeastern and southwestern rims @xcite . since the spectra of the rims are more or less featureless , the emission from these regions is reported to be nonthermal @xcite . at energies of less than 0.5 kev , a significant amount of emission is observed from the interior of the remnant , especially in the southwest @xcite . since the x - ray spectrum of the interior exhibits atomic emission - line features , a substantial amount of the emission from this region must be thermal bremsstrahlung . several of the spectral analyses of sn 1006 included two spectral components . invariably , one of the spectral components is thermal emission from an optically thin plasma . such a model is needed to describe the emission - line features of oxygen , neon , magnesium , silicon , and perhaps iron @xcite . some analyses have included an additional thermal component , while others have included a nonthermal component . for example , hamilton et al . ( 1986 ) modeled the spectrum using two thermal components . one component is used to describe the forward - shocked interstellar medium , and the other describes the reverse - shocked ejecta . to explain the nearly featureless nature of the global x - ray spectrum of the remnant , they used a model for the shocked ejecta in which most of the emission is produced by a nearly pure layer of fully ionized carbon . the elements more massive than carbon are assumed to be significantly underionized in the ejecta , to suppress the x - ray line emission that would otherwise be produced by these elements . however , these assumptions are inconsistent with the results of subsequent analyses . the results of an analysis of the ultraviolet absorption - line features along the line of sight to the s - m star suggest that most of the ejected silicon has already been shocked @xcite , and therefore may not be in a low ionization state . this result is supported by x - ray observations that reveal line emission from several highly ionized elements : oxygen , neon , magnesium , silicon , and perhaps iron @xcite . since the relative abundance of the shocked silicon producing the x - ray emission is much larger than the relative abundance of silicon in the solar system , most of the shocked silicon is ejecta . furthermore , the results of the analysis of @xcite suggest that the total amount of reverse - shocked carbon required by the model of hamilton et al . may be unreasonably large for the remnant of a type ia supernova ( i.e. , may be well in excess of the chandrasekhar mass ) . for these reasons , the thermal model of hamilton et al . provides a poor description of the x - ray emission of sn 1006 . @xcite present a nonthermal interpretation of the high - energy x - ray emission . they argue that the x - ray emission from the rims is produced by relativistic electrons that are beamed from an unobserved central compact object . this conclusion seems to be based largely on the x - ray morphology of sn 1006 , because no evidence of a compact object or of particle jets in the remnant has been reported . for comparison , the pulsar and the pulsar jet of the vela supernova remnant are observed to produce radio , x - ray , and gamma - ray emission @xcite . @xcite set an upper limit on the flux of a point source of @xmath9 erg @xmath10 s@xmath11 . the corresponding constraint on the blackbody temperature of a 10 km neutron star is @xmath12 k ( at a distance of 1.7 kpc ) . some of the models of cooling neutron stars indicate that the temperature of a neutron star is higher than this upper limit at an age of 990 yr @xcite . furthermore , the sketchy record of the optical light curve of the supernova @xcite , the lack of evidence of an ob association near sn 1006 ( hamilton et al . 1986 ) , the large distance of the remnant from the plane of the galaxy [ @xmath13 pc ] , measurements of the average expansion rate of the remnant @xcite , and measurements of the relative abundances of the elements producing the ultraviolet absorption lines @xcite suggest that sn 1006 was produced by a type ia supernova . type ia supernovae are generally not expected to produce compact objects . for these reasons , it seems unlikely that sn 1006 contains an undetected compact object that is responsible for the high - energy x - ray emission from the rims . @xcite advance an alternative nonthermal explanation of the high - energy x - ray emission . they conclude that the emission from the rims is synchrotron radiation from electrons that have been accelerated to energies of about 100 tev . @xcite argues convincingly that this conclusion is the only plausible one and shows that the nonthermal x - ray emission of the rims can be described by a model in which both the radio and x - ray emission are produced by a common population of shock - accelerated electrons . the strong correlation of the features of a high - energy x - ray image and a radio image of the northeastern rim of sn 1006 @xcite supports this result . furthermore , the detection of inverse - compton tev gamma - ray emission from the northeastern rim of the remnant confirms the presence of very high energy electrons in sn 1006 @xcite . since the tev photons are produced by electrons that have energies similar to the energies needed to produce x - ray synchrotron radiation and since the radio , x - ray , and gamma - ray spectral data can be fitted with simple models of the spectrum of shock - accelerated electrons @xcite , the conclusion that the nonthermal x - ray emission is produced by synchrotron radiation seems unavoidable . to accurately model the shape of the entire x - ray spectrum of sn 1006 , from 0.12 to 17 kev , we analyzed data obtained using the proportional counter array ( pca ) of the _ rossi x - ray timing explorer _ ( _ rxte _ ) satellite , the solid - state imaging spectrometers ( siss ) of the _ advanced satellite for cosmology and astrophysics _ ( _ asca _ ) , and the position sensitive proportional counter ( pspcb , hereafter just pspc ) of the _ rntgen _ satellite ( _ rosat _ ) . between 1996 february 18 and 1996 february 20 , sn 1006 was observed for 21 ks at a location on the northeastern rim ( @xmath14 , @xmath15 ) and for 20 ks at a location on the southwestern rim ( @xmath16 , @xmath17 ) using the pca . the pca is a spectrophotometer that is comprised of an array of five coaligned proportional counter units that are mechanically collimated to have a field of view of @xmath18 fwhm @xcite . the array is sensitive to photons that have energies between about 2 and 60 kev and has an energy resolution @xmath19 at 6 kev . the maximum on - axis collecting area is about 7000 @xmath20 . the pca data were screened to remove the time intervals during which ( 1 ) one or more of the five proportional counter units is off , ( 2 ) sn 1006 is less than @xmath21 above the limb of the earth , ( 3 ) the background model is not well defined , and ( 4 ) the pointing direction of the detectors is more than @xmath22 from the nominal pointing direction in either right ascension or declination . the set of x - ray events that satisfies these four selection criteria includes 7 ks of the data of the northeastern pointing and 11 ks of the data of the southwestern pointing . spectra were constructed for each of the two pca pointings . the two spectra are quite similar because the field of view of the pca is large enough to include the entire remnant at both pointing locations . therefore , the pca data are insensitive to modest changes in the shape of the spectrum of the remnant as a function of position in the remnant . the pca background spectrum for sn 1006 was estimated using version 1.5 of the ftool _ pcabackest_. this version of _ pcabackest _ includes an estimate of the charged - particle and diffuse cosmic x - ray backgrounds based on the `` vle '' count rate during observations of `` source - free '' regions of the sky and an estimate of the background associated with the decay of radioactive material that is activated when the spacecraft passes through the south atlantic anomaly . the tool has been found to produce slightly inaccurate estimates of the background associated with astrophysical sources . for example , when the background spectrum is subtracted from the spectral data of sn 1006 , the resulting `` source '' spectrum of the remnant has a negative count rate at high energies . to compensate for this problem , the spectrum of the background associated with the south atlantic anomaly is multiplied by a factor of 0.2 . this factor is determined by requiring the average difference between the source spectrum and the source model to be zero at energies between 27 and 71 kev . in this energy band , the total number of events associated with sn 1006 is expected to be negligibly small compared to the number of events associated with the background . the adjustment reduces the count rate of the 210 kev activation background from 1.0 to 0.2 counts s@xmath11 for the top layer of anodes in the pca . after applying this adjustment , the average 210 kev background count - rate for the top layer is estimated to be @xmath23 counts s@xmath11 for both pointings . on 1993 august 19 - 20 , 1993 september 13 - 15 , and 1996 february 20 - 21 , the instruments on _ asca _ @xcite were used to observe sn 1006 for 43 ks at a location in the center of the remnant ( @xmath24 , @xmath25 ) , for 38 ks at a location on the northeastern rim ( @xmath26 , @xmath27 ) , and for 57 ks at a location on the southwestern rim ( @xmath28 , @xmath29 ) , respectively . the sis detectors are ccd devices @xcite sensitive to photons that have energies between about 0.5 and 10 kev . the devices have an energy resolution @xmath300.08 ( fwhm ) at 1.5 kev and a field of view of @xmath31 . the spatial resolution of the telescope @xmath32 ( 50% encircled - photon radius ) for the sis data used here . the maximum on - axis effective area of the system is about 220 @xmath20 . the sis data were screened using the standard `` rev2 '' criteria that exclude time intervals ( 1 ) associated with passages through the south atlantic anomaly and regions of relatively high particles fluxes , ( 2 ) during which the bright limb of the earth is in the field of view , and ( 3 ) during which the source is occulted by the earth . the set of x - ray events that satisfies these selection criteria includes 23 , 20 , and 43 ks of data for the center , northeastern rim , and southwestern rim , respectively . the sis spectra for these three spatially separate regions , which correspond to the central , northeastern , and southwestern ovals of figure [ fig1 ] , have event rates of 0.21 , 0.89 , and 0.66 counts s@xmath11 , respectively . on 1993 january 2829 , sn 1006 was observed for 6 ks at a location near the center of the remnant ( @xmath33 , @xmath34 ) using the pspc . the pspc is a multiple - wire proportional counter sensitive to photons that have energies between about 0.1 and 2 kev . the energy resolution @xmath35 at 0.93 kev . the telescope has a field of view of @xmath36 . the spatial resolution @xmath37 ( 50% encircled - photon radius ) for the pspc data used here . the maximum on - axis effective area of the system is about 260 @xmath20 @xcite . spectra were constructed using the pspc data for three spatially separate regions : the bright northeastern rim , the bright southwestern rim , and the center of the remnant ( the three rectangular regions of fig . [ fig1 ] ) . the total pspc event rates for these three spectra are 5.6 , 6.1 , and 11.9 counts s@xmath11 , respectively . the pspc background for sn 1006 is negligibly small . @xcite report that the event rate associated with unrejected cosmic - ray particles is @xmath38 counts s@xmath11 kev@xmath11 arcmin@xmath39 in the pulse - height range @xmath40 . in this case , the cosmic - ray event rate over the entire remnant is only about @xmath41 counts s@xmath11 kev@xmath11 . since this rate is considerably smaller than the pspc count - rate for sn 1006 at all energies from 0.12 to 2 kev , the backround is neglected . a joint fit to the _ rxte _ pca , _ asca _ sis , and _ rosat _ pspc data was performed using version 10.0 of the spectral - fitting software package xspec . since @xcite and others report evidence of both thermal and nonthermal emission , the spectral fits were performed using several combinations of thermal and nonthermal emission components . the thermal components include a bremsstrahlung model ( brem ) , a model of a thin thermal plasma in collisional ionization equilibrium ( rs ; * ? ? ? * ) , and a model of a thin thermal plasma that is not in ionization equilibrium ( nei ; * ? ? ? the nonthermal components include power law ( pl ) and broken power law ( bpl ) models . for each fit , only the absolute normalization of each spectral component was allowed to vary from one region to another . while the spectra most likely vary as a function of position in the remnant , we find no evidence of a significant variation in the shape of the nonthermal spectra of the different regions . furthermore , @xcite and @xcite report no evidence of spatial variation in the emission of the interior of the remnant . table [ tab1 ] summarizes the results of some of the fits . note that the values of @xmath42 @xmath43 for all of the fits . this result suggests that the value of @xmath44 may be sensitive to systematic errors in the responses of the detectors as well as differences between the spectral models and the spectral data . therefore , the @xmath44 test statistic does not provide an unambiguous measure of the goodness of fit of the models . however , if the contributions of the systematic errors to @xmath44 do not change from one fit to another , the differences in @xmath44 provide a useful measure of the relative goodness of fit of the different models . in this case , @xmath45 is distributed as a @xmath44 variable with the number of degrees of freedom equal to the number of free parameters in the fit @xcite . for example , 26 free parameters are used for the best - fit model of table [ tab1 ] . therefore , the values of @xmath46 , 35.6 , 39.3 , and 50.6 are expected to correspond to the 1 @xmath47 , 90% , 2 @xmath47 , and 3 @xmath47 confidence levels , respectively . these values of @xmath48 are used in this paper to determine the confidence intervals of the fit and inferred parameters . since the differences between the value of @xmath44 for the bpl@xmath49nei model and the values of @xmath44 for the other sets of models @xmath50 , the bpl@xmath49nei model seems to provide a significantly better fit to the entire broadband spectral data than the other sets of models listed in table [ tab1 ] . table [ tab2 ] lists the fit parameters of the bpl@xmath49nei model . the spectral data and the best - fit model are shown in figures [ fig2][fig4 ] . in general , the model fits the 0.1217 kev data quite well . the thermal component of the best - fit model is a nonequilibrium ionization model of a thin thermal plasma that has different electron and ion temperatures ( hamilton et al . 1983 ) . the shape of the x - ray continuum is specified by the temperature associated with the forward shock , @xmath51 , and the parameter @xmath52 , where @xmath53 is the ambient density of interstellar hydrogen and @xmath54 is the initial kinetic energy of the ejecta . this model is implemented in xspec version 10.0 using a fixed grid in the @xmath55-@xmath52 plane . the temperature is quantized by factors of @xmath56 from @xmath57 to @xmath58 k and @xmath52 is quantized by factors of 10 from @xmath59 to @xmath60 ergs @xmath61 . the best - fit values of these two parameters are @xmath62 k and @xmath63 ergs @xmath61 ( fig . [ fig5 ] ) . figure [ fig5 ] includes curves along which the ionization timescale , @xmath64 , and the characteristic electron temperature , @xmath65 , are constant . the elongated confidence level contours in this figure nearly lie along curves of constant ionization timescale . the best - fit values of @xmath51 and @xmath52 are driven by the thermal emission of sn 1006because the same values for these two parameters ( within the statistical uncertainties ) are obtained if the same set of spectral models is fitted to only the _ asca _ sis spectrum of the center of the remnant . the relative elemental abundances of the elements oxygen , neon , magnesium , and silicon are included as free parameters in the fit because the _ asca _ sis spectrum of the central region of sn 1006 exhibits atomic emission - line features associated with these elements @xcite . the abundances of sulphur and calcium are fixed to be the same as the abundance of silicon because this situation is more or less consistent with the type ia supernova model of @xcite . since the value of @xmath44 is not significantly reduced if the abundances of helium , carbon , nitrogen , iron , or nickel are included as free parameters in the fit , these five abundances are set to be the same as the relative abundances of these elements in the solar system @xcite . the fitted abundances of oxygen and magnesium are consistent with the solar abundances of these elements . the abundance of neon in sn 1006 is less than 80% of the relative solar abundance of neon . the fitted relative abundance of silicon ( and sulphur and calcium ) is 1018 times larger than the solar abundance of silicon . the uncertainties of the abundances are somewhat larger than the statistical uncertainties specified in table [ tab2 ] because ( hamilton et al . 1983 ) note that the intensities used in the nei model for the strong emission lines are generally uncertain by at least a factor of 2 . a broken power law model is used here to approximate a gradually steepening nonthermal x - ray spectrum . such a spectrum is expected if the nonthermal emission is produced by synchrotron radiation @xcite . as shown in figure [ fig6 ] , the nonthermal spectrum of sn 1006 steepens significantly with increasing energy . the difference between the high - energy and low - energy photon indices of the broken power law @xmath66 at the 3 @xmath47 confidence level . the value of the break energy of the broken power law ( @xmath67 kev ) is not physically meaningful . it does not correspond to a feature in the nonthermal x - ray spectrum of sn 1006 . this value is determined by the transition from the relatively high count rate portion of the _ rosat _ pspc spectrum to the relatively high count rate portion of the _ rxte _ pca spectrum . only one thermal emission model is used in the present analysis to describe the properties of the x - ray emitting material in sn 1006 . the use of one thermal component is undoubtedly an oversimplification , because both the forward - shocked interstellar material and the reverse - shocked ejecta produce x rays . the large fitted abundance of silicon suggests that most of the shocked silicon is ejecta . yet the inferred amount of swept - up mass implies that most of the shocked mass is interstellar . for these reasons , the use of a single thermal model to describe the global thermal emission properties of the remnant may provide a poor description of any given volume element of the x - ray emitting material . nevertheless , the use of a single thermal model is sufficient to accurately characterize the properties of the nonthermal x - ray emission ( our primary goal ) , because the x - ray emission of the remnant is dominated by nonthermal emission at energies above 1 kev , because the results for the broken power law component are insensitive to the details of the single thermal component used in the fits ( compare the results obtained for the broken power laws of the `` bpl@xmath49nei '' and `` bpl@xmath49rs '' models listed in table 1 ) , and because @xcite and @xcite report no evidence of spatial variation of the thermal x - ray spectrum of the interior of sn 1006 . models that include two thermal components and one nonthermal component were fitted to the data . however , the results for the fit parameters of these three - component models showed that the parameters of one of the two thermal components are essentially unconstrained . @xcite favor a model for the x - ray emission of the interior of sn 1006 that includes two thermal components instead of one thermal component and one nonthermal component . they find that the temperatures of the hotter thermal components @xmath68 kev for both the northern and southern regions of the interior . these temperatures are quite high for the observed galactic supernova remnants . of the young , shell - type remnants , only cassiopeia a is reported to have a thermal plasma with a temperature this high . @xcite show that it may be possible for a reverse - shocked silicon - dominated plasma to have electron temperatures this high , but such a metal - rich plasma would cool quickly . therefore , the cooling time would have to be comparable to the age of the remnant or longer . otherwise , the reverse - shocked electrons may be too cool to produce much x - ray emission . the model of @xcite can not explain nonthermal x - ray emission from the interior of the remnant . since @xcite show that about 25% of the radio emission at 843 mhz is produced in the interior of sn 1006 , a similar fraction of nonthermal x - ray synchrotron emission is expected from this region . our fits to the x - ray data with a model that includes one thermal and one nonthermal component suggest that @xmath69% of the nonthermal x - ray emission is produced in the region . since this fraction is comparable to the fraction of the radio emission from the interior of sn 1006 , we favor a model for the interior ( and the rims ) that includes one thermal component and one nonthermal component instead of two thermal components . it should be emphasized that the principal results of the present analysis concerning the properties of the cosmic rays in sn 1006 are insensitive to this choice , since only a small fraction of the high - energy x - ray emission is produced in the interior of the remnant . the parameters of the best - fit model are listed in table [ tab2 ] . table [ tab3 ] includes a list of the inferred values of the characteristic energy of the thermal electrons , @xmath65 ( see fig . [ fig5 ] ) , the velocity of the forward shock , @xmath70 , the radius of the forward shock , @xmath71 , the distance of the remnant , @xmath72 , the ionization timescale , @xmath64 ( see fig . [ fig5 ] ) , the age of the remnant , @xmath73 , the density of interstellar hydrogen , @xmath53 , and the swept - up mass , @xmath74 ( hamilton et al . 1983 ) . these inferences are based on the assumptions that sn 1006 is in a sedov phase , that the initial kinetic energy of the ejecta is @xmath75 ergs , that the x - ray emission is dominated by forward - shocked interstellar material , and that the distribution of the ambient material around sn 1006 is homogeneous . the best - fit value for the absorption column density of interstellar hydrogen @xmath76 atoms @xmath10 ( at the 90% confidence level ) . this value is consistent with the result of schaefer ( 1996 ) [ @xmath77 atoms @xmath10 ] , who analyzed a compilation of results from several sources . the relative elemental abundances of oxygen , neon , magnesium , and silicon are consistent with the results of @xcite , but vink et al . ( 2000 ) obtain a somewhat lower abundance for oxygen . the inferred energy of the thermal electrons ( @xmath79 kev at the 90% confidence level ) is similar to the results obtained by vink et al . ( 2000 ; @xmath80 and @xmath81 kev ) and dyer et al . ( 2001a ; @xmath82 ) , but it is somewhat lower than the result of koyama et al . ( 1995 ; @xmath83 kev ) . the results of an analysis of the ultraviolet emission - line data of the northwestern rim favor a low electron temperature ( @xmath84 kev ) , but the results can not exclude temperatures as high as @xmath85 kev @xcite . the inferred shock velocity ( @xmath86 km s@xmath11 at the 90% confidence level ) is consistent with the results of kirshner , winkler , & chevalier ( 1987 ; @xmath873870 km s@xmath11 ) , raymond , blair , & long ( 1995 ; @xmath88 km s@xmath11 ) , and laming et al . ( 1996 ; @xmath89 km s@xmath11 ) for the northwestern rim of sn 1006 , but @xcite argue that the average forward shock velocity is @xmath90 ks s@xmath11 . the inferred distance ( @xmath91 kpc at the 90% confidence level ) is consistent with the result of schaefer ( 1996 ; @xmath92 kpc ) and with a distance estimate that is based on the velocity and proper motion of the forward shock in the northwest . as described above , @xcite estimate that the velocity of the forward shock in the northwest @xmath89 km s@xmath11 . since neutral atoms have very short lifetimes after having passed through the shock into the hot postshock gas , the h@xmath93 filaments trace the location of the forward shock @xcite . the results of analyses of the mean proper motion of the filaments along the northwestern rim of sn 1006 suggest that @xmath94 yr@xmath11 ( long et al . 1988 ) or @xmath95 yr@xmath11 @xcite . the weighted average of these two results is @xmath96 yr@xmath11 . therefore the implied distance of sn 1006 @xmath97 kpc . these distance estimates are consistent with the upper and lower limits on the distance ( hamilton et al . 1986 , 1997 ; * ? ? ? * ; * ? ? ? * ) . collectively , the results suggest that @xmath98 kpc . the ionization timescale ( @xmath99650 @xmath2 yr ) , age ( @xmath1006300 yr ) , and ambient density of interstellar hydrogen ( @xmath1010.37 @xmath2 ) , are consistent with the ionization timescale of koyama et al . ( 1995 ; @xmath102 @xmath2 yr ) , the known age of the remnant , and the ambient density estimates of toor ( 1980 ; @xmath103 @xmath2 ) and vink et al . ( 2000 ; @xmath104 @xmath2 ) . @xcite find a somewhat larger ambient density [ @xmath105 @xmath2 ] , and @xcite report a somewhat lower density ( @xmath1070.1 @xmath2 ) . although the remnant is interacting with a relatively dense environment in the northwest ( @xmath108 @xmath2 ; * ? ? ? * ) , the average ambient denisty at a galactic height of 430 pc is about 0.06 @xmath2 @xcite . therefore , the average ambient density around sn 1006 may be about 0.1 @xmath2 . the inferred amount of shocked material @xmath10967 @xmath110 . however , this range of values is an overestimate because the relative abundance of helium used in the best - fit model ( table [ tab2 ] ) is the same as the relative abundance of heilium in the solar system . since @xcite and @xcite report much smaller densities of helium in the neighborhood of sn 1006 , a more appropriate abundance may be about 0.2 times the solar abundance . in this case , the best - fit amount of shocked material @xmath111 @xmath110 , and the 90% confidence level range is 647 @xmath110 . this range is consistent with the mass estimates of pye et al . ( 1981 ; @xmath11215 @xmath110 ) and vink et al . ( 2000 ; @xmath113 @xmath110 ) . however , the results of @xcite suggest a somewhat smaller mass ( @xmath1145 ) and @xcite obtain a much smaller mass [ @xmath115 @xmath110 ] . with the exeption of the estimate of @xcite , the mass estimates are substantially larger than the chandrasekhar mass . therefore , most of the hot shocked material is probably forward - shocked interstellar matter . unlike the parameters inferred from the thermal component of the bpl@xmath49nei model , most of the parameters inferred from the thermal component of the bpl@xmath49rs model are inconsistent with the results of other observations . for example , the inferred velocity of the forward shock @xmath117420 km s@xmath11 and the inferred ambient density @xmath1180.014 atoms @xmath2 . these estimates are about an order of magnitude smaller than the results of previously published analyses . the values inferred for most of the other parameters listed in table [ tab3 ] are also unrealistic for the raymond - smith component of the bpl@xmath49rs model . therefore , the raymond - smith model provides a poor description of the thermal emission of sn 1006 . the differences in the physical properties of the raymond - smith and nonequilibrium ionization models in xspec are that the raymond - smith model is based on the assumptions that the plasma is in a state of collisional ionization equilibrium and that the electron and ion temperatures have equilibrated . neither of these two conditions is assumed for the nonequilibrium ionization model . in fact , at least one of the assumptions is not appropriate for sn 1006 . @xcite show that the electron temperature is significantly smaller than the temperature of the ions in the northwest ( @xmath119 ) . this difference in the temperatures could explain why the parameters inferred from the raymond - smith model are inconsistent with other results . table [ tab4 ] lists estimates of the kinetic energy of the matter in the remnant , @xmath120 , the thermal energy of the shock - heated protons , @xmath121 , and electrons , @xmath122 , the energy in the magnetic field , @xmath123 , and the energy in cosmic rays , @xmath124 . some of these estimates are based on the assumptions that the remnant is in a sedov phase ( hamilton et al . 1983 ) and that the initial kinetic energy of the ejecta @xmath125 ergs . the scaling of the estimates with @xmath54 is indicated in the table . however , the results for sn 1006 may be insensitive to the assumption that the remnant is in a sedov phase . for example , @xcite suggest that sn 1006 has not yet reached the sedov phase . the results of their analysis suggest that @xmath126 @xmath127 , @xmath128 @xmath2 , @xmath129 pc , and @xmath130 km s@xmath11 . if these values ( instead of the values in table [ tab3 ] ) are used to compute the parameters in table [ tab4 ] , @xmath131 ergs and @xmath132 ergs . therefore , the same distribution of energy is obtained using the results of @xcite because the smaller estimates of the amount of shocked material and the ambient density are offset by the larger estimates of the shock radius and velocity . the high - energy x - ray emission from the northeastern and southwestern rims of sn 1006 is almost certainly nonthermal . an attempt to describe this emission using only thermal bremsstrahlung emission ( hamilton et al . 1983 ) was unsuccessful @xcite . as described in section [ obs ] , the suggestion that the emission is associated with electrons beamed from an unseen central object is untenable . @xcite argues convincingly that the only plausible explanation of the emission is synchrotron radiation from electrons accelerated to energies of about 10100 tev . this conclusion is supported by the results of the present analysis , which show that the nonthermal spectrum steepens with increasing energy ( fig . [ fig6 ] ) . figure [ fig7 ] displays a compilation of the radio flux - density results of @xcite , @xcite , @xcite , @xcite , and @xcite . the radio data can be fitted with a power law spectrum : @xmath133 jy . the results of analyses of data obtained using the _ iras _ @xcite , _ rosat _ pspc , _ rxte _ pca , egret @xcite , cangaroo @xcite , and janzos @xcite instruments are also shown . the figure includes estimates of the photon spectra associated with synchrotron radiation , inverse compton scattering of the cosmic microwave background radiation , nonthermal bremsstrahlung , and the decay of neutral pions @xcite . the synchrotron spectrum is extrapolated from the radio data to x - ray energies assuming that the electron spectrum is as described in section [ crs ] . the nonthermal bremsstrahlung and @xmath134 spectra were computed assuming the density of protons in the remnant is 0.7 @xmath2 . the nonthermal x - ray spectrum is consistent with a model of synchrotron radiation , but it is not consistent with models of nonthermal bremsstrahlung or inverse compton scattering . the tev gamma - ray data can be described by inverse compton scattering . as illustrated in figure [ fig7 ] , the detection of x - ray synchrotron emission from sn 1006 requires the emission of synchrotron radiation at all energies between radio frequencies and x rays . @xcite report the detection of faint diffuse emission ( @xmath135@xmath136 erg @xmath10 s@xmath11 arcsec@xmath39 ) in the h@xmath93 band around the entire circumference of the remnant . @xcite notes that the surface brightness of the synchrotron emission is about 2832 mag arcsec@xmath39 in the visible band . this range corresponds to a flux of 0.1@xmath137 erg @xmath10 s@xmath11 arcsec@xmath39 in the h@xmath93 band . therefore , some of the faint diffuse h@xmath93 flux may be produced by synchrotron radiation from electrons that have energies of about 1 tev . the measurement of the optical ( or infrared ) synchrotron flux would be very important because it would corroborate the claim that the high - energy x - ray emission is synchrotron emission and could help determine the shape of the cosmic - ray electron spectrum of sn 1006 . as shown in figure [ fig7 ] , the synchrotron and inverse compton spectra of sn 1006 have similar shapes . gamma - ray emission is detected near the peak of the inverse compton spectrum . the electrons that produce this emission are the same electrons that produce the emission near the peak of the synchrotron spectrum . therefore , the gamma - ray and nonthermal x - ray emission are produced by electrons that have similar energies . for this reason , the nonthermal x - ray and gamma - ray images are expected to be similar . @xcite report the detection of gamma rays from only the northeastern rim of sn 1006 . the 95% confidence level upper limit on the gamma - ray flux from the southwestern rim ( @xmath138 @xmath10 s@xmath11 at energies above 1.7 tev ) is no larger than about 40% of the flux of the northeastern rim [ @xmath139 @xmath10 s@xmath11 at energies above 1.7 tev ] . however , the nonthermal x - ray fluxes of the northeastern and southwestern rims are comparable . this discrepancy can be explained if the electron spectra , magnetic field strengths , or synchrotron and inverse compton emission - angle distributions of the two rims are different . of these possibilities , the x - ray data can be used to search for differences in the shapes of the electron spectra by searching for differences in the shapes of the synchrotron spectra of the two rims . figure [ fig8 ] shows a comparison of the nonthermal x - ray spectra of the two rims . for the purposes of this comparison , only the _ rosat _ pspc data were used . the pspc spectra of each rim were fitted with a model that includes a power law component ( @xmath140 ) and the thermal component described in table [ tab2 ] . the ratio of the nonthermal x - ray fluxes of the southwestern and northeastern rims @xmath141 . the 1 , 2 , and 3 @xmath47 confidence level contours for the parameter space defined by @xmath142 and @xmath143 are shown in figure [ fig8 ] . since the point defined by @xmath144 and @xmath145 is well inside the 1 @xmath47 confidence contour , there is no reason to believe that the shapes of the nonthermal x - ray spectra of the two rims are different in the energy range to which the _ rosat _ pspc is sensitive . the results of the present analysis of the _ rxte _ pca data of the two rims also suggest that the shapes of the nonthermal x - ray spectra of the two rims are consistent with each other in the energy range to which the pca is sensitive . figure [ fig8 ] includes curves along which the ratio @xmath146 is a constant . here , an energy @xmath147 kev is used to compute the value of @xmath146 , because we expect that the synchrotron emission at 0.1 kev and the tev emission are produced by electrons that have the same energies . at the 3 @xmath47 confidence level , the nonthermal flux from the southwestern rim at 0.1 kev is greater than 50% of the nonthermal flux from the northeastern rim at this energy . at face value , this lower limit seems to be inconsistent with the upper limit of 40% on the ratio of the gamma - ray fluxes . however , a significant difference between the electron spectra of the two rims produces a significant difference in the corresponding inverse compton spectra , but only a modest difference in the synchrotron spectra . therefore , the modest constraint on the ratio of the inverse compton fluxes corresponds to a poor contraint on the on the ratio of the synchrotron fluxes of the two rims , and the results of the present analysis can not exclude the possibility that the differences between the x - ray and gamma - ray images are due to differences between the electron spectra of the two rims . perhaps this issue will be settled after the tev and x - ray spectra of the two rims are better measured . the available radio and x - ray data for sn 1006 can be used to infer the parameters of the spectrum of the cosmic - ray electrons producing the synchrotron emission . for simplicity , the relativistic electron spectrum is assumed to have the form @xmath148 . this form is the high - energy limit of bell s ( 1978 ) expression : @xmath149 where @xmath150 @xmath151 is the kinetic energy of the particles and the exponential cutoff is added to the formula of bell . the electron spectral index @xmath152 can be determined from the spectral index of the radio data : @xmath153 . however , estimates of the normalization factor @xmath154 and the exponential cutoff energy @xmath155 of the electrons can not be uniquely determined from the synchrotron data alone because the normalization ( @xmath156 ) and the roll - off energy ( @xmath157 ) of the synchrotron spectrum depend on the strength of the magnetic field . ( note that @xmath158 kev in fig . [ fig7 ] . ) since the normalization of the inverse compton spectrum produced by the relativistic electrons depends on @xmath154 ( and the known spectrum of the cosmic microwave background radiation ) , but not @xmath159 , the tev gamma - ray data can be used to determine @xmath154 . then the flux and the roll - off energy of the synchrotron spectrum can be used to determine @xmath159 and @xmath155 , respectively . for example , if the cosmic - ray electrons and the magnetic field of sn 1006 have effective volume - filling factors @xmath160 , if the volume of the remnant @xmath161 @xmath162 , and if the inverse compton emission is dominated by scattering of the cosmic microwave background radiation , @xmath163 @xmath2 gev@xmath164 , @xmath165 @xmath5 g , and @xmath166 tev . similar estimates of the magnetic field strength are reported by tanimori et al . ( 1998 ; @xmath167 @xmath5 g ) , aharonian & atoyan ( 1999 ; @xmath168 @xmath5 g ) , and dyer et al . ( 2001a , @xmath168 @xmath5 g ) . however , these estimates of the magnetic field strength seem rather small . if the field strength at a galactic height of 430 pc is similar to the strength in the plane @xcite , a field of 10 @xmath5 g can be little more than a compressed ambient field . yet , the large and abrupt change in the radio surface brightness in the regions of the bright rims of sn 1006 @xcite , the presence of a radially oriented component of the field @xcite , and estimates of the total energy in cosmic - ray particles ( described below ) suggest that the field has been significantly amplified . perhaps one or more of the assumptions used to determine the strength of the magnetic field is wrong . in particular , the effective volume - filling factor of the magnetic field may be smaller than the filling factor of the cosmic rays @xcite . in this case , the magnetic field strength is underestimated . for example , if @xmath169 and @xmath170 , @xmath171 @xmath2 gev@xmath164 , @xmath172 @xmath5 g , and @xmath173 tev . this value for the magnetic field lies comfortably between the compressed value ( @xmath174 g ) and the value obtained using the minimum - energy condition ( @xmath175 g ) . for the remainder of the discussion , it is assumed that @xmath172 @xmath5 g and that the cosmic - ray electron spectrum is specified by equation ( [ equ1 ] ) , where @xmath154 , @xmath152 , and @xmath155 are listed in table [ tab5 ] . the inferred spectrum of cosmic - ray electrons is shown in figure [ fig9 ] . although there is no evidence that cosmic - ray nuclei are accelerated in sn 1006 ( cf . * ) , @xcite suggest that relativistic electrons and nuclei are accelerated in a similar manner . therefore , we estimate the parameters of the cosmic - ray nuclei spectra . for simplicity , only hydrogen and helium nuclei are considered . the nonthermal spectra of these particles are assumed to have the same functional form as equation ( [ equ1 ] ) . the spectral indices of the protons and alpha particles are assumed to be the same as the spectral index of the electrons ( @xmath176 ) . the exponential cutoff energies of the different cosmic - ray particles are assumed to be related by the magnetic rigidity of the particles ( i.e. , @xmath177 tev ) . this relationship is appropriate if the maximum energy of the electrons is limited by the escape of the particles from the remnant @xcite , but it is not appropriate if @xmath155 is limited by radiative losses . the values of @xmath178 and @xmath179 are determined using the value of @xmath154 and the technique described below . the ratio of the number density of protons to helium nuclei @xmath180 in the neighborhood of sn 1006 @xcite . if the shocked protons and helium atoms are fully ionized , then @xmath181 . ( the electron contribution from other nuclei is expected to be small . ) therefore , @xmath182 . the fractional numbers of nonthermal protons and helium nuclei are assumed to be the same as the fractional number of nonthermal electrons ( i.e. , @xmath183 ) . using this relationship and the equations @xmath184 and @xmath185 yields @xmath186 where @xmath187 kev , @xmath188 tev , @xmath189 kev , and @xmath190 tev . similarly , @xmath191 where @xmath192 kev and @xmath193 tev . the parameters of the cosmic - ray proton and helium spectra are listed in table [ tab5 ] , and the spectra are shown in figure [ fig9 ] . these parameters and equation ( [ equ1 ] ) yield a cosmic - ray injection efficiency @xmath194 if the magnetic field @xmath195 g . this efficiency is assumed to be the same for all particle species . our assumptions about the shapes of the cosmic - ray spectra and about the relative numbers of nonthermal protons and electrons yield a number density of protons that is 160 times larger than the number density of electrons at 1 gev ( i.e. , @xmath196 at 1 gev ; fig . [ fig9 ] ) . within the rather large uncertainties of our estimates , this ratio is consistent with the ratio observed at earth @xcite . the energies in cosmic - ray electrons , protons , and helium nuclei can be obtained from equation ( [ equ1 ] ) and the particle - spectra parameters listed in table [ tab5 ] and above : @xmath197 the sum of these three energies is @xmath7 ergs . as expected , the cosmic - ray energy is dominated by the energy in cosmic - ray protons . the estimates of the cosmic - ray and magnetic field energies are listed in table [ tab4 ] for the case in which @xmath198 g . @xcite estimate the energy in only cosmic - ray electrons and obtain an energy that is about an order of magnitude larger than @xmath199 ergs because they used a magnetic field of about @xmath200 g instead of the field strength of @xmath201 g used here . @xcite and @xcite modeled the evolution of a supernova remnant using conditions that are quite similar to the conditions appropriate for sn 1006 . these authors assume that 1 @xmath127 of material is ejected with an initial kinetic energy of @xmath202@xmath203 ergs into an ambient medium that has a density @xmath204 @xmath2 . the predicted results for the case in which the injection efficiency @xmath205 ( fig . 3 of @xcite and fig . 4 of @xcite ) are consistent with our estimates of the distribution of energy between thermal particles , cosmic rays , and kinetic energy ( table [ tab4 ] ) . note that the cosmic - ray injection efficiency parameter @xmath206 of @xcite and @xcite , which should not be confused with our use of @xmath206 for the exponential cutoff energy of the cosmic - ray spectra , is computed somewhat differently from the injection efficiency parameter @xmath207 used here . if @xmath195 g , @xmath194 and the corresponding value of @xmath208 . therefore , our results and those of @xcite and @xcite are consistent if the magnetic field is about 40 @xmath5 g . if the magnetic field strength is as small as 10 @xmath5 g , @xmath209 , @xmath210 , and the total energy in cosmic rays @xmath211 ergs ( i.e. , the energy is expected to be more or less equally distributed between kinetic energy , thermal energy , and cosmic rays ) . for comparison , figures 2 and 3 of @xcite suggest that almost all of the energy would be contained in the cosmic rays if @xmath212 . therefore , the results of @xcite are more compatible with a magnetic field of @xmath213 g than a field of @xmath200 g . the estimates of the parameters of the cosmic - ray spectra of sn 1006 have important implications for galactic cosmic - ray acceleration . it is generally accepted that galactic cosmic rays are accelerated predominantly in the shocks of supernova remnants . if this hypothesis is true , an average supernova remnant should transfer a sufficient amount of energy to cosmic rays , produce a cosmic - ray proton spectrum that has an appropriate spectral index , and accelerate protons to energies high enough to explain the properties of the cosmic rays observed at earth . since the average flux of cosmic rays in the solar system has been constant within a factor of 2 over the last @xmath214 yr @xcite , the mean rate at which cosmic rays are energized should be approximately equal to the rate at which energy is lost as cosmic rays diffuse from the galaxy . this rate ( @xmath215@xmath216 ergs s@xmath11 ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) and the assumption that one supernova remnant occurs every 30 yr suggest that an average supernova remnant produces 0.3@xmath217 ergs of cosmic rays over the lifetime of the remnant ( i.e. , 3%10% of the initial kinetic energy of the ejecta is transferred to cosmic rays ) . the estimate of the cosmic - ray energy in sn 1006 at the present is already consistent with this range if @xmath218 g . therefore , the total energy of the cosmic - ray protons in sn 1006 may be consistent with the energy inferred for the sources of galactic cosmic rays . the average spectral index of the relativistic cosmic - ray proton spectra produced by the accelerators of galactic cosmic rays ( @xmath219 ) is inferred from the spectral index of the cosmic - ray proton spectrum observed at earth ( @xmath220 ; * ? ? ? * ) and the spectral steepening due to the energy - dependent escape of cosmic rays from the galaxy ( @xmath221 ; * ? ? ? the radio synchrotron flux of sn 1006 is produced by electrons that have energies of about 1 gev . at these energies the spectral index of the cosmic - ray electrons @xmath222 . since relativistic protons and electrons are thought to be accelerated in a similar manner , the relativistic proton spectrum of sn 1006 may have a spectral index that is consistent with the index inferred for the accelerators of galactic cosmic rays . the spectral index of the all - particle cosmic - ray spectrum observed at earth is more or less the same up to an energy of about 3000 tev . above this energy , the spectrum gradually steepens to have an index of about 3 . since the observed cosmic rays that have energies below this `` knee '' feature are thought to be galactic , the galactic cosmic - ray accelerators are expected to be capable of accelerating particles to energies as high as 3000 tev . if the mechanism responsible for the acceleration of galactic cosmic rays depends on the magnetic rigidity of the particles ( which is true for diffusive shock acceleration in supernova remnants ) , it may be the case that the cosmic - ray particles at 3000 tev are principally iron and that protons are accelerated to energies of only about 100 tev @xcite . the estimated cutoff energy of the electron spectrum of sn 1006 ( @xmath223 tev ) , while uncertain , is lower than 100 tev . since relativistic electrons and protons that have the same energy have the same magnetic rigidity , the maximum energy of the protons is expected to be the same as the maximum energy of the electrons unless the maximum energy of the electrons is regulated by radiative losses . if the magnetic field @xmath224 @xmath5 g , an electron with an energy @xmath225 tev radiates half of its energy in about 400 yr . since this time is less than the age of the remnant , the maximum energy of shock - accelerated electrons may be limited by synchrotron losses . in this case , the value of the energy @xmath155 represents a lower limit on the exponential cutoff energy of the proton spectrum because radiative losses are only important for electrons , not nuclei . therefore , the maximum energy of the protons may be consistent with the expected value of @xmath226 tev . however , if the magnetic field strength @xmath227 g , the time required for a 20 tev electron to radiate half of its energy ( 3000 yr ) is significantly greater than the age of the remnant . if the magnetic field is this small , the maximum energy of the electrons in sn 1006 may be regulated by the free escape of the particles from the remnant @xcite . in this case , the maximum energy of the protons is the same as the maximum energy of the electrons and is well below the expected energy of 100 tev . we present the results of a spectral analysis of _ rxte _ pca , _ asca _ sis , and _ rosat _ pspc data of the supernova remnant sn 1006 . these data were fitted with several sets of thermal and nonthermal x - ray emission models to characterize the global spectral properties of the remnant . the present work represents the first attempt to model both the thermal and nonthermal x - ray emission over the entire x - ray energy band from 0.1217 kev . the best - fit model includes a nonequilibrium ionization component and a broken power law component , which are absorbed through an interstellar column density @xmath228 atoms @xmath10 . the thermal component is described by an electron temperature @xmath0 kev and an ionization timescale @xmath229 @xmath2 s. the large relative elemental abundance of silicon ( 1018 times larger than the relative elemental abundance of silicon in the solar system ) indicates that most of the x - ray emitting silicon is reverse - shocked ejecta . the inferred mass of the shocked material is several times larger than the chandrasekhar mass , which implies that most of the shocked material is interstellar . the nonthermal x - ray emission is well described by a broken power law component with a low - energy photon index @xmath230 and a high - energy photon index @xmath4 . since @xmath231 is significantly larger than @xmath232 , the broken power law represents an approximation to a nonthermal spectrum that is steepening with increasing energy . this result supports previous claims that the nonthermal x - ray emission from sn 1006 is produced by synchrotron radiation from very high energy electrons . using both the radio and x - ray synchrotron results , the spectrum of relativistic electrons in the remnant is inferred to have the form @xmath233 , with @xmath171 @xmath2 gev@xmath164 , @xmath8 , and @xmath223 tev . the values of @xmath154 and @xmath155 are based on the assumption that the mean strength of the magnetic field in the synchrotron - emitting region @xmath198 g . this value was arbitrarily chosen for the present work because it is between the value obtained if the magnetic field is merely a compressed ambient field ( 10 @xmath5 g ) and the value obtained using the minimum - energy condition ( 100 @xmath5 g ) . other published estimates of the field strengths are near 10 @xmath5 g , but this value seems to be incompatible with the large , abrupt change in the radio flux at the bright rims , the detection of a significant radial component of the field , and the models of @xcite . since sn 1006 is expected to accelerate cosmic - ray nuclei as well as electrons , we have also estimated the parameters of the proton and helium spectra . the results suggest that the number density of protons is 160 times larger than the number density of electrons at 1 gev . this ratio is comparable to the ratio observed at earth . integrating over the nonthermal particle spectra yields an estimate of the total energy in cosmic rays @xmath234 ergs . this energy and the differential spectral index of the electrons at 1 gev ( @xmath235 ) are consistent with the hypothesis that galactic cosmic rays are accelerated predominantly in the shocks of supernova remnants . however , the maximum energy of cosmic - ray electrons in sn 1006 is well below 100 tev ( @xmath223 tev ) . if the maximum energy of cosmic - ray protons @xmath236 , the remnant does not accelerate particles to the energy of the `` knee '' at about 3000 tev . yet , if @xmath237 g , the maximum energy of the electrons may be limited by synchrotron losses and the maximum energy of nuclei might be significantly larger than 10 tev . therefore , sn 1006 is clearly a significant source of galactic cosmic rays ( at least cosmic - ray electrons ) , but the remnant may or may not be capable of accelerating particles to energies as high as the energy of the knee . similar conclusions are reported for other young galactic shell - type remnants ( allen et al . 1997 , 1999 ; * ? ? ? * ) . unlike the x - ray and radio synchrotron emission , tev gamma - ray emission is only observed from the northeastern rim of the remnant . one possible explanation for this discrepancy is that the two rims have different electron spectra . the results of the present analysis are consistent with the idea that the electron spectra of the two rims are the same , but the possibility that the spectra are different can not be excluded . therefore , the nature of this puzzle remains a mystery . we thank stephen reynolds for many thoughtful and informative discussions about particle acceleration in supernova remnants and the photon emission processes of the accelerated particles . we are grateful for matthew baring s suggestions about the distribution of energy in a supernova remnant . we thank john houck for helpful discussions about the thermal emission properties of shocked plasmas and for carefully reviewing the manuscript . we appreciate the many useful suggestions of the anonymous referee that have helped improve the paper . the present work involved the use of the data analysis tools and data archives maintained by the heasarc at nasa / goddard space flight center . this work was performed , in part , while g. e. a. held an nrc - nasa / gsfc postdoctoral research associateship , and g. e. a. warmly acknowledges keith jahoda and the _ rxte _ pca team for their support and hospitality during his tenure at nasa / gsgc . the research efforts of g. e. a. are supported in part by the _ chandra _ x - ray center under contract svi-61010 with the smithsonian astrophysical observatory . bpl@xmath49nei & @xmath238 & @xmath239 & @xmath240 & 1109/729 + & & @xmath241 & @xmath242 & + & & @xmath243 & & + bpl@xmath49rs & @xmath244 & @xmath245 & @xmath246 & 1202/729 + & & @xmath247 & & + & & @xmath248 & & + pl@xmath49nei & @xmath249 & @xmath250 & @xmath251 & 1864/731 + & & & @xmath252 & + brem@xmath49nei & @xmath253 & @xmath254 & @xmath255 & 2338/731 + & & & @xmath252 & + @xmath256 ( @xmath257 atoms @xmath10 ) & 5.60 & 5.046.24 + @xmath232 & 2.08 & 1.942.20 + @xmath258 ( kev ) & 1.85 & 1.662.05 + @xmath231 & 3.02 & 2.863.19 + flux(0.12 kev ) ( @xmath259 erg @xmath10 s@xmath11 ) & 1.42 & 1.311.54 + ( k ) & @xmath260 & @xmath261@xmath260 + @xmath52 ( ergs @xmath61 ) & @xmath262 & @xmath263@xmath262 + he & 1 & + c & 1 & + n & 1 & + o & 1.6 & 0.82.8 + ne & 0.3 & 00.8 + mg & 2.0 & 05.9 + si & 16 & 1018 + s & 16 & 1018 + ca & 16 & 1018 + fe & 1 & + ni & 1 & + flux(0.12 kev ) ( @xmath259 erg @xmath10 s@xmath11 ) & 2.53 & 2.252.91 + @xmath65 ( kev ) & 0.58 & 0.310.60 + @xmath70 ( km s@xmath11 ) & 2700 & 10002900 + @xmath71 ( pc ) & 5.9 & 5.616.3 + @xmath72 ( @xmath264 ) ( kpc ) & 1.4 & 1.33.7 + @xmath64 ( @xmath214 @xmath2 s ) & 8.6 & 7.620.6 + @xmath73 ( yr ) & 870 & 7606300 + @xmath53 ( @xmath265/@xmath266 ) ( @xmath2 ) & 0.32 & 0.100.37 + @xmath54 ( @xmath75 ergs ) & 1.00 & + @xmath74 ( @xmath127 ) & 9.7 & 8.367 +	we present the results of a joint spectral analysis of _ rxte _ pca , _ asca _ sis , and _ rosat _ pspc data of the supernova remnant sn 1006 . this work represents the first attempt to model both the thermal and nonthermal x - ray emission over the entire x - ray energy band . the thermal flux is described by a nonequilibrium ionization model with an electron temperature @xmath0 kev , an ionization timescale @xmath1 @xmath2 s , and a relative elemental abundance of silicon that is 1018 times larger than the solar abundance . the nonthermal x - ray spectrum is described by a broken power law model with low- and high - energy photon indices @xmath3 and @xmath4 , respectively . since the nonthermal x - ray spectrum steepens with increasing energy , the results of the present analysis corroborate previous claims that the nonthermal x - ray emission is produced by synchrotron radiation . we argue that the magnetic field strength is significantly larger than previous estimates of about 10 @xmath5 g and arbitrarily use a value of 40 @xmath5 g to estimate the parameters of the cosmic - ray electron , proton , and helium spectra of the remnant . the results for the ratio of the number densities of protons and electrons ( @xmath6 at 1 gev ) , the total energy in cosmic rays ( @xmath7 ergs ) , and the spectral index of the electrons at 1 gev ( @xmath8 ) are consistent with the hypothesis that galactic cosmic rays are accelerated predominantly in the shocks of supernova remnants . yet , the remnant may or may not accelerate nuclei to energies as high as the energy of the `` knee , '' depending on the reason why the maximum energy of the electrons is only 10 tev .
tumors develop by accumulating different mutations within a cell , which affect the cell s reproductive fitness @xcite . as @xcite , we refer to the fitness of a mutated cell as the ratio between a cell s rate to proliferate and the cell s rate of apoptosis compared to wild type cells . the higher the fitness , the more likely it is for the cell to proliferate . for high fitness values , the population of cells growths very fast and stochastic effects play a minor role . in our model , this can be thought of as the formation of a tumor . however , many mutations have no impact on the cell s fitness , e.g. mutations not affecting coding or regulatory sequences . other mutations may lead to a fitness disadvantage , which implies that the cell s risk of apoptosis is higher than its chance of proliferation . however , the same mutations in combination with other mutations within the same cell might lead to a large fitness advantage . we were motivated by genetic studies in burkitt lymphoma , a highly aggressive tumor , where a single genetic alteration has an impact on a wide range of other genes , some of them affect cell growth while others induce apoptosis . more specifically , a chromosomal translocation between the _ myc _ protooncogene on chromosome 8 and one of three immunoglobulin ( _ ig _ ) genes is found in almost every case of burkitt lymphoma @xcite . this leads to deregulated expression of the _ myc _ rna and in consequence , to deregulated myc protein expression . the myc protein acts as a transcription factor and has recently been shown to be a general amplifier of gene expression @xcite , targeting a wide range of different genes . most importantly , _ myc _ expression induces cell proliferation . in burkitt lymphoma , the _ ig - myc _ fusion is evidently the key mutation for tumorigenesis @xcite . however , _ myc _ plays also a key role in inducing apoptosis @xcite . thus , the _ ig - myc _ translocation alone would lead to cell death . therefore , the _ ig - myc _ translocation has to be accompanied by additional mutations , which deregulate the apoptosis pathways , such as mutations affecting e.g. _ tp53 _ or _ arf _ @xcite . these additional mutations have probably only little direct impact on the cell s fitness , since apoptosis is rare . hence , these mutations can not be considered as primary driver mutations in the context of burkitt lymphoma . however , in combination with the _ myc _ mutation these additional mutations decrease the apoptosis rate . consequently , the cells proliferate fast and the population grows accordingly , leading to tumorigenesis . because all cells carry the _ mutation in burkitt lymphoma , but fast growth does not start immediately with that mutation , it seems to confer its large fitness advantage only in a certain genetical context . thus , interactions between different mutations may crucially affect the dynamics of cancer progression . due to the fact that those additional mutations do not confer a direct fitness advantage , they can not be considered as driver mutations . nevertheless , at least some of them are necessary in order for the _ myc _ mutation to become advantageous for the cell . therefore , they can not be regarded as true passenger mutations , either . throughout this manuscript , we therefore call these additional mutations `` secondary driver mutations '' . besides burkitt lymphoma , epistatic effects in cancer initiation seem also to be relevant for other cancers . for example , we can think of the inactivation of a tumor suppressor gene discussed by knudson in the context of retinoblastoma @xcite . this inactivation is neutral for the first hit but highly advantageous for the second hit , and can hence be viewed as an interaction of genes @xcite . another case is found in lung carcinomas , where activation of each of two oncogenes ( _ sox2 _ and _ prkci _ ) alone is insufficient , but in concert they initiate cancer @xcite . in other cases , there is clear evidence for sign epistasis : the _ ras _ family of proto - oncogenes is also discussed to underlie epistatic effects . amplification of _ ras _ leads to senescence in the cell . nevertheless , _ ras _ is a well known oncogenic driver gene . hence , the _ ras _ mutation needs to be accompanied by other mutations @xcite . moreover , the difficulty to distinguish between drivers and passengers @xcite suggests that for a full understanding of cancer initiation it is insufficient to think of these two types of mutations only . so far , most models have focused on the idea that passenger mutations have no effect or only a little effect , whereas each driver mutation increases the fitness of the cell @xcite . other models focus on the neutral accumulation of mutations @xcite . moreover , different mutations are typically treated as independent , which is a strong assumption that will often not be fulfilled . in our model , mutations are interacting in an epistatic way @xcite : the change in fitness induced by the driver mutation depends strongly on the genetic environment , i.e. in our case on the number of secondary driver mutations that are present in that cell . in addition we assume that the secondary driver mutations alone have almost no fitness advantage . such a dependence between mutations can strongly affect the dynamics of cancer initiation . in evolutionary biology , epistatic systems are often analyzed regarding the structure or ruggedness of the landscape and the accessibility of different pathways @xcite . the experimental literature also studies which factors can lead to epistasis @xcite . here , we are interested in the dynamics of such an epistatic model , which we illustrate by stochastic , individual based simulations . in addition , we derive analytical results for the average number of cells with different combinations of mutations and find a good agreement with the average dynamics in individual based computer simulations . furthermore , we discuss the computation of the waiting time until cancer initiation . our results show that the dynamics in such systems of epistatic interactions are distinct from previous models of cancer initiation @xcite , which may have important consequences for the treatment of such cancers . while in previous models there is a steady increase in growth with every new mutation , in our model there is a period of stasis followed by a rapid tumor growth . of course , the biology of burkitt lymphoma is much more complex than modelled herein . to make the model more realistic one would have to distinguish between the different secondary driver mutations , since different genes contribute differently to the cells fitness , especially in a cell where the _ ig - myc _ fusion is present . our model is not aimed to realistically describe such a situation in detail . instead , we focus on the extreme case of the so called _ all - or - nothing _ epistasis @xcite to illustrate its effect on the dynamics of cancer initiation . as there is no theoretical analysis of epistatic effects in cancer initiation so far , a well understood minimalistic model seems to be necessary in order to illustrate the potential impact of epistasis on cancer progression . our minimalistic model clearly shows that epistasis can lead to a qualitatively different dynamics of cancer initiation . we analyze cancer initiation in a homogenous population of initially @xmath0 cells with discrete generations . in every generation , each of the @xmath0 cells can either die or divide . if a cell divides , its two daughter cells can mutate with mutation probabilities @xmath1 for the driver mutation and @xmath2 for secondary driver mutations ( where the @xmath3 indicates that these would be called passenger mutations in closely related models ) . in principle , we could drop the assumption that these two mutation probabilities are independent on the cell of origin , but this would lead to inconvenient notation . we neglect back mutations and multiple mutations within one time step , because their probabilities are typically very small . figure [ fig : model ] summarizes the possible mutational pathways of the model . the variables @xmath4 denote the number of cells with or without the primary driver mutation ( @xmath5 or @xmath6 respectively ) , and @xmath7 secondary driver mutations . denote the number of cells with or without the primary driver mutation ( @xmath5 , or @xmath6 respectively ) , and @xmath7 secondary driver mutations . right : cells with only secondary driver mutations have neutral or nearly neutral fitness . the fitness of cells with the primary driver mutation depends on the number of secondary driver mutations within the cell , leading to an epistatic fitness landscape.,title="fig:",scaledwidth=50.0% ] + denote the number of cells with or without the primary driver mutation ( @xmath5 , or @xmath6 respectively ) , and @xmath7 secondary driver mutations . right : cells with only secondary driver mutations have neutral or nearly neutral fitness . the fitness of cells with the primary driver mutation depends on the number of secondary driver mutations within the cell , leading to an epistatic fitness landscape.,title="fig:",scaledwidth=49.0% ] a cell s probability for apoptosis and proliferation depends on the presence of the primary driver mutation and on the number of secondary driver mutations it has accumulated . for cells with no mutations , the division and apoptosis probabilities are both equal to @xmath8 . this implies that the number of cells is constant on average as long as no further mutations occur . we assume that the initial number of cells is high and thus we can neglect that the population would go extinct @xcite . for our parameter values , the expected extinction time of our critical branching process exceeds the average life time of the organism by far . for cells without the primary driver mutation , each secondary driver mutation leads to a change in the cell s fitness by @xmath9 . for cells with the primary driver mutation , the fitness advantage obtained with each secondary driver mutation is @xmath10 . the driver mutation increases both the apoptosis rate and the proliferation rate . the increase in the apoptosis rate is @xmath11 and the increase in the division rate is @xmath12 . with these parameters , the proliferation rate @xmath13 for cells with @xmath7 secondary driver mutations but without the primary driver mutation is @xmath14 whereas the proliferation rate @xmath15 for such cells with the primary driver mutation is @xmath16 the apoptosis rates , denoted as @xmath17 and @xmath18 are simply one minus the proliferation rate @xmath19 mutations occur at fixed rates @xmath1 and @xmath2 for primary and secondary drivers , respectively . for a long time , the overall fitness does not increase noticeably . for @xmath20 , it stays on average constant . hence , the total number of cells stays approximately constant . only when a cell with enough secondary driver mutations and also the primary driver mutation arises , the cell s fitness is increased substantially beyond the fitness of other cells and its chance of proliferation is significantly increased . at that point , the total number of cells starts to increase rapidly , see figure [ fig : total ] . in models presented in literature so far , the cell s fitness is increased independently with every ( driver ) mutation ( see e.g. * ? ? ? * ; * ? ? ? although the total number of cells increases exponentially , these models do not find a sudden burst in the number of cells . , secondary driver fitness advantage @xmath21 , the primary driver fitness advantage @xmath22 , primary driver disadvantage @xmath23 , advantage of a secondary driver mutation in the presence of the primary driver mutation @xmath24 , mutation rates for secondary driver mutations @xmath25 , mutation rate for the primary driver mutation @xmath26 ) . , scaledwidth=80.0% ] , @xmath21 , @xmath22 , @xmath23 , @xmath24 , @xmath25 , @xmath26),scaledwidth=100.0% ] in figure [ fig : single ] , the total number of cells is subdivided into the number of cells with different numbers of mutations . the left panel presents the cells that have not acquired the primary driver mutation , the right one shows cells with the primary driver mutation . cells with the primary driver mutation , but not enough secondary driver mutations , arise occasionally , but those cells die out quickly again thus , their average abundance is small . cells without the primary driver mutation do not die out , they also do not induce fast growth , cf . figure [ fig : single ] . only cells that have obtained enough secondary driver mutations and in addition acquire the primary driver mutation , divide so quickly that the population size increases rapidly . the parameters in our figures have been chosen such that a cells acquires a substantial growth advantage once the primary driver mutation co - occurs with 4 secondary driver mutations . this event can occur at any time and hence , in some simulation the number of cells can increase very early , whereas in other simulations the number of cells does not undergo fast proliferation for many generations . consequently , the rate of progression has an enormous variation . for the parameters from our figures , the time at which rapid proliferation occurs varied between @xmath27 and @xmath28 generations in 500 simulations . the distribution of these times is discussed in more detail below . we can calculate the average number of cells with a certain number of mutations at a given generation @xmath29 . the number of cells which do not have the primary driver mutation and @xmath30 secondary driver mutations ( i.e. @xmath31 ) changes on average by means of the cell s fitness and it decreases by the mutation rate @xmath32 where @xmath33 . the solution of equation for @xmath34 , i.e.if the secondary driver mutations are not neutral , is @xmath35 where @xmath0 denotes the initial number of cells . the mathematical proof of equation is given in [ subsec : withoutdriver ] . note , that the product can be written in terms of a @xmath36-binomial coefficient @xcite , @xmath37 for the case @xmath38 , we take the limit of the @xmath36-binomial coefficient ( e.g. * ? ? ? * ) @xmath39 and obtain @xmath40 which is the result that is also expected if the secondary driver mutations are neutral and accumulated independently of each other . intuitively , the term @xmath41 describes the probability of obtaining exactly @xmath30 mutations in @xmath29 generations . there are different possibilities when the mutations happen , these possibilities are captured by the binomial coefficient @xmath42 . thus , we have a growing polynomial term in @xmath29 and a declining exponential term in @xmath29 , since @xmath43 . in the case of @xmath44 , the interpretation is similar . here , additionally the fitness advantage for secondary driver mutations has to be taken into account . since the number of cells with @xmath7 secondary driver mutations grows with @xmath45 , also the number of cells that can mutate grows . hence , the factor @xmath46 is multiplied to the expression and the binomial coefficient turns into the @xmath36-binomial coefficient . [ tbl : abbrev ] .summary of our abbreviations [ cols= " < , < " , ] for cells that have obtained the primary driver mutation and @xmath30 secondary driver mutations , the situation is slightly more complex . there are @xmath47 different possibilities on how to obtain @xmath30 secondary driver and the primary driver mutation , since some of the secondary driver mutations may have occurred before the primary driver mutation has been acquired whereas others may have occurred afterwards . let @xmath48 denote the number of cells with the primary driver mutation and @xmath30 secondary driver mutations , when the primary driver mutation has happened in a cell with @xmath49 secondary driver mutations . note that @xmath50 . the change in the number of cells now depends on @xmath49 . using the abbreviations from table [ tbl : abbrev ] to simplify our notation , we have @xmath51 to express the average number of cells in total we need to sum over all possible pathways , @xmath52 in [ subsec : withdriver ] , we proof that the analytical solution of equation is @xmath53 , \end{split}\end{aligned}\ ] ] if @xmath34 . the summation over @xmath49 indicates the different mutational pathways . an intuitive explanation of this somewhat lengthy equation is given in [ sec : intuitive ] . interestingly , the case for @xmath20 is much more challenging . the underlying problem is that the normal binomial coefficient can not be expressed in a sum in the way the @xmath36-binomial coefficient can be expressed @xcite , @xmath54 when summing over all generations of the population with @xmath49 secondary driver mutations to derive the expression for the population of cells with @xmath55 secondary driver mutations , we have to calculate the sum @xmath56 when we go further and try to calculate the expression for the population with @xmath30 secondary driver mutations , we need to apply this sum @xmath57-times and hence we obtain a multi sum , @xmath58 only an analytical expression for this multi sum would allow a closed solution of the problem with @xmath20 . also taking the limit @xmath59 of our expression for @xmath44 is a substantial mathematical challenge . however , we can use our solution for @xmath44 for arbitrarily small values of @xmath9 . moreover , numerical considerations show that the result for @xmath20 is very close to the case of @xmath60 . in figure [ fig : dyn ] , the dynamics of the average number of cells with a certain number of mutations , is shown , both without and with the primary driver mutation . simulation results for @xmath20 agree very well with the analytical result obtained for @xmath44 . ( circles ) agree almost perfectly with the analytical result obtained for @xmath61 . the bars represent the standard deviation . cells with no mutation have a very small relative standard deviation and cells with one mutation ( i.e. one passenger only or the driver only ) have a relatively small standard deviation . in contrast , cells with two passenger mutations for instance have a very broad standard deviation in the beginning that is approximately four times the average number . only in few realizations , a primary driver mutations co - occurs with several secondary drivers , hence the simulation data for these cases shows a large spread ( parameters : @xmath62 , @xmath21 , @xmath22 , @xmath23 , @xmath24 , @xmath25 , @xmath26).,scaledwidth=100.0% ] next , let us calculate the distribution of the time it takes until rapid proliferation occurs . since we use a multi type , time discrete branching process , we can make use of the probability generating function to recursively calculate the probability for a certain cell type cell to be present at a certain time @xmath29 @xcite . of particular interest is the probability that a tumor initiating cell is present , i.e. in the example above a cell with the primary driver mutation and 4 secondary driver mutations . let @xmath63 be the probability generation function for the cell type @xmath4 in the branching process described above . the probability that there is no cell with the primary driver mutation and @xmath30 secondary driver mutations at time @xmath29 , when the process starts with one @xmath64-cell , can be computed by recursively calculating @xmath65 , where @xmath66 starting with @xmath0 cells , we have to consider @xmath67 . thus , the probability that there is at least one tumor cell at time @xmath29 is @xmath68 . if we are interested in the probability @xmath69 that there is at least one tumor cell exactly at time @xmath29 , we need to subtract the probability , that there has been a tumor cell before @xmath70 . figure [ fig : timecompare ] shows the good agreement between simulations and the recursive calculation . shown as a dashed line ( parameters : @xmath62 , @xmath21 , @xmath22 , @xmath23 , @xmath24 , @xmath25 , @xmath26 , distribution over 20000 independent realizations ) . , scaledwidth=70.0% ] the time distribution for low @xmath29 follows a power law , as shown in the inset of figure [ fig : timecompare ] . the exponent of the power law is approximately @xmath71 . if all mutations were neutral , one would expect a lead coefficient of approximately 4 to accumulate five mutations , as derived by @xcite . in our case , the curve increases slower . numerical considerations show that the main reason for this is that , in contrast to @xcite , we allow extinction : many lineages that have accumulated mutations go extinct before the final , cancer causing mutation arises . most models in literature assume that each mutation leads to an independent and steady increase in the cells fitness @xcite . in this context , neutral passenger mutations have no causal impact on cancer progression . only recently , some authors have considered passenger mutations not only as neutral byproducts of the clonal expansion of mutagenic cells , but as having a deleterious impact on the cells fitness @xcite . here , we have described a model in which the fitness of the driver mutation strongly depends on the number of passenger mutations the cell has acquired . these passenger mutations , which we have termed secondary driver mutations , lead only to a small change in fitness or no change in fitness at all . as illustrated in figures [ fig : total ] , [ fig : single ] , and [ fig : dyn ] , the number of cells stays roughly constant for a long time before it rapidly increases , despite the fact that mutations occur in the process permanently . this dynamic effect of cancer initiation is very different from models in which mutations do not interact with each other . we speculate that this kind of dynamics can have important implications for diagnosis and treatment . in principle , the dynamics presented in figure [ fig : total ] can also be the result of one highly advantageous , but very unlikely driver mutation . but in such a case , cells with the driver mutation should not be present in the population before tumorigenesis . this contradicts with current knowledge about the _ myc _ translocation which has also been detected in humans without lymphoma @xcite . this effect is well captured by our model , as shown in figure [ fig : single ] . in some tumors , such as burkitt lymphoma , the neoplasms is only diagnosed after fast tumor growth has started . in this case , sequencing studies have shown that several mutations are present at the time of examination @xcite . since the patients typically do not have any symptoms in before diagnosis of the cancer , it is possible that some mutations have virtually no direct impact on the cells fitness . nevertheless , they are necessary for the initiation of the cancer , as they indirectly allow the driver mutation to initiate rapid cell growth . this agrees well with our epistatic model , where ( nearly ) neutral secondary driver mutations occur at a fixed rate before the cancer can be diagnosed . of course , not all mutations have such an epistatic effect on primary driver mutations , some might even be considered deleterious @xcite . nevertheless , our work shows that mutations that appear to be neutral in one context should not only be regarded as a neutral byproduct of the clonal expansion of mutagenic cells . instead , in some cases passenger mutations can have a serious impact in cancer initiation , in particular when there are non - trivial interactions between different mutations . in this case the term passenger " may not be the most appropriate one . to understand the impact of those interactions can be essential for a deeper understanding of the initiation of cancer . we thank j. richter for stimulating discussions on burkitt lymphoma and b. werner for helpful comments on our manuscript . generous funding by the max - planck society is gratefully acknowledged . r.s . is supported the german ministry of education and science ( bmbf ) through the mmml - myc - sys network systems biology of myc - positive lymphomas ( 036166b ) we first consider the case without the primary driver mutation . we assume @xmath72 ( and consequently @xmath73 ) , as discussed in the main text . the rate change of cells with @xmath30 secondary driver mutations is @xmath74 where @xmath33 . the solution of is formulated in the following theorem : [ thm : x0k1 ] for any integer @xmath75 , the number of cells with @xmath30 secondary driver mutations and no primary driver mutation is @xmath76 if is a solution , then it must satisfy . since solutions for recursive functions are always unique , would be the only solution . hence , we proof by inserting the equation on the right hand side of . @xmath77 we can write each product as a @xmath36-binomial coefficient , @xmath78 . thus , with the @xmath36-pascal rule @xcite @xmath79 equation simplifies to @xmath80 which concludes the proof . we now turn to the cells which have obtained the primary driver mutation . as discussed in the main text , we only look at the case where the fitness change of the secondary driver mutation is not equal to zero , @xmath72 . while for cells without the primary driver mutation there is only one mutational pathway , cells with the primary driver mutation can be reached via different mutational pathways , because cells that get the primary driver mutation might have different amounts of secondary driver mutations . hence , we need to sum over all those possible pathways . let @xmath49 be the number of secondary drivers that are present in the cell which acquires the primary driver mutation . then @xmath48 denotes the number of cells with the primary driver mutation and @xmath30 secondary driver mutations , when the primary driver mutation has happened in a cell with @xmath49 secondary driver mutations ( @xmath50 ) . with this , the total number of cells with the primary driver mutation is @xmath81 the change in the number of cells now depends on @xmath49 . we have @xmath82 the solution of is given by the following theorem : [ thm : x1kp ] the average number of cells with the primary driver mutation and @xmath30 secondary driver mutations , given that the primary driver mutation happens in a cell with @xmath49 secondary driver mutations , is given by @xmath83 , \end{split } \label{driverexp}\end{aligned}\ ] ] where the function @xmath84 is defined @xmath85 again we proof the theorem by inserting in and showing that the equality holds true . we need to distinguish between the two cases as in . first , we proof the theorem for the case @xmath86 . @xmath87 \\ & \quad + \mu_\mathrm{p}\varsigma_\mathrm{d}\varsigma_\mathrm{dp}^{k-1 } n\mu_d\mu_p^{k-1 } \varsigma_\mathrm{d}^{k - p-1 } \varsigma_\mathrm{dp}^{((k-1)(k-2)-p(p-1))/2 } \frac{\varsigma_\mathrm{p}^{p(p+1)/2}}{\prod_{n=0}^{p-1}\left(1-\varsigma_\mathrm{p}^{n+1}\right ) } \\ & \quad \times \left [ \nu_{\mathrm{p}}^{t - p-1 } \psi_{p , k-1}(t-1 ) - \sum_{j = p}^{k-1}\nu_{\mathrm{p}}^{j - p}\left(\nu_\mathrm{d}\varsigma_\mathrm{dp}^j\varsigma_\mathrm{d}\right)^{t - k } \psi_{p , j}(j)\prod_{n = j}^{k-2}\frac{1-\varsigma_\mathrm{dp}^{t - n-2}}{1-\varsigma_\mathrm{dp}^{k - n-1 } } \right ] \\ & \quad = n\mu_d\mu_p^k \varsigma_\mathrm{d}^{k - p } \varsigma_\mathrm{dp}^{(k(k-1)-p(p-1))/2 } \frac{\varsigma_\mathrm{p}^{p(p+1)/2}}{\prod_{n=0}^{p-1}\left(1-\varsigma_\mathrm{p}^{n+1}\right ) } \\ & \quad \times \left [ \nu_\mathrm{p}^{t - p-1}\left ( \nu_\mathrm{d}\varsigma_\mathrm{d}\varsigma_\mathrm{dp}^k \psi_{p , k}(t-1 ) + \psi_{p , k-1}(t-1 ) \right ) - \nu_\mathrm{p}^{k - p}(\nu_\mathrm{d}\varsigma_\mathrm{d}\varsigma_\mathrm{dp}^k)^{t - k}\psi_{p , k}(k ) \right.\\ & \quad - \left . \sum_{j = p}^{k-1 } \nu_\mathrm{p}^{j - p } \psi_{p , j}(j ) \left(\nu_\mathrm{d}\varsigma_\mathrm{d}\varsigma_\mathrm{dp}^j \right)^{t - k}\left ( \varsigma_\mathrm{dp}^{k - j}\prod_{n = j}^{k-1}\frac{1-\varsigma_\mathrm{dp}^{t - n-2}}{1-\varsigma_\mathrm{dp}^{k - n}}+\prod_{n = j}^{k-2}\frac{1-\varsigma_\mathrm{dp}^{t - n-2}}{1-\varsigma_\mathrm{dp}^{k - n-1 } } \right ) \right ] \end{split}\end{aligned}\ ] ] when we compare and we see , that the two equations are equal if @xmath88 and @xmath89 for equation , we have @xmath90 for equation , we need to insert the definition of @xmath84 @xmath91 this concludes the proof for the case @xmath86 . now we look at the case @xmath92 . we have @xmath93\\ & \quad + \mu_\mathrm{d}\varsigma_\mathrm{p}^k n\mu_\mathrm{p}^k \nu_\mathrm{p}^{t - k-1 } \varsigma_\mathrm{p}^{k(k-1)/2 } \prod_{n=0}^{k-1}\frac{1-\varsigma_\mathrm{p}^{t - n-1}}{1-\varsigma_\mathrm{p}^{n+1}}\\ & \quad = n \mu_\mathrm{d}\mu_\mathrm{p}^k \frac{\varsigma_\mathrm{p}^{k(k+1)/2}}{\prod_{n=0}^{k-1}(1-\varsigma_\mathrm{p}^{n+1})}\left [ \nu_\mathrm{d}\varsigma_\mathrm{d}\varsigma_\mathrm{dp}^k\nu_\mathrm{p}^{t - k-1 } \psi_{k , k}(t-1 ) - \left(\nu_\mathrm{d}\varsigma_\mathrm{d}\varsigma_\mathrm{dp}^k\right)^{t - k } \psi_{k , k}(k ) + \nu_\mathrm{p}^{t - k-1}\prod_{n=0}^{k-1}\left ( 1-\varsigma_\mathrm{p}^{t - n-1 } \right ) \right].\end{aligned}\ ] ] in order for this to be equal to @xmath94 , we need @xmath95 analogue to this equation holds true if @xmath96 by writing the summation as a @xmath36-pochhammer symbol , we have @xmath97 this concludes the proof also for @xmath92 . here , we try to understand this equation in a more intuitive way . for each generation @xmath29 , the number of possibilities to distribute the @xmath49 secondary driver mutations over @xmath29 time steps is given by the @xmath36-binomial coefficient @xmath98 . but the growth of the cells depends on the time when the secondary driver mutations are first acquired . due to fitness advantage , the earlier the mutations have been acquired , the faster the population grows , and also the sooner the primary driver mutation can be obtained . as in equation ( 5 ) , the effect of the fitness advantage on the cells without the primary driver mutation itself is captured by multiplying @xmath99 . the effect on the primary driver mutation is more intricate . to capture this effect , we start from a @xmath36-binomial coefficient and rewrite the @xmath36-pochhammer symbol in the numerator @xmath100 in terms of a sum @xcite , @xmath101 to make this resemble the term in the parentheses in the second line of equation ( 11 ) , we divide the numerator by @xmath102 and we obtain @xmath103 with @xmath104 equation can be written as @xmath105 for the numerator of this modified @xmath36-binomial coefficient , we introduce the abbreviation @xmath85 in terms of this @xmath84-function equation ( 11 ) can be written in a more compact form as @xmath106 . \end{split}\end{aligned}\ ] ] alexandrov , l. b. , nik - zainal , s. , wedge , d. c. , aparicio , s. a. j. r. , behjati , s. , biankin , a. v. , bignell , g. r. , bolli , n. , borg , a. , brresen - dale , a .- l . , boyault , s. , burkhardt , b. , butler , a. p. , caldas , c. , davies , h. r. , desmedt , c. , eils , r. , eyfjrd , j. e. , foekens , j. a. , greaves , m. , hosoda , f. , hutter , b. , ilicic , t. , imbeaud , s. , imielinski , m. , imielinsk , m. , jger , n. , jones , d. t. w. , jones , d. , knappskog , s. , kool , m. , lakhani , s. r. , lpez - otn , c. , martin , s. , munshi , n. c. , nakamura , h. , northcott , p. a. , pajic , m. , papaemmanuil , e. , paradiso , a. , pearson , j. v. , puente , x. s. , raine , k. , ramakrishna , m. , richardson , a. l. , richter , j. , rosenstiel , p. , schlesner , m. , schumacher , t. n. , span , p. n. , teague , j. w. , totoki , y. , tutt , a. n. j. , valds - mas , r. , van buuren , m. m. , van t veer , l. , vincent - salomon , a. , waddell , n. , yates , l. r. , australian pancreatic cancer genome initiative , icgc breast cancer consortium , icgc mmml - seq consortium , icgc pedbrain , zucman - rossi , j. , futreal , p. a. , mcdermott , u. , lichter , p. , meyerson , m. , grimmond , s. m. , siebert , r. , campo , e. , shibata , t. , pfister , s. m. , campbell , p. j. , stratton , m. r. , 2013 . signatures of mutational processes in human cancer . nature 500 , 415421 . beerenwinkel , n. , antal , t. , dingli , d. , traulsen , a. , kinzler , k. w. , velculescu , v. e. , vogelstein , b. , nowak , m. a. , 2007 . genetic progression and the waiting time to cancer . plos computational biology 3 , e225 . birch , j. m. , blair , v. , kelsey , a. m. , evans , d. g. , harris , m. , tricker , k. j. , varley , j. m. , sep 1998 . cancer phenotype correlates with constitutional tp53 genotype in families with the li - fraumeni syndrome . oncogene 17 ( 9 ) , 10618 . bozic , i. , antal , t. , ohtsuki , h. , carter , h. , kim , d. , chen , s. , karchin , r. , kinzler , k. w. , vogelstein , b. , nowak , m. a. , 2010 . accumulation of driver and passenger mutations during tumor progression . proceedings of the national academy of sciences usa 107 , 1854518550 . frhling , s. , scholl , c. , levine , r. l. , loriaux , m. , boggon , t. j. , bernard , o. a. , berger , r. , dhner , h. , dhner , k. , ebert , b. l. , teckie , s. , golub , t. r. , jiang , j. , schittenhelm , m. m. , lee , b. h. , griffin , j. d. , stone , r. m. , heinrich , m. c. , deininger , m. w. , druker , b. j. , gilliland , d. g. , 2007 . identification of driver and passenger mutations of flt3 by high - throughput dna sequence analysis and functional assessment of candidate alleles . cancer cell 12 , 501513 . greenman , c. , stephens , p. , smith , r. , dalgliesh , g. l. , hunter , c. , bignell , g. , davies , h. , teague , j. , butler , a. , stevens , c. , edkins , s. , omeara , s. , vastrik , i. , schmidt , e. e. , avis , t. , barthorpe , s. , bhamra , g. , buck , g. , choudhury , b. , clements , j. , cole , j. , dicks , e. , forbes , s. , gray , k. , halliday , k. , harrison , r. , hills , k. , hinton , j. , jenkinson , a. , jones , d. , menzies , a. , mironenko , t. , perry , j. , raine , k. , richardson , d. , shepherd , r. , small , a. , tofts , c. , varian , j. , webb , t. , west , s. , widaa , s. , yates , a. , cahill , d. p. , louis , d. n. , goldstraw , p. , nicholson , a. g. , brasseur , f. , looijenga , l. , weber , b. l. , chiew , y .- e . , defazio , a. , greaves , m. f. , green , a. r. , campbell , p. , birney , e. , easton , d. f. , chenevix - trench , g. , tan , m .- h . , khoo , s. k. , teh , b. t. , yuen , s. t. , leung , s. y. , wooster , r. , futreal , p. a. , stratton , m. r. , mar 2007 . patterns of somatic mutation in human cancer genomes . nature 446 , 1538 . hummel , m. , bentink , s. , berger , h. , klapper , w. , wessendorf , s. , barth , t. f. e. , bernd , h .- w . , cogliatti , s. b. , dierlamm , j. , feller , a. c. , hansmann , m .- haralambieva , e. , harder , l. , hasenclever , d. , khn , m. , lenze , d. , lichter , p. , martin - subero , j. i. , mller , p. , mller - hermelink , h .- k . , ott , g. , parwaresch , r. m. , pott , c. , rosenwald , a. , rosolowski , m. , schwaenen , c. , strzenhofecker , b. , szczepanowski , m. , trautmann , h. , wacker , h .- h . , spang , r. , loeffler , m. , trmper , l. , stein , h. , siebert , r. , molecular mechanisms in malignant lymphomas network project of the deutsche krebshilfe , jun 2006 . a biologic definition of burkitt s lymphoma from transcriptional and genomic profiling . new england journal of medicine 354 ( 23 ) , 241930 . jones , s. , chen , w .- d . , parmigiani , g. , diehl , f. , beerenwinkel , n. , antal , t. , traulsen , a. , nowak , m. a. , siegel , c. , velculescu , v. , kinzler , k. w. , vogelstein , b. , willis , j. , markowitz , s. , 2008 . comparative lesion sequencing provides insights into tumor evolution . proceedings of the national academy of sciences usa 105 , 42834288 . justilien , v. , walsh , m. p. , ali , s. a. , thompson , e. a. , murray , n. r. , fields , a. p. , 2014 . the prkci and sox2 oncogenes are coamplified and cooperate to activate hedgehog signaling in lung squamous cell carcinoma . cancer cell 25 , 139151 . love , c. , sun , z. , jima , d. , li , g. , zhang , j. , miles , r. , richards , k. l. , dunphy , c. h. , choi , w. w. l. , srivastava , g. , lugar , p. l. , rizzieri , d. a. , lagoo , a. s. , bernal - mizrachi , l. , mann , k. p. , flowers , c. r. , naresh , k. n. , evens , a. m. , chadburn , a. , gordon , l. i. , czader , m. b. , gill , j. i. , hsi , e. d. , greenough , a. , moffitt , a. b. , mckinney , m. , banerjee , a. , grubor , v. , levy , s. , dunson , d. b. , dave , s. s. , 2012 . the genetic landscape of mutations in burkitt lymphoma . nat genet 44 , 13211325 . mcfarland , c. d. , korolev , k. s. , kryukov , g. v. , sunyaev , s. r. , mirny , l. a. , feb 2013 . impact of deleterious passenger mutations on cancer progression . proceedings of the national academy of sciences usa 110 , 29105 . mller , j. r. , janz , s. , goedert , j. j. , potter , m. , rabkin , c. s. , jul 1995 . persistence of immunoglobulin heavy chain / c - myc recombination - positive lymphocyte clones in the blood of human immunodeficiency virus - infected homosexual men . proceedings of the national academy of sciences usa 92 ( 14 ) , 657781 . nie , z. , hu , g. , wei , g. , cui , k. , yamane , a. , resch , w. , wang , r. , green , d. r. , tessarollo , l. , casellas , r. , zhao , k. , levens , d. , 2012 . c - myc is a universal amplifier of expressed genes in lymphocytes and embryonic stem cells . cell 151 , 6879 . nowak , m. , komarova , n. l. , sengupta , a. , jallepalli , p. , shih , i. , vogelstein , b. , lengauer , c. , 2002 . the role of chromosomal instability in tumour initiation . proceedings of the national academy of sciences usa 99 ( 25 ) , 1622616231 . pelengaris , s. , khan , m. , evan , g. i. , may 2002 . suppression of myc - induced apoptosis in beta cells exposes multiple oncogenic properties of myc and triggers carcinogenic progression . cell 109 ( 3 ) , 32134 . richter , j. , schlesner , m. , hoffmann , s. , kreuz , m. , leich , e. , burkhardt , b. , rosolowski , m. , ammerpohl , o. , wagener , r. , bernhart , s. h. , lenze , d. , szczepanowski , m. , paulsen , m. , lipinski , s. , russell , r. b. , adam - klages , s. , apic , g. , claviez , a. , hasenclever , d. , hovestadt , v. , hornig , n. , korbel , j. o. , kube , d. , langenberger , d. , lawerenz , c. , lisfeld , j. , meyer , k. , picelli , s. , pischimarov , j. , radlwimmer , b. , rausch , t. , rohde , m. , schilhabel , m. , scholtysik , r. , spang , r. , trautmann , h. , zenz , t. , borkhardt , a. , drexler , h. g. , mller , p. , macleod , r. a. f. , pott , c. , schreiber , s. , trmper , l. , loeffler , m. , stadler , p. f. , lichter , p. , eils , r. , kppers , r. , hummel , m. , klapper , w. , rosenstiel , p. , rosenwald , a. , brors , b. , siebert , r. , icgc mmml - seq project , 2012 . recurrent mutation of the id3 gene in burkitt lymphoma identified by integrated genome , exome and transcriptome sequencing . nature genetics 44 ( 12 ) , 131620 . sander , s. , calado , d. p. , srinivasan , l. , kchert , k. , zhang , b. , rosolowski , m. , rodig , s. j. , holzmann , k. , stilgenbauer , s. , siebert , r. , bullinger , l. , rajewsky , k. , aug 2012 . synergy between pi3k signaling and myc in burkitt lymphomagenesis . cancer cell 22 ( 2 ) , 16779 . schmitz , r. , young , r. m. , ceribelli , m. , jhavar , s. , xiao , w. , zhang , m. , wright , g. , shaffer , a. l. , hodson , d. j. , buras , e. , liu , x. , powell , j. , yang , y. , xu , w. , zhao , h. , kohlhammer , h. , rosenwald , a. , kluin , p. , mller - hermelink , h. k. , ott , g. , gascoyne , r. d. , connors , j. m. , rimsza , l. m. , campo , e. , jaffe , e. s. , delabie , j. , smeland , e. b. , ogwang , m. d. , reynolds , s. j. , fisher , r. i. , braziel , r. m. , tubbs , r. r. , cook , j. r. , weisenburger , d. d. , chan , w. c. , pittaluga , s. , wilson , w. , waldmann , t. a. , rowe , m. , mbulaiteye , s. m. , rickinson , a. b. , staudt , l. m. , oct 2012 . burkitt lymphoma pathogenesis and therapeutic targets from structural and functional genomics . nature 490 ( 7418 ) , 11620 . sjblom , t. , jones , s. , wood , l. , parsons , d. , lin , j. , barber , t. , mandelker , d. , leary , r. , ptak , j. , silliman , n. , szabo , s. , buckhaults , p. , farrell , c. , meeh , p. , markowitz , s. , willis , j. , dawson , d. , willson , j. , gazdar , a. , hartigan , j. , wu , l. , liu , c. , parmigiani , g. , park , b. , bachman , k. , papadopoulos , n. , vogelstein , b. , kinzler , k. , velculescu , v. , 2006 . the consensus coding sequences of human breast and colorectal cancers . science 314 , 268274 . szappanos , b. , kovcs , k. , szamecz , b. , honti , f. , costanzo , m. , baryshnikova , a. , gelius - dietrich , g. , lercher , m. j. , jelasity , m. , myers , c. l. , andrews , b. j. , boone , c. , oliver , s. g. , pl , c. , papp , b. , jul 2011 . an integrated approach to characterize genetic interaction networks in yeast metabolism . nature genetics 43 ( 7 ) , 65662 . szendro , i. g. , franke , j. , de visser , j. a. g. m. , krug , j. , 2013 . predictability of evolution depends nonmonotonically on population size . proceedings of the national academy of sciences usa 110 , 571576 . wang , c. , tai , y. , lisanti , m. p. , liao , d. j. , 2011 . c - myc induction of programmed cell death may contribute to carcinogenesis : a perspective inspired by several concepts of chemical carcinogenesis . cancer biology and therapy 11 , 615626 . wood , l. d. , parsons , d. w. , jones , s. , lin , j. , sjoblom , t. , leary , r. j. , shen , d. , boca , s. m. , barber , t. , ptak , j. , silliman , n. , szabo , s. , dezso , z. , ustyanksky , v. , nikolskaya , t. , nikolsky , y. , karchin , r. , wilson , p. a. , kaminker , j. s. , zhang , z. , croshaw , r. , willis , j. , dawson , d. , shipitsin , m. , willson , j. k. v. , sukumar , s. , polyak , k. , park , b. h. , pethiyagoda , c. l. , pant , p. v. k. , ballinger , d. g. , sparks , a. b. , hartigan , j. , smith , d. r. , suh , e. , papadopoulos , n. , buckhaults , p. , markowitz , s. d. , parmigiani , g. , kinzler , k. w. , velculescu , v. e. , vogelstein , b. , 2007 . the genomic landscapes of human breast and colorectal cancers . science 318 ( 5853 ) , 11081113 . zech , l. , haglund , u. , nilsson , k. , klein , g. , 1976 . characteristic chromosomal abnormalities in biopsies and lymphoid - cell lines from patients with burkitt and non - burkitt lymphomas . int j cancer 17 , 4756 .	we investigate the dynamics of cancer initiation in a mathematical model with one driver mutation and several passenger mutations . our analysis is based on a multi type branching process : we model individual cells which can either divide or undergo apoptosis . in case of a cell division , the two daughter cells can mutate , which potentially confers a change in fitness to the cell . in contrast to previous models , the change in fitness induced by the driver mutation depends on the genetic context of the cell , in our case on the number of passenger mutations . the passenger mutations themselves have no or only a very small impact on the cell s fitness . while our model is not designed as a specific model for a particular cancer , the underlying idea is motivated by clinical and experimental observations in burkitt lymphoma . in this tumor , the hallmark mutation leads to deregulation of the _ myc _ oncogene which increases the rate of apoptosis , but also the proliferation rate of cells . this increase in the rate of apoptosis hence needs to be overcome by mutations affecting apoptotic pathways , naturally leading to an epistatic fitness landscape . this model shows a very interesting dynamical behavior which is distinct from the dynamics of cancer initiation in the absence of epistasis . since the driver mutation is deleterious to a cell with only a few passenger mutations , there is a period of stasis in the number of cells until a clone of cells with enough passenger mutations emerges . only when the driver mutation occurs in one of those cells , the cell population starts to grow rapidly . cancer , modeling , somatic evolution , population dynamics
the two - dimensional ising model is one of the most well - studied models in statistical mechanics . given a discrete planar domain @xmath0 ( a bounded subset of the square grid ) , the ising model in @xmath0 can be viewed either as a random assignment of spins to the _ faces _ of @xmath0 , or a random collection of edges of @xmath0 , with an edge drawn between each pair of faces having different spins . the partition function of the model is given by @xmath1 \quad \text { or } \quad \sum\limits_{s\in{\mathrm{conf}}({\omega^{\delta } } ) } x^{\|s\|},\ ] ] respectively , where @xmath2 denotes the set of faces of @xmath0 and @xmath3 is the set of subgraphs @xmath4 of @xmath0 such that all vertices of @xmath0 have even degrees in @xmath4 . we refer the reader to section [ sec : notations ] for a more detailed discussion and notation . we will be interested in the properties of the model at the critical temperature @xmath5 , which corresponds to @xmath6 . this value of @xmath7 will be fixed throughout the paper . discrete holomorphic observables , also called holomorphic fermions or fermionic observables , were proposed by smirnov in @xcite as a tool to study the critical ising model , although similar objects appeared earlier in @xcite and @xcite without discussing corresponding boundary value problems . since then , these observables proved to be very useful for a rigorous analysis of the planar ising model at criticality in the scaling limit when @xmath0 approximates some continuous domain @xmath8 as the lattice mesh @xmath9 tends to zero . recall that smirnov s fermionic observable is defined as @xmath10 where @xmath11 is the set of edge subsets @xmath4 , such that @xmath4 can be decomposed into a disjoint collection of loops and a simple lattice path @xmath12 connecting a boundary edge @xmath13 to the midpoint @xmath14 of an inner edge @xmath15 ; @xmath16 is the winding number of @xmath12 ; and @xmath17 denotes the orientation of the outgoing boundary edge @xmath13 . with this definition , the observable has been shown to be discrete holomorphic and satisfy riemann - type boundary conditions @xmath18 this led to a proof of its convergence to a conformally covariant scaling limit @xcite . this result has been the main ingredient of the recent progress in rigorous understanding of conformal invariance in the critical two - dimensional ising model . the martingale property of @xmath19 ( see further details in @xcite ) allows one to prove convergence of the ising interfaces to the chordal schramm s @xmath20 curves . using a slightly different version of this observable , hongler and smirnov @xcite were able to compute the scaling limit of the energy density in the critical ising model on the square grid , including the lattice dependent constant before the conformally covariant factor . this result was later extended to all correlations of the energy density field and certain boundary spin correlations @xcite . at the same time , similar observables proved to be very useful in the analysis of the random cluster ( fortuin - kasteleyn ) representation of the critical ising model @xcite . in particular , it was shown by beffara and duminil - copin @xcite that they can be used to give a short proof of criticality of the ising model at the self - dual point . many of these results generalize beyond the case of square grid approximations . thus , convergence of fermionic observables has been proven for isoradial lattices @xcite , which reappeared in the connection with the critical ising model in the paper of mercat @xcite . this proved the universality phenomenon , i.e. , the fact that a microscopic structure of the lattice does not affect macroscopic properties of the scaling limit . moreover , discrete complex analysis technique developed in @xcite and @xcite provides a general framework for such universal proofs . on the other hand , one of the most natural questions about the ising model the rigorous proof of conformal covariance of spin correlations in the scaling limit remained out of reach until recently . the goal of the present work is to introduce a new tool spinor holomorphic observables that allows to attack this problem . in particular , we prove convergence of ratios of spin correlations corresponding to different boundary conditions to conformally invariant limits . in a subsequent joint paper with clment hongler @xcite , using a more elaborate version of the spinor observables , we prove conformal covariance of spin correlations themselves . in this paper we extend the study of fermionic observables to the case of multiply connected domains . given a double cover @xmath21 of such a domain , we define the observable @xmath22 by @xmath23 where @xmath24 , but the sum is taken over _ the same set of configurations _ as before ; @xmath25 is the number of loops in @xmath4 that do not lift as closed loops to @xmath26 , and @xmath27 is the indicator of the event that @xmath12 lifts to @xmath26 as a path running from @xmath13 to @xmath28 ( and not to the other sheet ) , see section [ sec : observables_limits ] for detailed discussion . in other words , we plug into ( [ smirnovf ] ) an additional sign that depends on homology class of @xmath4 modulo two . it is worth to mention that our observables should be closely related to the vector bundle laplacian technique applied to uniform spanning trees and double dimers by kenyon @xcite , although at the moment we do not know any exact correspondence of that sort . our main observation is that @xmath29 are discrete holomorphic and satisfy the boundary conditions ( [ eq : rim_bc ] ) , just like smirnov s observable @xmath19 . the definition implies that @xmath30 , if @xmath31 belong to a fiber of the same point ; hence , we call @xmath32 _ holomorphic spinors_. to describe the scaling limits of @xmath33 , we will introduce the continuous holomorphic spinors @xmath34 . roughly speaking , these are fundamental solutions to the continuous riemann boundary value problem ( [ eq : rim_bc ] ) on the double - cover @xmath35 , with a singularity at @xmath13 and the property @xmath36 . postponing precise definitions until section [ sec : observables_limits ] , we now state our first main result ( see theorem [ thm : mainconv_int_1 ] ) : suppose that @xmath0 is a sequence of discrete domains of mesh size @xmath37 approximating ( in the sense of carathodory ) a continuous finitely connected domain @xmath8 , and that @xmath38 tends to @xmath39 as @xmath40 . then there is a sequence of normalizing factors @xmath41 such that @xmath42 uniformly on compact subsets of @xmath8 . this convergence also holds true up to the `` nice '' parts of the boundary ; moreover , considering ratios of observables corresponding to different @xmath43 s , one can get rid of normalization issues . we work this out in theorem [ thm : mainconv ] . a striking feature of our new observables is their direct relation to spin correlations . let @xmath0 be a _ simply connected _ domain with @xmath44 punctures , that is , @xmath44 single faces @xmath45 removed , and let @xmath46 be the cover that branches around each of these punctures . then , it turns out that @xmath47 , @xmath48 , is ( up to a fixed complex factor , see proposition [ prop : obs_corr ] ) equal to @xmath49,\ ] ] where @xmath50 and @xmath51 stand for the partition function and the expectation for the ising model with dobrushin boundary conditions : `` @xmath52 '' on the @xmath53 boundary arc and `` @xmath54 '' on @xmath55 . this , together with convergence results for the observables , gives the following corollary ( with the notation `` @xmath56 '' referring to `` @xmath54 '' boundary conditions everywhere on @xmath57 ) : [ cor : intratio2 ] let @xmath58 approximate @xmath59 as @xmath60 . then @xmath61 } { { { \mathbb e}}_+[\,\sigma(z_1^{\delta})\dots\sigma(z_m^{\delta})\,]}\ \mathop{\rightarrow}\limits_{{\delta}\to 0}\ \vartheta(\phi(z_1),\dots,\phi(z_m)),\ ] ] where @xmath62 are explicit functions and @xmath63 is a conformal map from @xmath8 onto the upper half - plane @xmath64 sending @xmath13 to @xmath65 and @xmath66 to @xmath67 . in section [ sec : explicit_half_plane ] we give explicit formulae for @xmath68 in @xmath64 , and hence , by conformal invariance , for all simply connected domains . for example , @xmath69\,,\ ] ] where @xmath70 stands for the harmonic measure of the arc @xmath71 in @xmath8 as viewed from @xmath28 . these formulae for @xmath72 were previously conjectured by means of conformal field theory , see @xcite and earlier papers . to the best of our knowledge , the explicit formulae for @xmath73 are new . corollary [ cor : intratio2 ] admits a number of generalizations . let @xmath0 approximate a finitely connected domain @xmath8 with @xmath74 inner boundary components @xmath75 ( possibly macroscopic ) . then , for any @xmath76 , one has @xmath77 } { { { \mathbb e}}_+[\,\sigma(\gamma_1^{\delta})\dots\sigma(\gamma_m^{\delta})\,]}\ \mathop{\rightarrow}\limits_{{\delta}\to 0}\ \vartheta_{ab}^{\omega}(\gamma_1,\dots,\gamma_m)\,,\ ] ] where the functions @xmath78 are conformally invariant , the expectations @xmath51 , @xmath56 are taken for the ising model with dobrushin ( respectively , `` + '' ) boundary conditions on the outer boundary component and monochromatic on inner components @xmath79 , meaning that we constrain the spins to be the same along each component , but do not specify a priory whether it is plus of minus . in this case , @xmath80 denotes the ( random ) spin of the component @xmath79 . further , closely following the route proposed by hongler in @xcite , we prove a pfaffian formula which generalizes ( [ intratio1 ] ) to the case of @xmath81 boundary change operators ( in other words , `` @xmath82 '' boundary conditions with @xmath81 marked boundary points , see section [ sec : pfaffian ] ) . for @xmath72 this pfaffian formula ( along with the expressions for @xmath68 ) was previously derived by means of conformal field theory @xcite , whereas we give it a rigorous proof for general @xmath44 both in discrete , and , thanks to the convergence theorem , in continuous settings . another application of our new observables @xcite is the proof of convergence of ( multiple ) ising interfaces to sle curves in multiply connected domains . in that context , a proper choice of the observable @xmath83 ( i.e. , the corresponding double cover @xmath43 ) guarantees its martingale property with respect to the growing interface . to prove that property , it is important to relate the values of @xmath83 to the partitions function of the model with relevant boundary conditions . in section [ sec : pfaffian ] , we show how to do it in the most general case , see proposition [ prop : obsz ] . the simplest example of an sle process treated in this way ( for which the use of a non - trivial double cover is essential ) is a radial ising interface converging to radial sle@xmath84 . for simplicity , in the present paper we work on the square grid , but all our proofs remain valid for the self - dual ising model defined on isoradial graphs ( e.g. , see @xcite ) . we refer the reader interested in a detailed presentation of the basic notions of discrete complex analysis on those graphs to the paper @xcite and the reader interested in the history of the ising model to the paper @xcite and references therein . in section [ sec : notations ] , we fix the notations and conventions regarding discrete domains and the ising model . in section [ sec : observables_limits ] , we give the definition of the spinor observable and discuss its properties ( in particular , discrete holomorphicity and boundary conditions ) , as well as the connections to spin correlations . we then define the continuous counterparts of the observables and briefly discuss their properties . section [ sec : proof ] is devoted to the proof of main convergence results for spinor observables : theorem [ thm : mainconv_int_1 ] ( convergence in the bulk ) and theorem [ thm : mainconv ] ( convergence on the boundary ) . we generalize our results to the case of multiple boundary change operators in section [ sec : pfaffian ] . finally , in section [ sec : explicit_half_plane ] we give explicit formulae for the continuous observables @xmath85 in the punctured half - plane and for the scaling limits @xmath86 appearing in corollary [ cor : intratio2 ] . we would like to thank stanislav smirnov who involved us into the subject of the critical planar ising model for many fruitful discussions . we are also grateful to clment hongler for many helpful comments and remarks . some parts of this paper were written at the ihs , bures - sur - yvette , and the crm , bellaterra . the authors are grateful to these research centers for hospitality . this research was supported by the chebyshev laboratory ( department of mathematics and mechanics , saint - petersburg state university ) under the russian federation government grant 11.g34.31.0026 . the first author was partly funded by p.delignes 2004 balzan prize in mathematics ( research scholarship in 20092011 ) and by the grant . the second author was supported by the european research council ag confra and the swiss national science foundation . [ sec : notations ] by _ ( bounded ) discrete domain _ ( of mesh @xmath37 ) @xmath0 we mean a ( bounded ) _ connected _ subset of the square lattice @xmath87 ( an example of a discrete domain is given on fig . [ fig : grid ] ) . more precisely , a discrete domain is specified by three sets @xmath88 ( vertices ) , @xmath2 ( faces ) and @xmath89 ( interior edges and boundary half - edges , respectively ) , with the following requirements : * all four edges and vertices incident to any face @xmath90 belong to @xmath91 ; * every vertex in @xmath92 is incident to four edges or half - edges in @xmath91 ; * every vertex that is incident to at least one edge or half - edge @xmath93 belongs to @xmath92 ; * at least one of two faces incident to any edge @xmath94 belongs to @xmath2 . for an interior edge @xmath95 we denote by @xmath96 its midpoint . for a boundary half - edge @xmath97 we denote by @xmath96 its endpoint which is not a vertex of @xmath0 . when no confusion arises we will identify an edge ( or half - edge ) @xmath15 with a point @xmath96 . by the _ boundary _ @xmath98 of @xmath0 we will mean the set of all its boundary half - edges @xmath99 or , if no confusion arises , the set of corresponding endpoints @xmath96 . double cover _ of a discrete domain @xmath0 is a graph @xmath26 with a two - to - one local graph isomorphism @xmath100 . given a marked boundary half - edge @xmath101 , one can describe points @xmath28 on a double cover by lattice paths @xmath12 running from @xmath13 to @xmath28 in @xmath0 , modulo homotopy and modulo an appropriate subgroup of the fundamental group . if @xmath0 is @xmath102-connected , that is , has @xmath74 holes , then there are @xmath103 double covers , including the trivial one . namely , to define a cover , for each hole one has to specify whether a loop surrounding this hole lifts to a loop in the double - cover , or to a path connecting points on different sheets . in the latter case we will say that @xmath46 _ branches _ around that hole . if @xmath28 is a point on a double cover @xmath26 , we let @xmath104 be defined by @xmath105 and @xmath106 . we will also use the obvious notation @xmath107 , @xmath108 etc . we will work with the low - temperature contour representation of the critical ising model in @xmath0 ( see @xcite ) . we call a subset @xmath4 of edges and half - edges in @xmath0 ( see fig . [ fig : grid ] , note that we admit inner half - edges in @xmath4 ) a _ generalized configuration _ or a _ generalized interfaces picture _ for this model , if * each vertex in @xmath0 is incident to @xmath109 or @xmath110 edges and half - edges in @xmath4 ; * if an edge @xmath111 consists of two halves @xmath112 , then at most one of those three belongs to @xmath4 . we will denote the set of all generalized configurations in @xmath0 by @xmath113 . by the _ boundary _ @xmath114 of @xmath115 we will mean the set of all half - edges @xmath116 or corresponding points @xmath96 , if no confusion arises . the _ partition function _ of the critical ising model is given by @xmath117 ( the value @xmath118 will be fixed throughout the paper ) . here and below @xmath119 is the total number of edges and half - edges in @xmath4 , and @xmath120 the formula ( [ zfree ] ) endows the set @xmath3 of configurations , which corresponds to free boundary conditions in the spin representation , with a probability measure , the probability of a particular configuration @xmath4 being @xmath121 . we will mostly work with subsets of @xmath3 , and restrictions of the probability measure to those subsets . thus , we denote @xmath122 where `` mod 2 '' means that if some of @xmath123 appear several times in the subscript ( it will be useful for us to allow this ) , we keep in @xmath114 only those appearing an odd number of times . in the spin representation , the subset @xmath124 corresponds to _ locally monochromatic _ boundary conditions , that is , along each boundary component the spins are required to be the same ( although they may be different on different components ) . if @xmath125 , then @xmath126 corresponds to the configurations where the spins change from `` @xmath54 '' to `` @xmath52 '' at the boundary points ( half - edges ) @xmath127 . to simplify the notation , we will write @xmath128 instead of @xmath129 when @xmath130 . one should remember that we always consider ising configurations or generalized interfaces pictures in the planar domain @xmath0 itself , even though we will define observables on double covers @xmath26 . [ sec : observables_limits ] an example of four - connected discrete domain @xmath0 and two generalized interfaces pictures @xmath131 , each decomposed into a collection of loops and a simple lattice path @xmath132 , as required in definition [ maindef ] . for a double cover @xmath26 , branching , say , around two small holes but not around the big central one , one has @xmath133 , @xmath134 , and @xmath135 as , being lifted on @xmath26 , @xmath136 and @xmath137 end on different sheets . ] in this subsection we will construct spinor observables and prove their discrete holomorphicity . these observables should be considered as natural generalizations of fermionic observables introduced by smirnov @xcite to the multiply connected setup . a discrete domain @xmath0 , its double cover @xmath138 , and a boundary half - edge @xmath139 will be fixed throughout this subsection . in order to give a consistent definition for all discrete domains , we need the following ( technical ) notation . the half - edge @xmath13 , oriented from an inner vertex to @xmath140 , can be thought of as a complex number . then we set @xmath141 for some fixed choice of the sign . note that , if @xmath13 is south - directed , then @xmath142 . given a point @xmath14 ( i.e. , the midpoint of an edge or the endpoint of a boundary half - edge @xmath143 ) and a configuration @xmath144 , we introduce the _ complex phase _ of @xmath4 with respect to @xmath28 . first , we decompose @xmath4 into a collection of loops and a path @xmath12 running from @xmath145 to @xmath146 so that there are no transversal intersections or self - intersections ( see fig . [ fig : grid ] ) . a loop in such a decomposition will be called _ non - trivial _ if it does not lift to a closed loop on the double cover ( that is , if it lifts to a path between points on different sheets ) . we denote by @xmath147 the number of non - trivial loops in @xmath4 . we also introduce a sign @xmath148 , if @xmath12 lifts to a path from @xmath13 to @xmath28 on @xmath26 , and @xmath149 , if it lifts to a path from @xmath13 to @xmath150 . [ maindef ] the complex phase of a configuration @xmath151 with respect to a point @xmath28 lying on a double cover @xmath21 is defined as @xmath152 where @xmath16 denotes the winding ( i.e. , the increment of the argument of the tangent vector ) along @xmath12 . then , we define a * spinor observable * on the double cover @xmath26 as @xmath153 \(i ) @xmath154 does not depend on the way one chooses the decomposition of a given configuration @xmath4 into loops and the path @xmath12 . the proof is elementary , and we leave it to the reader . note that it is sufficient to check that the second factor @xmath155 is independent of a decomposition , as the rest is well known ( e.g. , see ( * ? ? ? * lemma 7 ) ) . \(ii ) by definition , @xmath156 , thus we call @xmath29 a _ spinor_. \(iii ) if @xmath46 is the trivial cover , then definition [ maindef ] reproduces the original construction due to smirnov ( e.g. , see ( * ? ? ? * eq . ( 2.10 ) ) ) . we denote this observable by @xmath157 and the corresponding complex phase by @xmath158 . the most important `` discrete '' properties of the observable ( [ observable ] ) are revealed in theorem [ thm : shol ] below , which states its _ s - holomorphicity _ ( see ( * ? ? ? * section 3 ) or ( * ? ? ? * definition 3.1 ) ) and describes the boundary conditions . we introduce the following notation : given a vertex @xmath159 , we consider four nearby corners of faces incident to @xmath160 , and identify them with the points @xmath161 , @xmath162 . we denote sets of all corners of @xmath0 and its double cover @xmath26 by @xmath163 and @xmath164 , respectively . similarly to ( [ defetaa ] ) , for a corner @xmath165 we set @xmath166 ( again , the particular choice of square root signs is unimportant , so we fix it once forever for each of four possible orientations of @xmath167 ) . we denote by @xmath168 the orthogonal projection of a complex number @xmath169 onto the line @xmath170 . [ thm : shol ] for any corner @xmath171 formed by edges or half - edges @xmath172 , one has @xmath173 moreover , if @xmath174 is a boundary half - edge , then @xmath175 , i.e. , @xmath176 since our observables satisfy @xmath156 , the identities ( [ shol ] ) at two corners @xmath177 , @xmath178 such that @xmath179 are equivalent . the same is fulfilled for the boundary condition ( [ bc_0 ] ) . let @xmath160 denotes the vertex incident to both @xmath180 and @xmath181 . there exists a natural bijection @xmath182 , provided by taking `` xor '' of a generalized configuration @xmath4 with two half - edges @xmath183 and @xmath184 . the well known proof of the theorem for the trivial cover ( e.g. , see ( * ? ? ? * proposition 2.5 ) or ( * ? ? ? * lemma 45 ) ) assures that , for any @xmath185 , @xmath186 clearly , the same holds true with @xmath187 replaced by @xmath188 , unless @xmath189 changes the number of non - trivial loops @xmath147 or @xmath190 . however , it is easy to see that @xmath189 always preserves the factor @xmath191 : for instance , if there was a non - trivial loop in @xmath4 that disappeared in @xmath192 , then this loop has become a part of the path @xmath193 , leading to the simultaneous change of the sign @xmath194 . to derive the boundary condition ( [ bc_0 ] ) , it is sufficient to note that the winding of any curve @xmath12 running from @xmath13 to @xmath66 is equal to @xmath195 modulo @xmath196 . the next proposition relates the boundary values of @xmath197 to spin correlations in the ising model . for a given double cover @xmath21 of a @xmath102-connected domain @xmath0 , let us fix the enumeration of inner boundary components @xmath75 so that _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ @xmath46 branches around each of @xmath198 but not around @xmath199 . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ for simplicity , below we assume that two marked boundary points @xmath200 belong to the _ outer _ boundary of @xmath26 . we denote by @xmath50 and @xmath51 the partition function and the expectation in the ising model with `` @xmath52 '' boundary conditions on the counterclockwise boundary arc @xmath201 , `` @xmath54 '' on the complementary arc @xmath202 , and _ monochromatic _ on all inner boundary components @xmath75 . recall that these boundary conditions require the spin to be constant along each @xmath79 . we denote this ( random ) spin by @xmath203 . if some hole @xmath79 is a single face , then we do not impose any boundary condition there and @xmath203 is just a spin assigned to this face . [ prop : obs_corr ] if @xmath204 belong to the outer boundary and @xmath205 , then @xmath206\ ] ] ( the choice of sign is explained in remark [ rem : bcinversion ] ) . in particular , @xmath207 . moreover , @xmath208.\ ] ] [ rem : bcinversion ] the sign @xmath209 in ( [ obs_corr ] ) depends on particular choices of @xmath210 , @xmath211 and the sheets of @xmath200 on @xmath26 . one way to fix it is as follows : let @xmath66 be such that the counterclockwise boundary arc @xmath212 lifts to @xmath213 ( otherwise , consider @xmath214 ) . then , one can replace @xmath215 in ( [ obs_corr ] ) by @xmath216 , where @xmath217 denotes the winding of the arc @xmath218 . the second identity is clear from the definition ( [ observable ] ) , since each configuration @xmath219 contributes the same amount @xmath220 to both sides of ( [ obs_corr_a ] ) . to prove ( [ obs_corr ] ) , note that each @xmath221 contributes @xmath222 to both sides , thus we only need to check that all the signs are the same . given a configuration @xmath4 , decompose it into a path @xmath223 and a collection of loops . a loop contributes to @xmath147 ( i.e. , changes the sheet in @xmath26 ) if and only if it has an odd number of components @xmath198 inside . hence , removing all those loops from the configuration results in the same sign change @xmath224 for both sides . removing the other loops ( having an even number of those @xmath79 s inside ) does not change the signs . after removing all loops , moving @xmath12 across any of @xmath79 will change @xmath203 and the sheet on which the lifting of @xmath12 ends ( i.e. , the sign @xmath225 ) , again resulting in @xmath226 factor at both sides . if finally @xmath12 goes along the counterclockwise arc @xmath218 , then @xmath4 contributes @xmath227 to @xmath228 $ ] , while its contribution to the left - hand side is equal to @xmath229 , if @xmath12 lifted to the double cover ends at @xmath66 , and @xmath230 , if it ends at @xmath231 . in this section we introduce continuous counterparts of the discrete holomorphic spinor observables , which we will later prove to be scaling limits thereof . for a moment , let us assume that @xmath8 is a bounded finitely connected domain whose boundary components are single points @xmath232 and _ smooth _ curves @xmath233 . given a double cover @xmath234 and a point @xmath235 , we denote by @xmath236 ( or just @xmath85 for shortness ) an analytic function which does not vanish identically and satisfies the following properties : ( @xmath237 ) [ lista ] @xmath238 everywhere in @xmath35 ; [ listb ] @xmath34 is continuous up to @xmath239 except , possibly , at the single - point boundary components and at @xmath13 , and satisfies riemann boundary conditions @xmath240 where @xmath241 denotes the _ normal to @xmath8 at @xmath28 ; [ listc ] for each single - point boundary component @xmath242 the following is fulfilled : + if @xmath46 branches around @xmath243 , then there exists a _ constant @xmath244 such that @xmath245 otherwise @xmath85 is bounded near @xmath243 , and thus has a removable singularity there ; [ listd ] in a vicinity of the point @xmath13 , one has @xmath246 for a _ real _ constant @xmath247 . the properties ( ) ( ) should be thought of as natural continuous analogues of those satisfied by @xmath33 on the discrete level . namely , ( ) corresponds to the boundary condition ( [ bc_0 ] ) ; ( ) turns out to be the correct formulation of this boundary condition for microscopic holes ; and ( ) states that @xmath85 has the simplest possible behaviour near @xmath13 , which roughly resembles the fact that @xmath33 fails to satisfy ( [ bc_0 ] ) at one boundary edge only . lemma [ lemma : uniqueness_bis ] below shows that properties ( ) ( ) define the function @xmath85 uniquely up to multiplication by a real constant ; moreover , @xmath248 unless @xmath85 vanishes identically . sometimes it is convenient to fix this constant so that @xmath249 however , below we also work with non - smooth domains , for which @xmath250 is not well - defined ; therefore , we prefer to keep the multiplicative normalization of @xmath85 unfixed the boundary value problem ( ) ( ) is not easy to analyse directly . however , the following trick relates it to a much simpler dirichlet - type problem : given a spinor @xmath85 , denote @xmath251 note that the function @xmath252 is analytic in @xmath8 , so @xmath253 is locally well - defined and harmonic . [ lemma : h ] a holomorphic spinor @xmath85 satisfies the conditions ( ) ( ) if and only if @xmath253 is a single - valued harmonic function satisfying the following properties : ( @xmath254 ) @xmath253 is continuous up to @xmath255 ; moreover , @xmath256 and @xmath257 on all macroscopic inner boundary components @xmath258 and the outer boundary of @xmath8 , where @xmath259 stands for the outer normal derivative ; @xmath253 is bounded from above near single - point boundary components @xmath242 ; @xmath253 is bounded from below near the point @xmath13 . let @xmath85 be a holomorphic spinor such that ( ) ( ) are fulfilled . on the smooth boundary @xmath260 , one can reformulate the boundary condition ( ) as @xmath261 , which is equivalent to say that @xmath262 and @xmath263 , as stated by ( ) . further , ( ) is equivalent to say that @xmath264 as @xmath265 , hence ( ) holds true . in particular , @xmath253 is single - valued in @xmath8 as it is single - valued near each of @xmath243 and constant along each of macroscopic boundary components . similarly , ( ) can be rewritten as @xmath266 as @xmath267 , which is equivalent to ( ) . vice versa , if @xmath253 satisfies dirichlet boundary conditions on smooth macroscopic boundary components , then @xmath268^{1/2}$ ] is continuous up to . then , one can easily apply the similar arguments as above to deduce ( ) ( ) from ( ) ( ) . [ lemma : uniqueness_bis ] the holomorphic spinor @xmath269 with the properties ( ) ( ) above , if exists , is unique up to multiplication by a positive constant . moreover , if @xmath270 is a conformal map , then , again up to multiplicative constants , @xmath271 and @xmath272 , with @xmath273 being the pushforward of @xmath46 by @xmath63 . suppose that @xmath274 , @xmath275 both satisfy ( ) ( ) . then , one can compose a linear combination @xmath276 with non - zero coefficients , such that @xmath277 satisfies and is _ bounded _ in a neighborhood of @xmath13 . as above , define a harmonic function @xmath278 , and note that it is continous up to @xmath279 . as ( ) implies @xmath263 everywhere on macroscopic boundary components , the function @xmath280 can not attain its maximum there . also , ( ) says that @xmath280 is bounded from above in a neighborhood of each @xmath243 . then , the maximum principle gives @xmath281 and @xmath282 everywhere in @xmath8 , i.e. , @xmath274 and @xmath275 are proportional to each other . for the second claim , it is sufficient to check the properties ( ) ( ) for the right - hand side of ( [ eq : conf_inv ] ) and apply uniqueness , which we leave to the reader . the existence of a non - trivial solution @xmath85 to the above boundary value problem will follow from theorem [ thm : mainconv_int_1 ] ; we also refer the reader to @xcite for a purely analytic techniques developed for boundary problems of this kind . the conformal covariance property ( [ eq : conf_inv ] ) immediately allows one to extend the definition of @xmath283 to non - smooth of unbounded domains : if @xmath8 is a finitely connected bounded domain , such that @xmath284 consists of smooth arcs and single points , we define @xmath285 as the unique , up to a multiplicative constant , non - zero solution to the boundary value problem ( ) ( ) . otherwise we define it by ( [ eq : conf_inv ] ) , where @xmath286 is any smooth bounded domain . further , we choose a harmonic function @xmath287 so that @xmath288 on the boundary component of @xmath8 containing the point @xmath13 , thus @xmath289 is defined up to a multiplicative constant as well . an _ equivalent _ definition of @xmath290 for non - smooth domains would be to impose the conditions ( ) ( ) given in lemma [ lemma : h ] and the condition that @xmath291^{1/2}$ ] is a spinor on @xmath292 . indeed , the only condition in in lemma [ lemma : h ] that relies on smoothness of @xmath293 is @xmath257 which can be reformulated in the following conformally invariant way : @xmath294 for our convergence results , we assume that discrete domains @xmath0 approximate @xmath8 in the carathodory topology , see @xcite or ( * ? ? ? * section 3.2 ) . the reader unfamiliar with that notion can think of the ( stronger ) hausdorff convergence . to simplify notation , we also assume that @xmath0 has the same topology as @xmath8 . the first theorem says that discrete spinors defined in section [ subsec : discr_hol_spinors ] ( with respect to a fixed double cover @xmath46 of the refining domains @xmath0 ) are uniformly close to their continuous counterparts in the bulk of @xmath8 . [ thm : mainconv_int_1 ] suppose that @xmath0 is a sequence of discrete domains of mesh size @xmath37 approximating ( in the sense of carathodory ) a continuous finitely connected domain @xmath8 , and that @xmath295 tends to some @xmath296 which is not a single - point boundary component . then , there exists a sequence of normalizing factors @xmath41 such that @xmath297 uniformly on compact subsets of @xmath8 . see section [ sec : proof ] . when extending this convergence to the boundary , we will impose additional regularity assumptions : [ def : reg_conv ] we say that a sequence of discrete domains @xmath0 with marked boundary points @xmath298 approximating a planar domain @xmath8 with a marked boundary point @xmath66 is _ regular _ at @xmath66 , if * near @xmath66 , the boundary @xmath299 locally coincide with a horizontal or vertical line ; * there exist @xmath300 , such that , for any @xmath37 , @xmath301 contains a discrete rectangle @xmath302 shifted and rotated so that @xmath298 is the midpoint of its boundary side , and @xmath98 coincides with that side in the @xmath194-neighborhood of @xmath298 . in fact , all our results can be directly extended to the case of a straight , but not necessarily vertical or horizontal boundary ( cf . * theorem 5.6 ) ) . some additional technicalities are required to prove this result in the full generality : note that @xmath303 is not even continuous or bounded on the non - smooth boundary , so one is forced to work with ratios , as , e.g. , in theorem [ thm : mainconv ] below . [ thm : mainconv ] let @xmath304 be two double covers of a bounded finitely connected domain @xmath8 with two marked points @xmath200 on the outer boundary component , and let @xmath0 converge to @xmath8 in the carathodory sense , @xmath305 be boundary points converging to @xmath200 as @xmath40 , and this convergence is regular at @xmath13 and @xmath66 . then , @xmath306 where both @xmath307 and @xmath308 are normalized by ( [ norm_h_a ] ) . formally , above we should have used different notation @xmath309 etc , to denote points lying on different double covers , with @xmath310 . we prefer to keep a simpler notation for the shortness . see section [ sec : proof ] . in the next corollary , let @xmath46 be a fixed double cover of @xmath8 , and @xmath311 be those inner components of @xmath299 for which @xmath46 branches around @xmath79 . denote by @xmath312 the corresponding components of @xmath313 . [ cor : ratio ] under the conditions of theorem [ thm : mainconv ] , as @xmath40 , one has @xmath314 } { { { \mathbb e}}_+[\,\sigma(\gamma^\delta_1)\dots \sigma(\gamma^\delta_m)\,]}\ \rightarrow\ \frac{f^{\omega}_{{\varpi}}(a , b ) } { f^{\omega}_0(a , b)}=:\vartheta_{ab}^{\omega}(\gamma_1,\dots,\gamma_m)\,,\ ] ] where both @xmath315 and @xmath316 are normalized by ( [ norm_h_a ] ) . the limit @xmath78 is a conformal invariant of @xmath317 . denote @xmath318 . due to proposition [ prop : obs_corr ] , one has @xmath319 } { { { \mathbb e}}_+[\sigma(\gamma ) ] } = \frac{\rz_{a^{\delta}b^{\delta}}{{\mathbb e}}_{a^{\delta}b^{\delta}}[\sigma(\gamma)]}{\rz_+{{\mathbb e}}_+[\sigma(\gamma)]}\cdot \frac{\rz_+}{\rz_{a^{\delta}b^{\delta } } } = \frac{f_{\varpi}(a^{\delta},b^{\delta})}{f_{\varpi}(a^{\delta},a^{\delta})}\cdot \frac{f_0(a^{\delta},a^{\delta})}{f_0(a^{\delta},b^{\delta})}\,.\ ] ] thus , ( [ ratio_exp_conv ] ) immediately follows from theorem [ thm : mainconv ] . the limit is conformally invariant due to the same conformal covariance property ( [ eq : conf_inv ] ) of both @xmath303 and @xmath320 ( observe also that the coefficients @xmath321 for both @xmath85 and @xmath322 change by the same factor @xmath323 when applying ( [ eq : conf_inv ] ) ) . [ rem : ratio_ab_ad ] ( i ) corollary [ cor : ratio_many ] gives a generalization of this result for the case of @xmath324 marked boundary points and `` @xmath82 '' boundary conditions . \(ii ) let a third point @xmath325 be marked on the outer boundary of @xmath8 and the convergence of @xmath326 to @xmath8 is regular at @xmath325 as well . then , @xmath327 and this limit is a conformal covariant of the multiply connected domain @xmath328 ( namely , it is multiplied by the factor @xmath329 when applying a conformal map @xmath63 ) . for simply connected @xmath8 s , this is given by ( * ? ? ? * corollary 5.7 ) , and we give a proof for multiply connected domains in the end of section [ sec : proof ] . in this section we prove theorems [ thm : mainconv_int_1 ] and [ thm : mainconv ] following the scheme developed in @xcite for simply connected domains . first of all , in order to transform the boundary conditions ( [ bc_0 ] ) to the dirichlet ones , we consider a discrete integral @xmath330 . further , we observe that , under a proper normalization , the functions @xmath331 and @xmath332 have non - trivial subsequential limits on compact subsets of @xmath8 as @xmath333 . we then show that any such subsequential limit is a solution to the boundary value problem ( ) ( ) , and lemma [ lemma : uniqueness_bis ] guarantees that all those limits are the same , concluding the proof of theorem [ thm : mainconv_int_1 ] . finally , we treat the behaviour of @xmath332 near the boundary points @xmath66 and @xmath13 to prove theorem [ thm : mainconv ] . for technical purposes , we extend our domain slightly : denote by @xmath334 and @xmath335 the subsets of faces and vertices that are adjacent but do not belong to @xmath2 and @xmath92 , respectively . more precisely , @xmath335 can be identified with the set @xmath99 of boundary half - edges and should be formally considered as a set of pairs @xmath336 ( e.g. , see ( * ? ? ? * section 2.1 ) ) , and @xmath334 should be treated in the same way . we set @xmath337 we work with the discrete spinor @xmath338 defined on a double cover @xmath26 of a discrete multiply connected domain @xmath0 , which we do not include in the notation unless needed . recall that it is s - holomorphic ( i.e. , satisfies ( [ shol ] ) ) and obeys the boundary conditions ( [ bc_0 ] ) at all boundary half - edges @xmath15 , except for one edge @xmath13 on the boundary . we denote the corresponding vertex of @xmath335 by @xmath339 . observe also that @xmath32 is not identically zero , since the positivity of spin correlations and proposition [ prop : obs_corr ] yield @xmath340 . these are the only properties of @xmath341 that we will use in this section . recall that , in the continuous case , it was proved to be useful to transform the boundary value problem ( ) ( ) into a dirichlet - type one by integrating @xmath342 . the extension of this construction to the discrete setting is delicate , because the square of discrete analytic function need not be discrete analytic . fortunately , the tools to treate this issue have already been developed in @xcite . proposition [ prop : propertiesh ] below summarizes these tools . namely , properties ( 1)(3 ) thereof show that one can define the discrete analog of @xmath343 as a pair of functions @xmath344 and @xmath345 , one of which is subharmonic and another superharmonic ; properties ( 4)(6 ) handle the boundary conditions , and properties ( 7),(8 ) show that these two functions can not be too far from each other . this allows one to work with a pair @xmath346 as if it was a single harmonic function . essentially , our analysis is based on a priori bounds for @xmath347 derived from simple harmonic measure estimates combined with the uniqueness of solution to the boundary value problem ( ) ( ) . we define two functions @xmath344 and @xmath345 on @xmath348 and @xmath349 , respectively , by the following rule : if @xmath350 and @xmath93 , @xmath351 , @xmath352 are all incident to @xmath177 , then @xmath353 thanks to the square , this definition does not depend on the choice of the sheet , and thanks to the basic definition ( [ shol ] ) of s - holomorphicity , it does not depend on the choice of @xmath15 between the two edges ( or boundary half - edges ) adjacent to @xmath177 . [ prop : propertiesh ] the functions @xmath344 and @xmath345 obey the following properties : ( ) - they are well - defined up to an additive constant ; [ h : welldef ] [ defh ] if @xmath93 is incident to @xmath354 and @xmath355 , then : @xmath356 ; \\ h_\circ(f)-h_\circ(f')&={\mathrm{im } } [ ( f({\varpi}^{-1}(e))^2(f - f')];\end{aligned}\ ] ] [ h : subharm ] @xmath357 and @xmath358 for all @xmath359 , @xmath360 , where @xmath361 stands for the discrete laplacian @xmath362 [ h : normal ] if @xmath363 , then @xmath364 ; [ h : boundary ] @xmath365 along each component of @xmath334 ( then we fix an additive constant in the definition of @xmath347 so that @xmath366 on the boundary component that contains @xmath13 ) ; [ h : modify ] one can modify the discrete laplacian at all vertices @xmath367 incident to @xmath363 and set values of @xmath368 on @xmath369 to be equal to the corresponding @xmath370 , so that ( [ h : subharm ] ) and ( [ h : normal ] ) will still hold ( moreover , this modification do not destroy any estimates or convergence results for discrete harmonic functions defined on @xmath92 ) ; [ h : harn1 ] if @xmath371 and @xmath372 are adjacent to some inner face @xmath90 and @xmath373 , then , for all @xmath374 , @xmath375 with some universal constant ; [ h : harn2 ] if @xmath372 and @xmath376 are adjacent to some inner vertex @xmath359 and @xmath377 , then , for all @xmath374 , @xmath378 with some universal constant . all these properties are known in the simply connected case ( e.g. , see ( * ? ? ? * section 3 ) ) . since ( [ defh ] ) , ( [ h : subharm ] ) , ( [ h : harn1 ] ) and ( [ h : harn2 ] ) are local consequences of s - holomorphicity ( [ shol ] ) , they extend immediately to the multiply connected setup . properties ( [ h : normal ] ) and ( [ h : boundary ] ) follow from the boundary condition ( [ bc_0 ] ) using ( [ defh ] ) . the property ( [ h : modify ] ) is also a local property of s - holomorphic function satisfying the boundary condition ( [ bc_0 ] ) , see ( * ? ? ? * section 3.6 ) or ( * ? ? ? * proof of proposition 8) . so we only need to check ( [ h : welldef ] ) , i.e. that the summation of ( [ defh_basic ] ) along any loop gives zero . for homotopically trivial loops this follows from the local consistency of the definition ( [ defh_basic ] ) exactly as in the simply connected case , and extends to loops running around holes by ( [ h : boundary ] ) . [ remark : homogeneous ] note that we have an immediate corollary of properties ( [ h : welldef])([h : normal ] ) : if @xmath379 is an s - holomorphic spinor satisfying the boundary condition ( [ bc_0 ] ) _ everywhere _ on @xmath57 including the point @xmath13 , then @xmath380 . indeed , ( [ h : subharm ] ) implies that the corresponding function @xmath344 attains its maximal value on the boundary , and then @xmath381 due to the property ( [ h : normal ] ) which now holds true everywhere on @xmath57 . moreover , similar arguments show that a solution to the discrete boundary value problem ( [ shol ] ) , ( [ bc_0 ] ) is unique up to a multiplicative constant ( the proof mimics ( * ? ? ? * remark 5.1 ) ) . in what follows , we denote by @xmath382 the _ discrete harmonic measure _ of a set @xmath383 in the discrete domain @xmath0 viewed from a vertex @xmath160 . recall that @xmath382 is given by the probability of the event that the simple random walk starting at @xmath160 hits @xmath12 before @xmath384 . we use the same notation @xmath385 for the discrete harmonic measure of a set @xmath386 viewed from a face @xmath277 . essentially , we will use only the following elementary properties of @xmath387 , which also are fulfilled for the discrete laplacian modified near the boundary as it is mentioned in proposition [ prop : propertiesh ] , property ( [ h : modify ] ) ( see ( * ? ? ? * section 3.6 ) or ( * ? ? ? * proof of proposition 8) ) : * _ weak beurling - type estimate : _ there exist absolute ( i.e. , independent of @xmath9 , @xmath0 and @xmath12 ) constants @xmath388 such that the following estimate holds true : @xmath389^{\!p},\ ] ] where @xmath390 means the smallest @xmath391 such that @xmath28 and @xmath392 are connected inside @xmath393 . * _ uniform estimates for the exit probabilities in rectangles : _ if @xmath394 is a discretization of the rectangle @xmath395 with @xmath396 , @xmath397 , @xmath398 denote correspondingly the bottom and the top side of @xmath399 , then @xmath400 for all @xmath401 , where @xmath402 are some absolute constants . note that ( [ beurling ] ) and ( [ hmrectangles ] ) hold true for arbitrary isoradial graphs ( see @xcite ) . now we proceed to the definition of normalizing factors @xmath403 used in theorem [ thm : mainconv_int_1 ] . let a sequence of discrete domains @xmath0 approximate a finitely connected planar domain @xmath8 , whose boundary @xmath293 consists of single - point inner components @xmath404 and continua @xmath405 , where _ @xmath406 denotes the outer boundary of @xmath8_. let @xmath407 be chosen sufficiently small so that , for any @xmath408 , @xmath409\]]has the same topological structure as @xmath8 , where @xmath410 denotes the proper connected component of @xmath411 . further , let @xmath412 be chosen small enough so that , for any @xmath413 , one has @xmath414 and @xmath415 for all @xmath416 . let @xmath417 be the spinor observable ( [ observable ] ) in @xmath0 and @xmath418 be the corresponding discrete integrals @xmath419 defined by ( [ defh_basic ] ) and proposition [ prop : propertiesh ] . we introduce normalizing factors @xmath420 by @xmath421 where , similarly to the continuous setup , @xmath422 $ ] and @xmath423 stand for discrete @xmath424-neighborhoods of @xmath243 and @xmath13 in @xmath326 . \(i ) note that the right - hand side of ( [ normdef ] ) does not vanish . indeed , if it did , @xmath425 would also vanish identically . [ rem : hnonzero ] \(ii ) since the limiting function @xmath85 has singularities at @xmath13 and @xmath243 , it is natural to cut them off when taking a maximum to get a correct normalizing factor . this choice of @xmath426 guarantees that @xmath427 in @xmath428 . our next goal is to show that , for a fixed @xmath424 , the functions @xmath429 are uniformly bounded on _ any _ compact subset of @xmath8 . this is achieved by the following lemma : ( a ) the proof of the estimate @xmath430 in lemma [ lemma : comparenorm ] . if @xmath431 is big at @xmath277 , it is also big at vertices near some path @xmath432 . since the harmonic measure of @xmath12 seen from @xmath160 is uniformly bounded from below , @xmath433 has to be big as well . ( b ) similar considerations near @xmath13 give the estimate @xmath434 . ] [ lemma : comparenorm ] for any @xmath435 , the ratio @xmath436 is bounded uniformly in @xmath437 . denote @xmath438 and @xmath439by definition , @xmath440 , hence @xmath441 . we first show that @xmath442 is uniformly bounded . recall that @xmath443 is subharmonic , @xmath444 is superharmonic , @xmath445 for adjacent @xmath160 and @xmath446 , and @xmath447 on the boundary component containing @xmath13 . therefore , @xmath448 note that in fact @xmath449 indeed , the function @xmath443 is subharmonic in @xmath450 , @xmath451 , and it can not attain its maximum on the boundary component @xmath452 because of property ( 4 ) of proposition [ prop : propertiesh ] ( and , if @xmath453 consists of one face only , then @xmath443 is subharmonic _ everywhere _ in @xmath454 ) . thus , either @xmath455 and then there is nothing to prove , or @xmath456 since @xmath444 is superharmonic , in this case there exists a path of consecutive neighbors @xmath457 such that @xmath458 . this path can only end up at @xmath312 , where the superharmonicity of @xmath444 fails ( see fig . [ fig : paths]a ) . denote by @xmath459 the set of _ vertices _ adjacent to @xmath12 . then , the property ( 8) in proposition [ prop : propertiesh ] ensures that @xmath460 where we have used the estimate @xmath461 which was explained after ( [ eq : maxmax ] ) . hence , for some absolute constant @xmath462 , @xmath463 further , there exists a constant @xmath464 independent of @xmath424 and @xmath9 such that , for any vertex @xmath465 , one has @xmath466 ( see fig . [ fig : paths]a ) . by subharmonicity of @xmath443 , this implies @xmath467 since @xmath468 , we infer that @xmath469 . it remains to prove that @xmath470 is uniformly bounded , which is done by exactly the same argument with the roles of @xmath443 and @xmath444 interchanged , inequalities reversed , and @xmath13 playing the role of @xmath243 . note that the uniform lower bound for the harmonic measure @xmath471 still holds ( this time , @xmath459 will be the set of _ faces _ adjacent to a _ vertex _ path @xmath12 which terminates at @xmath13 , see fig . [ fig : paths]b ) , provided that @xmath277 is chosen , say , to be the closest face to some fixed @xmath472 ( e.g. , see ( * ? ? ? * lemma 3.14 ) ) . now we are able to claim precompactness of the families @xmath473 as @xmath40 . [ lemma : precompactness ] fix some sufficiently small @xmath391 . then , there exists a subsequence such that the functions @xmath474 converge to a harmonic function uniformly on compact subsets of @xmath8 . moreover , the functions @xmath475 converge to @xmath476 uniformly on compact subsets of @xmath8 . definition of @xmath426 and lemma [ lemma : comparenorm ] guarantee that @xmath474 are uniformly bounded in @xmath477 for any @xmath478 , and hence on all compact subsets of @xmath8 . due to * theorem 3.12 ) , the functions @xmath474 and the spinors @xmath479 are thus equicontinuous on compact subsets of @xmath0 , and the arzela - ascoli theorem implies their subsequential convergence to a continuous function @xmath347 and a spinor @xmath379 , respectively . morera s theorem , together with the discrete holomorphicity of @xmath480 , implies that @xmath379 is holomorphic . since increments of @xmath474 are given by discrete integrals @xmath481 $ ] , one has @xmath482 , so @xmath347 is a harmonic function defined in @xmath8 . the next step is to show that all these subsequential limits solve the correct boundary value problem . it is convenient to work with the function @xmath347 . recall that , due to proposition [ prop : propertiesh ] ( properties ( [ h : boundary ] ) and ( [ h : modify ] ) ) , @xmath483 on each of macroscopic boundary components @xmath484 . by definition ( [ normdef ] ) of the normalizing factors @xmath426 , we have @xmath485 . thus , taking a subsequence once more , we can assume that @xmath486 for some constants @xmath244 such that @xmath487 . [ lemma : identlimits ] for any subsequential limit @xmath347 from lemma [ lemma : precompactness ] satisfying ( [ betacjmesh->cj ] ) , the conditions ( ) ( ) from lemma [ lemma : h ] hold true . moreover , @xmath488 . below we use the equivalent reformulation ( [ eq : d_n_h_le_0 ] ) of the boundary condition @xmath489 which does not rely on the smoothness of @xmath299 . _ property ( ) _ . our first goal is to prove that @xmath347 satisfies the dirichlet boundary conditions on all macroscopic boundary components @xmath79 , @xmath490 . fix some small @xmath491 , and recall that @xmath492 everywhere in @xmath493 , where @xmath494 does not depend on @xmath9 . due to superharmonicity of @xmath444 , for any face @xmath495 , one has @xmath496 similarly , subharmonicity of @xmath443 implies that , for any vertex @xmath497 , one has @xmath498 now let @xmath277 and @xmath160 approximate some point @xmath499 as @xmath40 . since both discrete harmonic measures converge to the continuous one , and for incident @xmath277 and @xmath160 , ( [ unifh ] ) and ( [ unifh1 ] ) yield @xmath500 in particular , @xmath501 as @xmath28 tends to @xmath79 . the boundary condition ( [ h : normal ] ) from proposition [ prop : propertiesh ] survives in the limit and gives ( [ eq : d_n_h_le_0 ] ) due to ( * ? ? ? * remark 6.3 ) . _ properties ( ) and _ by subharmonicity of @xmath443 , the inequality @xmath502 extends from @xmath503 to the whole discrete disc @xmath504 , thus @xmath505 in a vicinity of @xmath243 . similarly , superharmonicity of @xmath506 and the condition @xmath507 on the boundary component containing @xmath13 imply @xmath508 near @xmath13 . _ normalization @xmath488 . _ recall that , similarly to ( [ tbeta = max(max , min ) ] ) , @xmath509 attains its maximum @xmath510 on the boundary @xmath511 note that ( [ betacjmesh->cj ] ) and the property ( ) yield @xmath512 moreover , as @xmath474 uniformly converge to @xmath347 on all compact subsets of @xmath8 , one has @xmath513 finally , the convergence of @xmath429 to @xmath347 inside of @xmath8 also imply @xmath514 since the estimates ( [ unifh]),([unifh1 ] ) guarantee that @xmath509 are uniformly small near the boundary component of @xmath8 containing @xmath13 . we set @xmath515 for a small _ fixed _ @xmath391 . by lemma [ lemma : precompactness ] , the functions @xmath516 and @xmath517 have subsequential limits @xmath347 and @xmath379 , respectively . lemmas [ lemma : identlimits ] , [ lemma : h ] and [ lemma : uniqueness_bis ] guarantee that all these possible limits @xmath347 are the same . moreover , this unique limit is nontrivial due to the normalization condition in lemma [ lemma : identlimits ] , and thus coincides with @xmath289 normalized so that @xmath518 . since @xmath519 , we conclude that @xmath520 . the next lemma shows that the convergence of @xmath517 holds true at the boundary point @xmath66 ( in fact , it follows from our proof that this convergence is uniform up to straight parts of the boundary @xmath299 ) . note that a similar result has been obtained in ( * ? ? ? * theorem 5.6 ) under milder assumptions . for convenience of the reader , we give a shorter proof here using our ( stronger ) regularity assumptions for @xmath0 near @xmath66 . [ lemma : onboundary ] suppose that , under the conditions of theorem [ thm : mainconv_int_1 ] , we are given a sequence of marked points @xmath521 , @xmath522 , and the convergence @xmath523 is regular at @xmath66 . then , @xmath524 as @xmath40 . below we assume that @xmath8 and @xmath0 are shifted and rotated so that @xmath525 , @xmath0 contains a discrete rectangle @xmath526 for some fixed @xmath300 , and @xmath98 locally coincides with the boundary @xmath527 of the discrete upper half - plane @xmath528 ( see definition [ def : reg_conv ] ) . we also assume that the additive normalization of the functions @xmath331 is chosen so that they vanish on the macroscopic boundary component containing @xmath66 . recall that , on compact subsets of @xmath8 , @xmath517 and @xmath516 converge to the properly normalized functions @xmath85 and @xmath253 , respectively . moreover , the uniform estimates ( [ unifh]),([unifh1 ] ) guarantee that the convergence @xmath529 remains true up to @xmath530 . in particular , the functions @xmath516 converge to @xmath253 uniformly in the fixed rectangle @xmath531 around @xmath66 . let @xmath532 and @xmath533 ( note that @xmath534 can be defined by ( [ defh_basic ] ) starting with the constant s - holomorphic function @xmath535 which satisfies the boundary conditions ( [ bc_0 ] ) on @xmath527 ) . then , for any @xmath536 , one can find a small @xmath537 such that @xmath538 for all sufficiently small @xmath9 s . since both @xmath539 and @xmath540 are discrete harmonic and satisfy the same dirichlet boundary conditions as @xmath331 near @xmath298 , the sub- and super - harmonicity of @xmath443 and @xmath444 implies that @xmath541 let @xmath542 be the inner vertex of the boundary half - edge @xmath66 ( see notation on fig . [ fig : proof]a ) . estimating the discrete harmonic measure @xmath543 by ( [ hmrectangles ] ) , one obtains @xmath544 since @xmath545 , this gives @xmath546 note that this bound also holds true for the value @xmath547 , where @xmath548 denotes the neighboring boundary edge and @xmath549 is the corresponding inner vertex ( see notation on fig . [ fig : proof]a ) . now let @xmath550 be the midpoint of the edge @xmath551 , and @xmath552 be an inner face incident to both @xmath553 and @xmath554 . using ( [ h_vs_h_circ_bullet ] ) and estimating @xmath555 by ( [ hmrectangles ] ) , one obtains @xmath556 note that for any complex number @xmath557 one has @xmath558^{-1}\cdot(\|{\mathbf{pr}}_{e^{{-i\pi}/{8}}}(\xi)\|^2+\|{\mathbf{pr}}_{e^{{i\pi}/{8}}}(w)\|^2).\ ] ] therefore , we arrive at the inequalities @xmath559^{-1}\cdot(\|{\mathbf{pr}}_{e^{{-i\pi}/{8}}}(\beta({\delta})f^{\delta}(b''))\|^2 + [ 2\cos^2\!{\textstyle\frac{\pi}{8}}]^{-1}\cdot(\|{\mathbf{pr}}_{e^{{-i\pi}/{8}}}(\beta({\delta } ) f^{\delta}(b))\|^2 + { \textstyle\frac{1}{2}}(\|\beta({\delta } ) f^{\delta}(b)\|^2+\|\beta({\delta } ) f^{\delta}(b')\|^2)\le \mu+{\mathrm{const}\ , } \cdot{\varepsilon}.\end{aligned}\ ] ] since @xmath560 can be chosen arbitrary small , this yields latexmath:[$\lim_{{\delta}\to 0 } ( a ) notation near @xmath298 used in the proof of lemma [ lemma : onboundary ] . ( b ) notation near @xmath562 used in the proof of lemma [ lemma : diffcovers ] . ] finally , we work out the relation of the values of discrete observables @xmath480 at the points @xmath563 to the growth rate of their limit @xmath85 near @xmath13 . [ lemma : diffcovers ] under the conditions of theorem [ thm : mainconv_int_1 ] , assume also that the convergence @xmath523 is regular at @xmath13 . for @xmath564 , let @xmath565 be double covers of @xmath0 , @xmath566 , and @xmath567 denote the corresponding coefficients in the expansions of @xmath568 near @xmath13 ( see ( ) in the definition of @xmath568 given in sect . [ sec : cont_spinors ] ) . then , @xmath569 below we assume that @xmath8 and @xmath0 are shifted and rotated so that @xmath570 , @xmath0 contains a discrete rectangle @xmath526 , and @xmath98 locally coincides with @xmath527 . we also write @xmath571 for @xmath572 and @xmath573 for @xmath574 . denote @xmath575 recall that both @xmath576 are positive multiples of some fixed complex number @xmath577 ( see proposition [ prop : obs_corr ] ) , thus @xmath578 . taking a subsequence , we may assume that @xmath579 as @xmath40 ( if @xmath580 , then swap @xmath581 and @xmath582 and consider the inverse ratio @xmath583 ) . note that @xmath584 therefore , it is sufficient to prove that the function @xmath585 defined in @xmath586 , converges to a limit which remains bounded near @xmath587 , since this will immediately give @xmath588 for any subsequential limit . note that , being a real linear combination of @xmath581 and @xmath582 , the function @xmath480 is s - holomorphic in the discrete rectangle @xmath586 and satisfies the boundary condition ( [ bc_0 ] ) on its bottom side , including the point @xmath562 , where @xmath589 . thus , ( [ defh_basic ] ) allows one to define a discrete integral @xmath590 inside @xmath586 so that @xmath447 everywhere on the bottom side of @xmath586 . we claim that both functions @xmath591 , and hence @xmath480 and @xmath331 , are uniformly bounded on the top , left and right sides of the smaller rectangle @xmath592 ( see fig . [ fig : proof]b for the notation ) , where @xmath593 . on the top side , this follows from the uniform convergence of @xmath591 to a continuous limit . in order to prove a uniform bound on the left and the right sides , note that , for @xmath594 , the second terms in ( [ unifh ] ) and ( [ unifh1 ] ) can be uniformly estimated using ( [ hmrectangles ] ) in the following way : @xmath595 therefore , @xmath596 in neighborhoods of the left and the right sides of @xmath592 . hence , @xmath597 on those sides due to ( * ? ? ? * theorem 3.12 ) . thus , @xmath331 is uniformly bounded on the top , left and right sides of the rectangle @xmath592 and vanishes on its bottom side . using super-/sub - harmonicity of @xmath331 on faces / vertices , and uniform estimates ( [ hmrectangles ] ) , one easily deduces from here that @xmath598 everywhere in @xmath599 . applying ( * ? ? ? * theorem 3.12 ) once again , we conclude that @xmath600 in @xmath599 . hence , @xmath588 . one has @xmath601 where we have applied lemmas [ lemma : diffcovers ] and [ lemma : onboundary ] to the first and the second factors . let @xmath391 be chosen small enough . as above , it is sufficient to prove ( [ ratio_ab_ad ] ) for any subsequence @xmath602 such that @xmath603 converge to some nontrivial continuous limit . but this immediately follows from lemma [ lemma : onboundary ] applied to both boundary points @xmath604 and @xmath605 , since the properly normalized observables @xmath606 converge to @xmath607 at both @xmath66 and @xmath325 . in this section , we follow @xcite to extend the definition of the spinor observables to the case of _ multiple _ marked points on the boundary . we show that these observables ( we call them _ multi - source _ ones ) are still s - holomorphic . moreover , by analysing boundary value problems they solve , we prove _ recurrence relations _ that eventually allow one to express all these observables in terms of the basic ones introduced in section [ sec : observables_limits ] . the reason to introduce the multi - source observables is revealed in propositions [ prop : obs_corr_many ] and [ prop : obsz ] , where we establish their relation to spin correlations and partition functions , respectively . the latter is especially important in view of two applications . first , it allows one to prove the discrete martingale property of those observables with respect to interfaces growing in multiply connected domains , leading to a description of scaling limits thereof @xcite . second , in the case of @xmath81 microscopic holes carrying one boundary change operator each , the kramers - wannier duality relates the corresponding partition function to the @xmath81-points spin - spin correlations in the critical ising model with free boundary conditions . this can be used to prove the conformal covariance of their scaling limits , which would complement the results of @xcite and this paper . consider a domain @xmath0 with @xmath608 marked points @xmath609 and an inner edge @xmath28 . each configuration @xmath610 can be decomposed into a collection of ( mutually disjoint , non - self - intersecting ) loops and @xmath611 curves connecting @xmath612 s and @xmath28 in some manner . in order to define the complex phase of @xmath4 , we draw @xmath613 artificial arcs @xmath614 connecting @xmath615 to @xmath616 , @xmath617 , @xmath618 to @xmath619 , respectively , and fix the way how they lift to the double cover @xmath292 . adding these arcs to a configuration promotes it to a collection of loops and a single curve @xmath12 running from @xmath620 to @xmath621 . as we admit intersections of the artificial arcs with curves constituting @xmath4 , this time the loops can be self - intersecting , see figure [ fig : multi ] . [ def : multi_source ] we define the complex phase @xmath154 to be @xmath622 _ times _ @xmath623 , where @xmath624 is the number of loops in @xmath625 that have zero winding modulo @xmath626 , and other factors are as in definition [ maindef ] . further , we define the multi - source observable by the formulae @xmath627 [ rem : artificial_arcs ] in fact , the only data we use concerning each of the artificial arcs @xmath628 is its winding @xmath629 modulo @xmath626 and the way how it lifts to @xmath292 . the reader may check that altering the choice of @xmath630 can only result in a sign change of the observable . ( a ) a doubly connected domain @xmath0 and a configuration @xmath631 . adding artificial arcs ( dashed lines ) , we promote @xmath4 to a collection of loops and a simple path @xmath12 running from @xmath620 to @xmath621 . the loop containing @xmath632 and @xmath633 has winding @xmath67 , thus contributing to @xmath624 . ( b ) a configuration contributing to @xmath634 . it also contributes the same value ( up to a complex factor @xmath635 ) to @xmath636 . ] the following straightforward generalization of theorem [ thm : shol ] holds true : [ prop : s - hol_many ] the observables ( [ observable_mp ] ) are s - holomorphic spinors satisfying the boundary condition ( [ bc_0 ] ) everywhere on @xmath637 . the proof is essentially the same as the one of theorem [ thm : shol ] . the only difference is that now the bijection @xmath189 between the sets of configurations @xmath638 and @xmath639 can create or destroy a loop which is self - intersecting . if such a loop has winding @xmath196 modulo @xmath626 , then there is no difference with the case of a simple loop . if it has winding @xmath67 modulo @xmath626 , then its contribution to @xmath640 after it becomes a part of @xmath12 is _ minus _ that of a simple loop , which is compensated by the simultaneous change of @xmath624 by one . the boundary conditions follow in the same way as before . we now turn to the relation of the multi - source observables with spin correlations and partition functions . let @xmath324 distinct boundary points @xmath641 , be chosen on the _ outer _ boundary of @xmath26 , and let @xmath46 be a double cover that branches around boundary components @xmath198 and does not branch around the others . write also @xmath642 , with @xmath643 if @xmath644 . [ prop : obs_corr_many ] we have @xmath645,\ ] ] where the subscripts refer to the model with boundary change operators at all the marked points @xmath646 . the proof follows from definition of @xmath647 similarly to ( [ obs_corr ] ) , so we leave it to the reader . [ rem : signs ] let us describe one way to fix the sign in ( [ obs_corr_many ] ) . suppose that , starting at @xmath13 and tracing the outer boundary of @xmath26 in the counterclockwise direction , we find the marked points @xmath646 in that order . let the artificial arcs @xmath628 run outside the domain , following the boundary arcs @xmath648 . then , assuming that the arcs @xmath649 carry `` @xmath54 '' boundary conditions , one can replace @xmath215 in the right - hand side of ( [ obs_corr_many ] ) by @xmath650 , where @xmath217 is the winding of the counterclockwise boundary arc @xmath218 , cf . remark [ rem : bcinversion ] . for the next proposition , we allow the marked points @xmath651 to be _ anywhere _ on the boundary . denote by @xmath652 the double cover that branches around all boundary components of @xmath0 carrying an _ odd _ number of marked points and does not branch around the others . [ prop : obsz ] let @xmath0 be a discrete domain . one has @xmath653 any configuration @xmath654 contributes the same value @xmath655 to both sides of the equation , so it is sufficient to check that @xmath656 does not depend on @xmath4 . it is convenient to add one more artificial arc @xmath657 connecting @xmath66 to @xmath13 , and to think that the arcs @xmath658 are actually drawn on @xmath0 as simple , mutually non - intersecting curves , cf . remark [ rem : artificial_arcs ] . by definition , an inner boundary component has an odd number of artificial arcs @xmath628 issuing therefrom if and only if @xmath652 branches around that component . hence , cutting along , say , the right - hand side of each @xmath628 , we get a sheet of @xmath26 , and @xmath26 can be obtained from two copies of that sheet by gluing along the cuts , in such a way that crossing any cut would change the sheet . we now want to show that @xmath659 does not depend on @xmath4 . adding @xmath657 to @xmath625 , we end up with a collection of loops ; denote @xmath660 . observe that a loop has winding @xmath67 modulo @xmath196 ( and hence contributes to @xmath624 ) if and only if it has an odd number of self - intersections , and it contributes to @xmath25 if and only if it intersects an odd number of cross - cuts . so , we can write @xmath661 , where @xmath662 is the number of intersections of @xmath663 with _ other _ loops ( self - intersections contribute to both @xmath147 and @xmath624 ) . those intersections can only occur between @xmath4 and artificial arcs , and each intersection contributes at most once to the sum . further , observe that @xmath664 . just as above , @xmath665 counts the number of self - intersections of @xmath397 , and @xmath225 describes the number of intersections of @xmath12 with the other cross - cuts . therefore , @xmath666 since the exponent is the total number of intersections between the loops ( not counting self - intersections ) , it is always even . in accordance with our convention `` @xmath667 '' in ( [ defconf ] ) , definition [ def : multi_source ] also gives the values of @xmath32 at marked boundary points . denote @xmath668 . [ prop : ind_formula ] the identity @xmath669 is fulfilled for any @xmath670 . both sides of ( [ ind_formula ] ) are discrete s - holomorphic spinors ( defined on the same double cover @xmath26 ) satisfying the boundary condition ( [ bc_0 ] ) everywhere on @xmath671 except the marked points @xmath672 . moreover , for any @xmath673 , there is only one term in the sum ( [ ind_formula ] ) that fails to satisfy ( [ bc_0 ] ) at @xmath612 . however , its value at @xmath612 coincides with the left - hand side value @xmath674 . hence , these two morphic spinors are equal to each other due to remark [ remark : homogeneous ] , since their difference satisfies the boundary condition ( [ bc_0 ] ) _ everywhere _ on @xmath671 . note that configurations contributing to @xmath674 actually have @xmath81 boundary points instead of @xmath324 . thus , the right - hand side of ( [ ind_formula ] ) can be expressed in terms of the similar observables with smaller number of marked points , and , recursively , in terms of the basic observables @xmath675 . in order to do this in a convenient way , we need an additional notation . recall that , for each @xmath612 , we fix the complex number @xmath676 according to ( [ defetaa ] ) . then , we define the _ real _ antisymmetric @xmath677 matrix @xmath678 $ ] by setting , for @xmath679 , @xmath680 where we have used that @xmath681 for any curve @xmath682 running in @xmath0 . further , let @xmath683 $ ] denote the sub - matrix of @xmath684 obtained by _ removing _ rows and columns with indices @xmath685 . [ prop : obs_pfaff ] we have @xmath686\cdot f_{\varpi}(a_{k};z),\ ] ] with the sign depending on the choices made for @xmath687 and @xmath628 . in particular , the sign is `` plus '' with conventions described in remark [ rem : pfs_sign ] below . [ rem : pfs_sign ] ( i ) the sign in the left - hand side of ( [ eq : obs_pfaff ] ) depends on the choice of artificial arcs @xmath628 , while for the sum in the right - hand side it depends on the sheets of @xmath612 and the signs of @xmath688 , @xmath689 . we will assume that each @xmath628 lifts to a path from @xmath690 to @xmath691 on @xmath26 , and that the signs of @xmath688 are chosen so that @xmath692=\frac{\eta_{2s}}{i\eta_{2s-1}}\,.\ ] ] \(ii ) one could also write ( [ eq : obs_pfaff ] ) as the pfaffian of a @xmath693 matrix , obtained from @xmath684 by adding the last column with entries @xmath694 and a corresponding row . we prove the claim by induction in @xmath613 , starting with the trivial case @xmath695 and using ( [ ind_formula ] ) . for @xmath689 , let @xmath696 , if @xmath74 is odd , and @xmath697 , if @xmath74 is even . by definition , any configuration contributing to @xmath698 contains a curve running from @xmath699 to @xmath700 , appended with an artificial arc connecting @xmath700 to @xmath612 . removing that arc , and taking into account ( [ eq : def_eta ] ) , we get the following identity : @xmath701;a_{k'}),\ ] ] where @xmath702:=a\setminus\{a_{k},a_{k'}\}$ ] . thus , using the induction hypothesis and observing that @xmath703 does not depend on @xmath74 , we get ( for @xmath689 ) @xmath704\frac{f_{\varpi}(a_j;a_{k'})}{f_{\varpi}(a_{k};a_{k})}\\ & = ( -1)^{k } \sum\limits_{0\leq j\neq k , k'\leq 2n}(-1)^j \,\mathrm{pf}\,g_{\varpi}[j , k , k']\frac{f_{\varpi}(a_j;a_{k'})}{i f_{\varpi}(a_{k'};a_{k'})}\\ & = ( -1)^{k } \sum\limits_{0\leq j\neq k , k'\leq 2n}(-1)^{j+ { { \mathbf 1}}[{j > k'}]}\,\mathrm{pf}\,g_{\varpi}[j , k , k']\cdot(g_{\varpi})_{j , k'}\,.\end{aligned}\ ] ] due to the standard recursive formula for pfaffians applied to the matrix @xmath705 $ ] , this can be written as @xmath706,\quad k=1,\dots,2n.\ ] ] similarly , @xmath707;a_1 ) = \frac{i\eta_0}{\eta_1}\sum\limits_{k=3}^{2n } \frac{f_{\varpi}(a_2,a[1,2];a_k ) } { f_{\varpi}(a_{k};a_{k})}\ , { f_{\varpi}(a_{k};a_{1})}.\ ] ] applying ( [ eq : f0ak / fkk= ] ) and using @xmath708 , one arrives at @xmath709}\cdot ( g_{\varpi})_{k,1}= \mathrm{pf}\,g_{\varpi}[0].\ ] ] plugging this and ( [ eq : f0ak / fkk= ] ) into ( [ ind_formula ] ) , we obtain ( [ eq : obs_pfaff ] ) . [ cor : ratio_many ] with the notation as above and conventions of remark [ rem : signs ] , let @xmath0 approximate @xmath8 as @xmath40 , regularly at all marked points . then , @xmath710 } { { { \mathbb e}}_+[\sigma(\gamma)\,]}\,\rightarrow\ , \frac{\mathrm{pf}\,[\,\zeta_{a_ja_k}^{-1}\vartheta_{a_ja_k}^{\omega}(\gamma_1,\dots,\gamma_m)\,]\,_{0\le j < k\le 2n+1 } } { \mathrm{pf}\,[\,\zeta_{a_ja_k}^{-1}\,]\,_{0\le j < k\le 2n+1}}\,,\ ] ] where the conformal invariants @xmath711 are given by ( [ ratio_exp_conv ] ) , and @xmath712 . below we omit @xmath9 for the shortness . using the relations ( [ obs_corr_many ] ) and ( [ obs_corr_a ] ) for @xmath46 as in proposition [ prop : obs_corr_many ] and for the trivial cover , and then ( [ eq : obs_pfaff ] ) , we get @xmath713 } { { { \mathbb e}}_+[\sigma(\gamma ) ] } & = \frac{f_{{\varpi}}(a_0,a;{a_{2n+1}})\cdot f_{0}(a_{2n+1},a_{2n+1})}{f_{0}(a_0,a;{a_{2n+1}})\cdot f_{{\varpi}}(a_{2n+1},a_{2n+1})}=\cr & = \frac{\sum_{k=0}^{2n } ( -1)^{k}\,\mathrm{pf}\,g_{{\varpi}}[k]\cdot { f_{{\varpi}}(a_{k};a_{2n+1})}\cdot ( i f_{\varpi}(a_{2n+1};a_{2n+1}))^{-1}}{\sum_{k=0}^{2n } ( -1)^{k}\,\mathrm{pf}\,g_0[k]\cdot f_{0}(a_{k};a_{2n+1})\cdot ( i f_0(a_{2n+1};a_{2n+1}))^{-1 } } \cr & = \frac{\sum_{k=0}^{2n } ( -1)^{k}\,\mathrm{pf}\,g_{{\varpi}}[k]\cdot ( g_{\varpi})_{k,2n+1}}{\sum_{k=0}^{2n } ( -1)^{k}\,\mathrm{pf}\,g_0[k]\cdot ( g_0)_{k,2n+1 } } = \frac{\mathrm{pf}[\,(g_{{\varpi}})_{j , k}\,]_{0\le j < k\le 2n+1}}{\mathrm{pf}[\,(g_{0})_{j , k}\,]_{0\le j < k\le 2n+1}}\,,\end{aligned}\ ] ] where @xmath714 and @xmath715 are given by ( [ eq : def_g ] ) for corresponding double covers . this can be further rewritten as @xmath716 } { { { \mathbb e}}_+[\sigma(\gamma ) ] } = \frac{\displaystyle\mathrm{pf}\left[\frac{(g_{\varpi})_{j , k}}{(g_0)_{j , k}}\cdot \frac{(g_0)_{j , k}}{(g_0)_{0,2n+1}}\right]_{0\le j < k\le 2n+1 } } { \displaystyle\mathrm{pf}\left[\frac{(g_0)_{j , k}}{(g_0)_{0,2n+1}}\right]_{0\le j < k\le 2n+1}}\,.\ ] ] note that definition ( [ eq : def_g ] ) and convergence ( [ mainconv ] ) yield @xmath717 as @xmath40 . similarly , ( [ eq : def_g ] ) and ( [ ratio_ab_ad ] ) give @xmath718 and the factors @xmath719 cancel out in the ratio of pfaffians . \(i ) one may assume that @xmath720 is normalized by ( [ norm_h_a ] ) ; in this case one has a conformal covariance rule @xmath721 . the linearity of pfaffians implies that the normalization is not important , and that the ratio ( [ ratio_exp_conv_many ] ) is conformally invariant . \(ii ) clearly , @xmath316 does not change if one adds boundary singletons inside the domain . in particular , if @xmath722 and @xmath723 , then one has @xmath724 ( with the normalization given by ( [ norm_h_a ] ) ) . \(iii ) the formula ( [ ratio_exp_conv_many ] ) was predicted by means of conformal field theory , see ( * ? ? ? * equation ( 17 ) ) . in this section we explicitly compute the holomorphic spinors @xmath725 . by ( [ ratio_exp_conv ] ) , this immediately gives us the quantities @xmath726 and , by conformal invariance of @xmath68 s , all the limits @xmath727 } { { { \mathbb e}}_+[\,\sigma(w_1^{\delta})\dots \sigma(w_m^{\delta})\,]}= \vartheta(\phi(w_1),\dots,\phi(w_m))\,,\ ] ] for discrete domains @xmath728 approximating an arbitrary _ simply connected _ domain @xmath317 and inner _ faces _ @xmath729 tending to @xmath730 , where @xmath731 is a conformal map such that @xmath732 and @xmath733 . due to the pfaffian formula ( [ ratio_exp_conv_many ] ) , this result easily extends to the case of @xmath324 marked boundary points . it is convenient to choose @xmath734 . specializing the boundary value problem and ( [ norm_h_a ] ) to the case @xmath735 , @xmath587 and using the conformal map @xmath736 , we get the following conditions for @xmath737 : in order to find @xmath277 , we introduce an auxiliary spinor @xmath747^{\frac{1}{2}}}\,.\ ] ] note that it satisfies ( ) , ( ) and ( ) . moreover , it has real zeros at @xmath748 ; thus , ( ) , ( ) and ( ) also hold for the product @xmath749 , where @xmath750 is any function of the form @xmath751 ( and if some @xmath752 coinside , we can add higher - order poles , so that @xmath750 is always a linear combination of @xmath44 linearly independent functions ) . we are looking for parameters @xmath753 such that @xmath749 satisfy ( ) as well . denote @xmath754^{\frac{1}{2}}.\ ] ] then , the condition ( ) for @xmath749 can be restated as @xmath755={\mathrm{im}}r_k + \sum_{j=1}^m \lambda_j\cdot{\mathrm{im}}\frac{r_k}{t_j\!-\!w_k}=0\quad \text{for~all~}k=1,\dots , m\,.\ ] ] note that ( [ system ] ) is an @xmath756 linear system in @xmath753 . we argue that this system is always non - degenerate . indeed , if @xmath757 is a solution to the corresponding _ homogeneous _ system , then @xmath758\cdot f_{w_1,\dots , w_m}(z)$ ] is a spinor satisfying and such that @xmath759 as @xmath745 . but it follows from the proof of lemma [ lemma : uniqueness_bis ] that any such spinor is identically zero ( after mapping to a bounded domain , the condition at infinity yields boundedness near @xmath13 ) . thus , @xmath760 . taking into account that one has explicitly @xmath761^{\frac{1}{2}}}\,.\ ] ] and solving the linear system ( [ system ] ) , one obtains @xmath753 as ratios of certain explicit @xmath756 determinants ( depending of @xmath243 s ) , and then the ratio of spin correlations ( [ thetaw1 m ] ) is given by @xmath762\cdot \prod_{j=1}^m\frac{{\mathrm{re}}\,w_j}{\|w_j\|}\,.\ ] ] here we have used conventions of remark [ rem : bcinversion ] to determine the sheet of the double cover of @xmath763 ( that is , signs of the square roots ) : in order to get the value @xmath764 , one starts with the value @xmath765 and continuously moves the boundary point to @xmath67 along the _ counterclockwise _ boundary arc @xmath766 , thus arriving to @xmath767 . the other way to fix the sign is given by taking all @xmath243 close to the boundary arc @xmath768 : this should yield a positive correlation as that arc carries `` + '' boundary conditions . in particular , for a _ single _ point @xmath769 ( i.e. , for the case @xmath770 ) one has @xmath771 , @xmath772 , thus @xmath773 and @xmath774\,.\ ] ] since harmonic measure is conformally invariant , this implies @xmath775 } { { { \mathbb e}}_+[\,\sigma(w^{\delta})\ , ] } \to \cos\,[\pi\text{hm}_{\omega}(w,(ab))]\,,\ ] ] if @xmath776 approximate @xmath317 and faces @xmath777 tend to an inner point @xmath778 .	we introduce a new version of discrete holomorphic observables for the critical planar ising model . these observables are holomorphic spinors defined on double covers of the original multiply connected domain . we compute their scaling limits , and show their relation to the ratios of spin correlations , thus providing a rigorous proof to a number of formulae for those ratios predicted by cft arguments .
studies of the processes that lead to star formation are critical for our understanding of how galaxies form and evolve with time . in massive galaxies , star formation is triggered through a variety of processes : e.g. , compression of molecular cloud cores in spiral density waves , tidal interactions , and bar formation . on scales of a few hundred parsecs , star formation is observed where the gas density exceeds a threshold , whose value varies with the critical density for local gravitational stability ( e.g. , martin & kennicutt 2001 ) . models that rely on thermal instabilities to determine the star formation threshold predict a constant threshold column density ( schaye 2004 ) . studies of star formation ( sf ) in tidal arms of interacting galaxies offer the possibility to study sf processes at conditions that are distinctly different from those in normal galactic disks ( e.g. , given the absence of spiral density waves and bars ) . nearby , interacting systems give us the unique opportunity to study the properties of tidal features ( e.g. atomic and molecular gas content and region properties ) in detail . tidal dwarf candidates have been detected in tidal tails of interacting galaxies in a number of cases ( e.g. , see list in braine et al . 2001 ) ; but selection effects limit these to bright , and therefore fairly evolved , tidal dwarfs . the detection of sf regions around interacting systems ( ` intergalactic ' regions ) has been reported in ryan weber et al . ( 2004 ) . given the much higher incidence of interacting galaxies in the past , a significant fraction of today s dwarf galaxies may have formed out of tidal debris ( hunsberger et al . 1996 , hunter et al . 2000 , okazaki & taniguchi 2000 , duc et al . 2004 ) . also of interest in this context is the self - enrichment of these tidal debris , and their possible incidence as intervening metal line systems . the nearby m81 triplet ( consisting of m82 , m81 and ngc3077 ) is an ideal testbed to study tidal arms around interacting galaxies in detail . the map of yun et al . ( 1994 ) shows a geometrical cross section of roughly 50 kpc @xmath12 100 kpc at column densities of n(hi)@xmath111.8@xmath1210@xmath13 @xmath3 . the impressive tidal arm near ngc3077 is the subject of the present study ( total mass : @xmath14m@xmath5 , walter et al . according to various numerical simulations ( thomasson & donner 1993 , brouillet et al . 1991 , yun et al . 1993 ) this tail has been created by a recent ( @xmath8yr ) interaction with m81 . barbieri et al . ( 1974 ) detected ` a fragmentary complex of almost stellar objects ' in this region ( the ` garland ' , see also karachentsev et al . 1985 , sharina 1991 , sakai & madore 2001 ) . the presence of extended molecular gas in this tidal feature has been reported by walter & heithausen ( 1999 ) and heithausen & walter ( 2000 ) . in this paper we report on the discovery of low level star formation over roughly 24 kpc@xmath1 and present new high resolution observations of the molecular gas distribution in this feature . in section 2 we summarize our observations obtained at the 2 m telescope at kitt peak , and at the ovro millimeter interferometer . in sec . 3 we present the distribution of the regions , the luminosity function and the high resolution co maps ; sec . 4 summarizes our results . we obtained deep narrowband imaging of the region including ngc3077 and the prominent tidal arm with the 2 m telescope at kitt peak on 2000 january 210 . the seeing yielded a median image quality of 13 fwhm . the individual frames were corrected for fixed pattern noise in the standard way using iraf , aligned using foreground stars , and combined by filter subset . the h@xmath15 image presented here was constructed from an on - band image taken through filter kp1468 , central wavelength ( cwl ) of @xmath16 and a full width half maximum ( fwhm ) of 87 , and an appropriately - scaled continuum image taken through filter kp809 ( cwl = 6488 , fwhm=67 ) . observations of spectrophotometric standards were used to flux calibrate the h@xmath15 image . the h@xmath15 image contains [ nii ] @xmath17 emission . the correction is h@xmath15/([nii]+h@xmath15)=0.75 in the center of ngc3077 and increases to h@xmath15/([nii]+h@xmath15 ) = 0.56 in the extended gas , where the ionization parameter is low ( martin 1997 ) . this correction has not been measured in the tidal gas ; but if the gas phase metallicity is lower in the arm , then the contribution from [ nii ] will be smaller than in ngc3077 . the foreground galactic extinction is about 0.11 magnitudes at h@xmath15 ( schlegel et al . 1998 ; burstein & heiles 1982 ) . we observed ngc3077 in the co(1@xmath180 ) transition at 115ghz ( @xmath192.7 mm ) using the owen s valley radio observatory s millimeter array ( ovro ) in c ( baselines ranging from @xmath20 ) and l ( @xmath21 ) configurations during 18 tracks during the years 2000 - 2002 ( typical length of one track : 8 hours ) . to cover most of the co emission detected in our earlier observations using the iram 30 m single dish telescope ( walter & heithausen 1999 , heithausen & walter 2000 ) we observed 4 pointings ( primary beam size of one ovro antenna : @xmath191@xmath22 ) . the northern pointing ( which was centred on the maximum of the iram 30 m observations ) was observed for 10 tracks in 2000 , the additional three pointings towards the south were added in 2002 ( total : 8 tracks ) . data were recorded using a correlator setup resulting in velocity resolutions of 2.6 @xmath23 ( after online hanning smoothing ) with a total bandwidth of @xmath24 @xmath23 ( i.e. fully covering the profiles detected in the single dish observations ) . flux calibration was determined by observing 1328 + 307 ( 3c286 ) and neptune for approximately 20 minutes during the observing runs . these calibrators and an additional noise source were used to derive the complex bandpass corrections . the nearby calibrators b0716 + 714 , j1048 + 717 and j0841 + 708 were used as secondary amplitude and phase calibrators . the data for each array were edited , calibrated and mapped with the mma and the miriad packages . the most northern pointing has a beamsize of 7.0@xmath25 6.0@xmath26 ( natural weighting ) and an rms of @xmath1920mjybeam@xmath6 in a 2.6kms@xmath6 channel ( conversion factor : 1k=0.45jybeam@xmath6 ) . the southern pointings have slightly higher resolution ( [email protected]@xmath26 , due to the different uv coverage ) but have significantly higher noise ( @xmath1965mjybeam@xmath6 , due to the shorter integration times ) . the emission in individual channels is faint and all the detected features are narrow ( typically 35 channels ) ; therefore it is not possible to calculate integrated co maps directly from the 60 channel data cube . we thus blanked emission that was not present in consecutive channels to create a map of the integrated co emission . fig . 1 shows a b band image of the region around ngc3077 ; the contours show the distribution of neutral hydrogen ( walter et al . 2002 , here smoothed to 40@xmath26 resolution ) . 2 is a blown up h@xmath15 image of the region indicated by a box in fig . 1 and shows the newly detected regions in the tidal arm near ngc3077 . in total , we catalogued 36 regions all are associated with the prominent tidal arm near ngc3077 . for each individual region we determined the coordinates , fluxes , luminosities , diameters and column densities ( tab . 1 ) . in the following we will adopt a distance of 3.6 mpc to the the tidal arm ( leading to a scale of 17.5pc/[@xmath26 ] ) . this corresponds to a distance modulus of 27.80 mag ( values used in previous studies for m81 , ngc3077 and the surroundings : 27.76 mag : sharina 1991 , 27.93 mag : sakai & madore 2001 , 27.80 mag : freedman et al . 1994 ) . the regions are distributed over an impressive area of @xmath0kpc@xmath28=24kpc@xmath28 . most of the regions are found where the column density reaches values @xmath29@xmath3 ( in a 13@xmath26 [ @xmath19 200pc ] beam , tab . 1 ) . most of the regions are rather compact ( tab . 1 ) , we derive an average diameter for the regions ( here defined as the fwhm of the regions , after deconvolving for the point spread function of our h@xmath15 observations , fwhm : 1.3@xmath26 ) of 1.9@xmath26 ( 2.3@xmath26 before the deconvolution ) . this corresponds to a linear size of @xmath1935pc , i.e. , similar to the radius of a stromgen sphere of a massive o star . the thick lines in fig . 2 represent the distribution of molecular gas in the tidal features as traced by the co(1@xmath180 ) emission observed with the iram 30 m single dish telescope ( heithausen & walter 2000 ) . the southern region coincides with the highest density of star formation . the higher star formation rate in the southern molecular complex is also evident from the excitation conditions of the molecular gas : the co(2@xmath181)/co(1@xmath180 ) line ratio is higher in the south as compared to the regions in the north ( heithausen & walter 2000 ) . 3 shows the distribution of the molecular gas as seen by the ovro interferometer ( contours ) . the circles indicate the primary beam sizes of our four pointings . recall that the noise in the northern pointing is a factor 3 lower than in the remaining pointings ( sec . 2.2 ) . in total we catalog six regions ( a f ) - regions c , d and f are marginal detections and need confirmation by follow - up observations . the properties of these six regions as derived from the ovro observations ( centre coordinates , fluxes , adopted masses and systemic velocities ) are summarized in tab . 2 . 4 we compare the ovro ( co ) , iram 30 m ( co ) and vla ( hi ) spectra for each of the six regions . the ovro spectra ( thick full line ) are plotted in units of jybeam@xmath6 , as indicated on the left hand axis , the iram 30 m data ( dashed line ) are in units of t@xmath30 ( right hand axis ) . for comparison , the vla hi data are plotted as the thin dotted line . from this it is evident that the co velocities are in good agreement with the peaks of the spectra at the same position . it is also clear that regions c , d and f need follow up confirmation . although the signal to noise of the ovro observations is low , we can still attempt to derive a virial mass estimate for the brightest two regions ( regions a and b ) . from the intergrated co map we derive an effective ( deconvolved ) radius @xmath31 of @xmath32pc for both regions . the line width @xmath33 for both regions is also similar ( fwhm : @xmath19 7.5kms@xmath6 ) . using m@xmath34=250@xmath35[pc]@xmath36[kms@xmath6]@xmath1 we thus derive approximate virial masses of @xmath38m@xmath5 for both regions ( with an error of order 3050% ) . if we take the measured co flux ( s@xmath39 ) of these two regions and apply a galactic conversion factor to convert co luminosities to molecular gas masses we derive h@xmath40 masses of 1.3@xmath41m@xmath5 and @xmath42m@xmath5 for regions a and b , respectively ( see caption of tab . 2 for details ) . as these numbers ( virial mass compared to mass derived using a galactic conversion factor ) are in agreement within the errors , this implies that the co to conversion factor in the tidal arm appears to be similar to the galactic value . we note that a galactic conversion factor was also found to hold for the centre of ngc3077 ( meier et al . 2001 , walter et al . 2002 ) . we thus derive a total molecular gas mass recovered by the ovro observations of 2.4@xmath41m@xmath5 ( using a galactic conversion factor for all regions , tab . the difference in morphology between the co map obtained at the iram 30 m single dish telescope ( fig . 2 ) and the ovro interferometer ( fig . 3 ) is striking . it can be explained by the fact that our interferometer observations are only sensitive to compact and clumpy structures and will not recover more diffuse emission . in addition , the 30 m single dish map is not fully sampled ( see heithausen & walter 2000 for details ) . with our ovro observations we only recover about one sixth of the total flux detected by the earlier 30 m observations ( there we find a total flux of of 22 kkms@xmath6 [ 100 jykms@xmath6 ] in the area covered by ovro3 times higher than the galactic factor used here . ] ) which is entirely consistent with the picture that most of the molecular gas is more extended . only one molecular cloud complex is associated with a bright region ( complex a , region 17 ) . regions c and d are also associated with faint h@xmath15 emission , however most of the other prominent regions , in particular in the southern pointing , do not show corresponding co emission down to our sensitivity limit . one bright co complex ( b ) has no associated star formation . it is likely that new star formation will commence in this region in the future . given the ( likely ) low metallicity in the tidal arm , we don not expect extinction to play a major role in these regions . observations of the dust phase ( e.g. using spitzer ) would be needed to check if highly extincted , embedded star formation may be present in some areas of the tidal feature . we now investigate how the properties of the 36 regions in the tidal feature compare to regions found in other dwarf galaxies in the same group . the luminosity function of the regions in the tidal arm is presented in fig . 5 ( thick histogram ) . also plotted in this figure are the luminosity functions derived for other m81 group dwarf galaxies by miller & hodge ( 1994 ) . note that the decrease in the luminosity function towards low luminosities is due to the sensitivity of the observations ( down to a completeness limit of log(l[ergs@xmath6])@xmath1936.4 ) for both our observations and the observations presented in miller & hodge ( 1994 ) . at the bright end of the luminosity function , values for regions in the tidal feature do not reach values as high as found in big spiral galaxies or more massive dwarf galaxies ( e.g. , ic2574 ) . however , keeping in mind the low number statistics , the shape of the region luminosity function is similar to what is found in other low mass dwarfs such as holmberg i ( fig . 5 ) . even for our brightest regions the h@xmath15 luminosity can in principle be emitted by a single massive star ( l@xmath43ergs@xmath6 , l@xmath44ergs@xmath6 , devereux et al . the total h@xmath15 and [ nii ] flux ( luminosity ) for all the regions is : f(h@xmath45=1.88@xmath46ergs@xmath6@xmath3 ( l(h@xmath45=2.91@xmath47ergs@xmath6 ) . this corresponds to a total star formation rate ( sfr ) in the entire tidal arm of sfr @xmath48m@xmath49 yr@xmath6 , where sfr(0.1 - 100m@xmath5 stars ) = l(h@xmath15)/1.26@xmath50ergs@xmath6 [ m@xmath49 yr@xmath6 ] ( kennicutt 1998 ) . the ionizing stars are part of stellar clusters , as revealed by deep acs imaging obtained with the hst ( proposal i d 9381 , these data will be presented elsewhere ) . we present the discovery of wide - spread , low - level star formation in the tidal feature near ngc3077 ( total of 36 regions ) . the column densities exceed @xmath2@xmath3 over the @xmath1924kpc@xmath1 region where the regions are found . the total star formation rate in the tidal arm is sfr=@xmath51m@xmath49 yr@xmath6 . these properties place the complexes amongst the faintest star forming regions detected in tidal arms around galaxies so far ( cf . ryan - weber et al . the presence of a huge reservoir of molecular gas ( @xmath52m@xmath5 , heithausen & walter 2000 , using the galactic conversion factor ) as well as atomic gas ( m@xmath53m@xmath5 in the region shown in fig . 2 ) suggests that star formation may continue at the current rate for a hubble time . if some 10 percent of the gas were transformed into stars , the total luminosity of the tidal dwarf will eventually be of order @xmath54m@xmath55 ( here we simplisticly adopt a stellar mass to light ration of 1m@xmath56/l@xmath55 ) , similar to other faint dwarf galaxies in the m81 group . it is interesting to note that no clear correlation exists between the sites of on going star formation and gas column densities ( both molecular and atomic ) : regions are not always found where the or molecular column density is highest ( e.g. , fig . 3 ) . however , star formation is only seen where the column density reaches values larger than @xmath57@xmath3 ( tab . 1 ) this value is consistent with what is typically found in ( dwarf ) galaxies ( e.g. , skillman 1987 , walter & brinks 1999 , martin & kennicutt 2001 ) . also , the derived luminosity function is , within the errors , similar to other low - mass dwarf galaxies in the same group ( such as holmberg i ) . no regions are found with log(l[ergs@xmath6])@xmath1137.4 this is more than an order of magnitude fainter than what is found in more massive dwarf galaxies in the same group ( e.g. ic2574 ) and bigger spiral galaxies ( e.g. , in m51 luminosities reach values of log(l)[ergs@xmath6]=39 , e.g. , rand 1993 ) . none of the regions are coincident with the x ray sources found in the same region of the sky ( ott , martin & walter 2003 ) . it is difficult to estimate when star formation started in the tidal arm . from the fact that we see h@xmath15 emission today it is clear that star formation is occuring well after the creation of the structure some @xmath58yr ago ( i.e. , ` in situ ' formation ) . if we assume a constant star formation rate over the last @xmath59yr we estimate a total mass of newly formed stars of @xmath60m@xmath5 . this star formation is enriching the ism of the tidal tail with heavy elements : taking our derived assembled stellar mass since the creation of the feature , and a yield of 2% , we derive a metal production of 1.4@xmath1210@xmath61m@xmath5 over the last @xmath59yr . we divide this number by the total amount of gas within the @xmath194@xmath126kpc region enclosed by the 10@xmath62@xmath3 contour ( 3@xmath63m@xmath5 ) to get a metallicity of 4.7@xmath64 , or roughly [email protected] solar , well below the values measured for the gas phase in the most metal poor dwarfs . we note that the total metallicity in the tidal feature is likely higher as models for the tidal arm show that it was stripped from the outskirts of ngc3077 ( thomasson & donner 1993 , brouillet et al . 1991 , yun et al . 1993 ) . measurements of the gas phase metallicity in the tidal feature would thus clearly help to shed light on its origin and history . we conclude that wide spread low level star formation may be a common phenomenon in tidal tails with column densities that exceed @xmath2@xmath3 , leading to chemical enrichment in these tails . this type of low - level star formation in tidal debris should be detectable , if it is common , as intervening absorption in quasar spectra . the tidal system discussed here would be classified as a damped lyman alpha absorber ( dla ) with a cross section of roughly 30 kpc@xmath1 . a tidal origin for ( at least some ) dlas would naturally explain their lack of enhanced [ @xmath65fe ] abundance ratios , which suggest that star formation proceeded relatively slowly ( pettini et al . direct observations of dlas at low redshift are consistent in that they indicate associations with a wide variety of galaxy types and relatively small impact parameters ( of order 10kpc , e.g. , kanekar & chengalur 2003 ) . for comparison , the tidal system discussed here has a mean projected distance of @xmath195kpc from ngc 3077 ( @xmath1950kpc distance from m81 ) . this interesting object should be further explored via metallicity measurements and simulations that would establish the fate of the metals formed in this system . clm thanks the david and lucile packard foundation and the alfred p. sloan foudation for supporting this work . we thank bryan miller for sharing his luminosity functions with us and the anonymous referee for providing useful comments . research with the owens valley radio telescope , operated by caltech , has been supported by nsf grant ast9613717 . support of this work was also provided by a grant from the k.t . and e.l . norris foundation . freedman , w.l . , hughes , s.m . , madore , b.f . , mould , j.r . , lee , m.g . , stetson , p. , kennicutt , r.c . , turner , a. , ferrarese , l. , ford , h. , graham , j.a . , hill , r. , hoessel , j.g . , huchra , j. , illingworth , g.d . 1994 , , 427 , 628 rllcccc nr . & ra ( 2000 ) & dec ( 2000 ) & f(h@xmath66 & log(l(h@xmath45 ) & d@xmath67 & @xmath68 & hh mm ss & @xmath69 @xmath22 @xmath26 & 10@xmath70ergs@xmath6@xmath3 & erg s@xmath6 & @xmath26 & 10@xmath62@xmath3 1 & 10 03 33.73 & 68 39 37.9 & 1.98 & 36.5 & & 0.9 2 & 10 03 34.18 & 68 39 33.6 & 3.47 & 36.7 & & 0.7 3 & 10 03 35.68 & 68 42 21.9 & 7.63 & 37.1 & 1.6 & 0.9 4 & 10 03 36.35 & 68 41 19.5 & 2.00 & 36.5 & 1.9 & 1.0 5 & 10 03 39.31 & 68 41 28.0 & 8.92 & 37.1 & 4.8 & 1.2 6 & 10 03 44.56 & 68 42 27.4 & 6.52 & 37.0 & & 1.5 7 & 10 03 46.29 & 68 41 25.0 & 4.13 & 36.8 & 1.8 & 1.5 8 & 10 03 46.51 & 68 42 12.1 & 2.06 & 36.5 & & 1.8 9 & 10 03 46.68 & 68 43 09.7 & 0.49 & 35.9 & 1.3 & 0.9 10 & 10 03 46.85 & 68 41 40.2 & 1.37 & 36.3 & 2.1 & 1.5 11 & 10 03 47.24 & 68 42 12.5 & 3.45 & 36.7 & 1.8 & 1.8 12 & 10 03 51.88 & 68 42 39.8 & 6.87 & 37.0 & 3.2 & 2.1 13 & 10 03 52.10 & 68 42 31.9 & 2.25 & 36.5 & 2.9 & 2.0 14 & 10 03 52.16 & 68 42 48.6 & 0.56 & 35.9 & 1.3 & 2.0 15 & 10 03 52.89 & 68 42 52.0 & 0.79 & 36.1 & 1.7 & 1.9 16 & 10 03 55.74 & 68 42 50.4 & 1.00 & 36.2 & & 1.7 17 & 10 03 56.18 & 68 42 10.3 & 11.9 & 37.3 & 1.7 & 1.6 18 & 10 03 56.38 & 68 40 42.0 & 6.23 & 37.0 & 2.3 & 1.6 19 & 10 03 57.79 & 68 41 27.4 & 3.13 & 37.7 & 1.5 & 2.3 20 & 10 03 57.85 & 68 42 03.9 & 0.97 & 36.2 & 2.0 & 1.9 21 & 10 03 58.68 & 68 41 21.9 & 3.73 & 36.8 & & 2.5 22 & 10 03 58.72 & 68 40 41.7 & 9.55 & 37.2 & 2.7 & 1.3 23 & 10 03 58.93 & 68 43 02.6 & 1.93 & 36.5 & 2.5 & 2.5 24 & 10 03 59.34 & 68 40 57.2 & 10.9 & 37.2 & 1.8 & 1.8 25 & 10 03 59.46 & 68 41 07.0 & 2.61 & 36.6 & & 2.5 26 & 10 03 59.85 & 68 41 11.2 & 11.8 & 37.3 & 2.1 & 2.7 27 & 10 04 00.63 & 68 41 10.6 & 5.53 & 36.9 & 2.1 & 2.5 28 & 10 04 01.92 & 68 41 24.0 & 0.12 & 37.3 & 2.5 & 1.8 29 & 10 04 07.33 & 68 41 05.1 & 0.15 & 37.4 & 4.1 & 1.5 30 & 10 04 10.02 & 68 41 56.5 & 4.28 & 36.8 & 1.5 & 1.7 31 & 10 04 14.73 & 68 42 37.8 & 4.13 & 36.8 & 3.2 & 1.4 32 & 10 04 26.02 & 68 42 29.7 & 3.10 & 36.7 & 2.4 & 1.2 33 & 10 04 27.25 & 68 42 32.4 & 3.46 & 36.7 & 2.4 & 1.0 34 & 10 04 28.37 & 68 42 33.0 & 8.57 & 37.1 & & 0.9 35 & 10 04 28.54 & 68 42 38.5 & 0.12 & 37.3 & 3.3 & 1.0 36 & 10 04 29.32 & 68 42 44.6 & 3.35 & 36.7 & 2.4 & 0.9 lllccc region & ra & dec & flux & h@xmath40 mass@xmath71 & v@xmath72 & ( j2000.0 ) & ( j2000.0 ) & jy kms@xmath6 & 10@xmath61 m@xmath49 & kms@xmath6 a & 10 03 55.8 & 68 42 11.8 & 8.2 & 130 & 12.7 b & 10 03 49.7 & 68 41 57.8 & 4.7 & 75 & 10.1 @xmath73 & 10 03 55.7 & 68 40 47.5 & 1.2 & 19 & 17.9 @xmath73 & 10 03 51.8 & 68 42 42.9 & 0.15 & 2.4 & 10.1 e & 10 03 55.5 & 68 42 50.8 & 0.90 & 14 & 20.5 @xmath73 & 10 03 57.4 & 68 42 42.5 & 0.35 & 5.6 & 12.7	we report the detection of ongoing star formation in the prominent tidal arms near ngc3077 ( member of the m81 triplet ) . in total , 36 faint compact regions were identified , covering an area of @xmath0 kpc@xmath1 . most of the regions are found at column densities above @xmath2@xmath3 ( on scales of 200pc ) , well within the range of threshold columns measured in normal galaxies . the luminosity function resembles the ones derived for other low mass dwarf galaxies in the same group ; we derive a total star formation rate of 2.6@xmath4 m@xmath5yr@xmath6 in the tidal feature . we also present new high resolution imaging of the molecular gas distribution in the tidal arm using co observations obtained with the ovro interferometer . we recover about one sixth of the co flux ( or m@xmath7m@xmath5 , assuming a galactic conversion factor ) originally detected in the iram 30 m single dish observations , indicating the presence of a diffuse molecular gas component in the tidal arm . the brightest co peak in the interferometer map ( comprising half of the detected co flux ) is coincident with one of the brightest regions in the feature . assuming a constant star formation rate since the creation of the tidal feature ( presumably @xmath8years ago ) , a total mass of @xmath9 m@xmath5 has been transformed from gas into stars . over this period , the star formation in the tidal arm has resulted in an additional enrichment of @[email protected] . the reservoir of atomic and molecular gas in the tidal arm is @xmath8 m@xmath5 , allowing star formation to continue at its present rate for a hubble time . such wide spread , low level star formation would be difficult to image around more distant galaxies but may be detectable through intervening absorption in quasar spectra .
supersymmetry ( susy ) provides perhaps the best motivated extension of the standard model . spontaneous susy breaking leads naturally to radiative electroweak symmetry breaking with masses of order the electroweak scale for the superpartners of the standard model ( sm ) particles . if the messenger interactions which couple the sm superpartners to the susy breaking sector are stronger than gravity , the lightest supersymmetric particle ( lsp ) is the goldstone fermion of supersymmetry breaking , the goldstino @xmath8 . the next to lightest supersymmetric particle ( nlsp ) is generally the lightest sm superpartner . if the intrinsic scale of supersymmetry breaking is below @xmath9 tev the nlsp can decay to its sm partner and the goldstino on laboratory length scales @xcite . this has an important impact on experimental susy signatures at high energy colliders . since superpartners are generally produced in pairs , these decays give rise to final states with two hard partons and missing energy ( @xmath10 ) carried by the goldstino pair , and with possibly other partons in the final state from cascade decays to the nlsp @xcite . the identity of the nlsp determines the type of final states which arise from decay to the goldstino @xcite . a neutralino nlsp , @xmath11 , which is gaugino - like , can decay by @xmath12 , leading to final states with @xmath13 . a slepton nlsp , @xmath14 , can decay by @xmath15 , giving @xmath16 final states . in this letter we consider in detail the possibility of a fermionic higgsino - like neutralino nlsp . because it is the superpartner of the higgs boson , @xmath0 , a higgsino nlsp can decay by @xmath17 . in addition , since the longitudinal component of the @xmath1 boson mixes with the goldstone mode of the higgs field , @xmath18 can also result . because of a strong phase space suppression of the @xmath0 and @xmath1 final states near threshold , decay to a photon can also be important for higgsinos not too much heavier than the @xmath1 boson . pair production of higgsinos which decay to goldstinos can then give rise to the di - boson final states @xmath19 @xcite . di - boson signatures which include higgs and @xmath1 bosons and @xmath10 are quite novel discovery modes for supersymmetry in the mass range accessible to the current generation of high energy collider experiments . in conventional susy signatures , in which the lightest neutralino , @xmath11 , is assumed to escape the detector without decay to the goldstino , the mass splittings between supersymmetric particles required in order for @xmath0 or @xmath1 to arise in a cascade decay , typically imply the superpartners are too heavy to be produced in sufficient numbers at present colliders . for this reason events with reconstructed @xmath1 bosons are in fact generally rejected in present susy searches . however , since the goldstino is essentially massless , sufficient phase space is available for the @xmath20 and @xmath21 modes for a higgsino somewhat heavier than @xmath0 or @xmath1 . and this mass range will be accessible at the upcoming run ii at the fermilab tevatron . the higgs final states also present the exciting possibility of discovering and studying the higgs boson in association with supersymmetry . if supersymmetry is broken at a low scale , as required for the di - boson sigatures discussed here , it is very likely that the sm gauge interactions play some role in coupling the susy breaking sector to the sm superpartners @xcite . however , such gauge - mediated susy breaking requires additional interactions between the higgs and susy breaking sectors in order to break certain higgs sector global symmetries and obtain acceptable electroweak symmetry breaking @xcite . these interactions can modify the higgsino mass from minimal expectations , and allow for a higgsino nlsp . so searches for di - boson signatures of a higgsino nlsp within theories of low scale gauge - mediated susy breaking are very well motivated as possible indirect probes for the existence of these additional couplings . the higgsinos @xmath22 and @xmath23 are fermionic superpartners of the higgs boson fields @xmath24 and @xmath25 . the neutral higgsinos mix with the gaugino superpartners of the @xmath2 and @xmath1 gauge bosons , while the charged higgsino mixes with the gaugino superpartner of the @xmath26 gauge boson . in the limit relevant here , in which the gauginos are heavier than the higgsinos , the two lightest neutralinos and lightest chargino , @xmath11 , @xmath27 , @xmath28 , are predominantly higgsino and approximately degenerate . the splitting between these states is on the order of 10 - 15 gev for masses in the range 120 - 250 gev discussed below . if the @xmath29 and @xmath30 gaugino mass parameters , @xmath31 and @xmath32 , have the same sign , @xmath33 then @xmath11 is the nlsp . for @xmath34 it is however possible in certain regions of parameter space that @xmath28 is the nlsp . in this letter only a @xmath11 nlsp , which leads to the interesting di - boson signatures , will be considered . the branching ratios @xmath35 are determined by the higgsino and gaugino content of @xmath11 @xcite . = 3.5 in this is illustrated in fig . [ br ] as a function of the neutralino mixing angle @xmath36 for fixed @xmath11 mass , where @xmath37 is the higgsino mass parameter , and @xmath38 is the ratio of higgs expectation values . for definiteness the higgs decoupling limit in which decays to the heavy scalar and pseudoscalar higgs bosons , @xmath39 and @xmath40 , are kinematically blocked is employed throughout . for gaugino - like @xmath11 the @xmath2 mode dominates , but for higgsino - like @xmath11 the @xmath0 and @xmath1 modes become important . the dependence on @xmath41 and @xmath42 apparent in fig . [ br ] can be understood in terms of the @xmath11 quantum numbers and couplings and will be presented elsewhere . the branching ratios also depend on the @xmath11 mass through the phase space available to the @xmath0 and @xmath1 modes which suffer a @xmath43 velocity suppression near threshold @xcite . so even a higgsino - like @xmath11 decays predominantly by @xmath44 for masses not too far above the @xmath0 and @xmath1 masses . = 3.5 in the mass dependence of the branching ratios is illustrated in fig . [ sigbr ] in which the @xmath45 signal cross section times branching ratio into the di - boson final states is given as a function of the @xmath11 mass for fixed higgsino - neutralino mixing . with @xmath11 higgsino - like the @xmath3 , @xmath7 , or @xmath5 modes dominate for very large masses , while the @xmath46 mode dominates for smaller masses . however , because of the strong phase space suppression near threshold there is a transition region which extends over a significant range of mass between these limits in which the mixed final states @xmath47 and/or @xmath48 ( depending on @xmath49 and @xmath50 are important . these final states are particularly useful for masses in the transition region since the photon is quite hard . the total cross section @xmath51 for @xmath52 in fig . [ sigbr ] , summed over all the higgsino - like states , is the relevant signal cross section since these states are approximately degenerate , and can all be produced at similar rates . the heavier states cascade decay to @xmath11 through neutral and charged current interactions . the partons from these cascade decays are relatively soft and probably not particularly useful at the trigger level . the final states with a @xmath1 boson can be significant for large @xmath42 , or at small @xmath42 with @xmath53 . the @xmath1 boson can decay invisibly , leptonically , or hadronically , @xmath54 , leading to many possible signatures . the @xmath55 and @xmath56 leptonic decays allow the possibility of precise reconstruction of the @xmath1 invariant mass , but suffer from small branching ratio , @xmath57 . in contrast , the invisible and hadronic decay modes can be useful because of larger branching ratios , @xmath58 , and @xmath59 . the @xmath60 di - boson mode dominates the total cross section in the transition region of masses as shown in fig . leptonic decay of the @xmath1 provides the cleanest final state , @xmath61 , which is similar to existing sm @xmath62 studies without @xmath10 @xcite . for a higgsino search , however , an additional large @xmath10 cut , as well as a more stringent photon @xmath63 cut should reduce the backgrounds to a negligible level . our monte carlo estimates indicate that this channel is practically background free , but is limited by the small leptonic branching ratio of the @xmath1 boson . the tevatron run iia with @xmath64 fb@xmath65 of integrated luminosity will have a reach at the 3@xmath66 discovery level for @xmath11 masses up to 155 gev for the parameters of fig . 2(b ) , while the reach in run iib with @xmath67 fb@xmath65 should approach 220 gev . invisible decay of the @xmath1 gives rise to the signature @xmath68 . this channel has been studied in run i as a probe for anomalous @xmath48 couplings @xcite . backgrounds include @xmath69 and @xmath70 with one jet faking a photon _ and _ in each case the remaining jet energy mismeasured to be below the minimum pedastool . the largest background in run i was from single @xmath26 production with @xmath71 and the electron misidentified as a photon . this background can be substantially reduced by raising the photon @xmath63 and @xmath10 cuts above 50 gev , beyond the jacobian peak for @xmath72 @xcite . this also reduces the hadronic background . the 3@xmath66 discovery reach in @xmath73 mass should then approach 150 ( 185 ) gev in run iia ( iib ) for the parameters of fig . [ sigbr](b ) . hadronic decay of the @xmath1 in the @xmath74 mode gives rise to the signature @xmath75 . backgrounds are similar to those of the @xmath68 channel . the @xmath75 channel has been studied in run i in order to place limits on squark and gluino masses in very specific supersymmetric models @xcite . further background suppressions not included in the run i study are possible with acoplanarity , sphericity and invariant dijet mass cuts to reconstruct the @xmath1 boson , and a lepton veto . in any case , the total background is expected to be smaller than for the @xmath76 channel , due to the presence of two additional hard partons . given the significant @xmath1 hadronic branching ratio , the @xmath75 channel should provide somewhat better reach than the @xmath77 or @xmath68 channels in run ii . the @xmath78 di - boson mode dominates at larger @xmath11 mass as shown in fig . leptonic decay of each @xmath1 boson gives rise to the spectacular signature @xmath79 , with the lepton pairs reconstructing the @xmath1 mass ( in one choice of pairing for @xmath80 ) . this channel is expected to be essentially background free , but suffers from small leptonic branching ratio . because of this run iib will not be sensitive to this channel for the paramters of fig . [ sigbr ] . but for @xmath81 and @xmath82 with larger @xmath83 ( c.f . [ br ] ) , the 3@xmath66 discovery reach in run iib for the @xmath11 mass is 170 gev . at the lhc @xmath79 would represent the gold plated channel for the @xmath84 di - boson mode from higgsino decay . hadronic decay of one of the @xmath1 bosons gives the signature @xmath85 . an important background in this channel comes from @xmath86 production with @xmath87 and @xmath88 with the @xmath89 pair reconstructing the @xmath1 mass , and each @xmath90-jet not identified as a heavy flavor . other backgrounds arise from @xmath7 and @xmath91 in association with jets . in run iib the 3@xmath66 discovery reach in @xmath11 mass should approach 195 gev for @xmath81 and @xmath92 . rejecting backgrounds for the other decay channels of the @xmath84 di - boson mode presents more serious challenges . the decay of higgsinos to real higgs bosons gives perhaps the most interesting di - boson final states because of the opportunity to study both supersymmetry and the higgs sector . higgs boson final states are important for small @xmath42 and @xmath93 or for large @xmath42 with sufficiently large @xmath11 mass , as shown in figs . [ br ] and [ sigbr ] . in the transition region of @xmath11 mass , @xmath94 is the most important di - boson mode . with the dominant decay @xmath95 this leads to the signature @xmath96 . backgrounds include @xmath97 and @xmath98 with @xmath99 and @xmath100 and @xmath101 with one jet misidentified as a photon _ and _ in each case the remaining jet energy mismeasured to be below the minimum pedastool . based on the work presented here @xcite it has been estimated @xcite that with a single @xmath90-tag the 3@xmath66 discovery reach in @xmath11 mass should approach 210 ( 250 ) gev in run iia ( iib ) for the parameters of fig . [ sigbr](a ) . for larger @xmath11 masses the @xmath102 and/or @xmath103 modes can become important , as shown in fig . [ sigbr ] . the @xmath104 di - boson final state gives rise to the signature @xmath105 . the sizeable qcd and electroweak backgrounds to this final state can be significantly reduced by requiring at least 3 tagged @xmath90-jets with large invariant mass for two @xmath90-jet pairs @xcite , as verified by monte carlo simulation @xcite . remaining backgrounds include @xmath106 with each @xmath99 , @xmath101 with one jet misidentified as a @xmath90-jet , and @xmath107 with in each case the jet energy mismeasured to be below the minimum pedastool , and @xmath108 production with @xmath87 and one hadronic decay @xmath109 with one jet misidentified as a @xmath90-jet , and one leptonic decay @xmath88 with @xmath110 not identified . accounting for the @xmath86 background @xcite , the 3@xmath66 discovery reach in @xmath11 mass at run iib should approach 240 gev for the parameters of fig . [ sigbr](a ) . the @xmath102 di - boson mode arising from higgsino decay is similar to direct @xmath5 production . invisible decay of the @xmath1 gives the signature @xmath111 , and would contribute slightly to searches for the sm higgs boson in this channel . leptonic decay of the @xmath1 gives the signature @xmath112 . unfortunately , the dominant background from @xmath86 production with @xmath113 and @xmath88 with the @xmath89 pair reconstructing the @xmath1 mass , is very similar to the signal . because of this , run ii is not expected to be sensitive to this channel . hadronic decay with @xmath99 gives the signature @xmath105 , similar to the @xmath103 mode . however , because of the smaller branching ratio , @xmath114 , run ii will just marginally not be sensitive to the @xmath102 mode in this channel for the parameters of fig . [ sigbr ] . the higgs boson final states of higgsino decay discussed above present the possibility of collecting a relatively clean sample of events which contain real higgs bosons . it is therefore interesting to consider the reach as a general function of both higgsino and higgs masses . the total cross section times branching ratio contours for the @xmath96 and @xmath105 channels as a function of the @xmath0 and @xmath11 masses are shown in fig . 3 . = 3.5 in these contours include @xmath115 for @xmath116 and @xmath117 and sm values for @xmath118 . the run iia 3@xmath66 dicovery reach quoted above for the @xmath96 channel corresponds to a signal times branching ratio cross section of 5 fb . for the parameters of fig . 3 this corresponds to a higgs mass of up to at least 120 gev for @xmath11 masses in the range 135 - 200 gev , with a maximum reach in higgs mass of just over 130 gev . this is to be contrasted with the search for the sm higgs from direct @xmath119 and @xmath120 production . these sm channels are background limited , and no sensitivity to a higgs mass beyond current limits is expected in run iia @xcite . so the @xmath96 channel presents the interesting possibility for run iia of a susy signal which contains real higgs bosons . the run iib 3@xmath66 dicovery reaches quoted above for the @xmath96 and @xmath105 channels correspond to signal times branching ratio cross sections of 1 fb and 4 fb respectively . for the parameters of fig . 3 the maximum reach in higgs mass then corresponds to just over 145 gev and 115 gev respectively . in order to identify the higgs boson directly in a sample of events arising from higgsino decays it is necessary to observe a peak in the @xmath121 invariant mass . the identifiable di - boson final states and large @xmath10 carried by the goldstinos render the supersymmetric higgs boson final states discussed here relatively clean . reconstructing the higgs mass peak should be relatively straightforward compared to sm @xmath119 and @xmath120 production modes which suffer from much larger continuum @xmath121 backgrounds . all the new signatures presented here involve hard photons , leptons , and/or @xmath90-jets , in association with significant missing energy . new triggers are therefore not required , but final state specific off line analysis should be implemented in order to search for supersymmetry and/or the higgs boson in these interesting channels . finally , higgsino decay with a measurable macroscopic decay length to the goldstino would render all the di - boson final states discussed here essentially background free . a search for such final states requires a special analysis for displaced @xmath122 , @xmath70 , or @xmath121 with large invariant mass and approximately uniform angular distribution with respect to the beam axis @xcite . we would like to thank r. demina , g. landsberg , and j. wells for invaluable discussions . the work of k. m. was supported by fermilab which is operated under doe contract de - ac02 - 76ch03000 , and that of s. t. by the us national science foundation under grant phy98 - 70115 , the alfred p. sloan foundation , and stanford university through the frederick e. terman fellowship . k. matchev , talk given at susy98 , july 11 - 17 1998 , oxford , england , http://hepnts1.rl.ac.uk/susy98 ; + s. thomas , talks presented at the susy / higgs run ii workshops , batavia , il , may 14 - 16 1998 , september 3 - 4 1998 . s. ambrosanio , g. l. kane , g. d. kribs , s. p. martin , and s. mrenna , hep - ph/9605398 , phys . d * 54 * 5396 ( 1996 ) ; + j. bagger , k. matchev , d. pierce and r .- j . zhang , hep - ph/9609444 , phys . d * 55 * 3188 ( 1997 ) . j. qian , hep - ph/9903548 , fermilab - fn-680 ; report of the low - scale supersymmetry breaking working group , conveners r. culbertson , s. martin , j. qian , and s. thomas , http://fnth37.fnal.gov/gm/gmrun2.html .	in supersymmetric theories of nature the higgsino fermionic superpartner of the higgs boson can arise as the lightest standard model superpartner depending on the couplings between the higgs and supersymmetry breaking sectors . in this letter the production and decay of higgsino pairs to the goldstone fermion of supersymmetry breaking and the higgs boson , @xmath0 , or gauge bosons , @xmath1 or @xmath2 are considered . relatively clean di - boson final states , @xmath3 , @xmath4 , @xmath5 , @xmath6 , or @xmath7 , with a large amount of missing energy result . the latter channels provide novel discovery modes for supersymmetry at high energy colliders since events with @xmath1 bosons are generally rejected in supersymmetry searches . in addition , final states with real higgs bosons can potentially provide efficient channels to discover and study a higgs signal at the fermilab tevatron run ii . # 1#2#3#4]#5 # 1 # 2 # 3 # 4 nucl . phys . * b#1 * , # 2 ( # 3)#4 # 1 # 2 # 3 # 4 phys . lett . * b#1 * , # 2 ( # 3)#4 # 1 # 2 # 3 # 4 phys . rev . * d#1 * , # 2 ( # 3)#4 # 1 # 2 # 3 # 4 phys . . lett . * # 1 * , # 2 ( # 3)#4 # 1 # 2 # 3 # 4 phys . rept . * # 1 * , # 2 ( # 3)#4 # 1 # 2 # 3 # 4 mod . . lett . * a#1 * , # 2 ( # 3)#4 # 1 # 2 # 3 # 4 z. phys . * c#1 * , # 2 ( # 3)#4
precise determinations of the strong coupling constant are important for strong interaction phenomenology and searches for new physics . lattice qcd provides a good methodology for such a determination @xcite . the staggered quark formulation allows for much faster simulation of dynamical quarks than other formulations . it maintains a remnant chiral symmetry at the cost of not entirely removing the doublers , which appear as additional ` tastes ' of quarks . the presence of the doublers gives rise to large lattice artefacts through taste - changing interactions . the realisation that these can be systematically reduced by a symanzik improvement programme opens the way to generating ensembles with light dynamical quarks and small lattice artefacts . the milc collaboration have generated a number of dynamical ensembles using the asqtad action for the dynamical quarks and either 2 dynamical flavors , 3 degenerate dynamical flavors or 2 degenerate plus 1 heavier dynamical flavors , which should be close to the real world @xcite . in the 2 + 1 flavor case they have been able to reduce the light dynamical mass to 1/5 of the heavier one ( which is approximately at @xmath1 ) . the asqtad staggered action includes a set of staples in place of the simple 1-link derivative , to suppress gluon exchange with momentum @xmath2 which couples the tastes together . these staples include a correction for low - momentum @xmath3 artefacts introduced by this procedure . in addition there is a 3-link derivative ( naik ) term to improve rotational discretisation errors through @xmath3 . the asqtad action then has @xmath4 rotational errors and @xmath5 taste - symmetry breaking interactions . this is combined in milc ensembles with a gluon action improved to the level of @xmath5 discretisation errors . we have calculated the @xmath6 spectrum on these ensembles using the nrqcd action for the @xmath7 quark @xcite . the physics of the bottomonium ( @xmath6 ) system is that of non - relativistic @xmath7 quarks ( @xmath8 10% ) and so the nrqcd action is appropriate @xcite . we use a hamiltonian correct through @xmath9 where all terms are tadpole - improved using an estimate of the landau gauge link @xcite . for the purposes of determining the lattice spacing to set the scale for @xmath10 we need only the radial and orbital excitation energies . these are experimentally well - determined and have the advantage that they are to a good approximation independent of heavy quark mass so that sophisticated tuning of the @xmath7 mass is not necessary . in addition they are good probes of the dynamical quark content of the theory because the @xmath6 has no valence light quarks . the leading contribution to the radial and orbital splittings is at @xmath11@xmath12 and so the expected systematic error from higher order ( e.g. @xmath13 or @xmath14 ) terms not included in the nrqcd action is expected to be at the 1% level . the splittings can be determined very precisely on the lattice because of the speed of nrqcd correlator calculations . we use the perturbative expansion of the plaquette in terms of a physically well - motivated coupling , @xmath15 , defined from the heavy quark potential : @xmath16 we then define @xmath17 by the following expansion of the logarithm of the plaquette , where @xmath18 is equal to @xmath19 to the order ( second ) to which we are working here . @xmath20\end{aligned}\ ] ] the numbers in brackets indicate numerical uncertainties in the perturbative coefficients from the numerical integration involved in their calculation . this perturbative expansion is for 1-loop symanzik - improved gluon action using tadpole - improvement with @xmath21 from the plaquette and @xmath22 flavors of aqtad improved staggered quarks , as appropriate to the milc ensembles . the scale , @xmath23 , is set using the blm scheme and is @xmath24 in this case . the perturbative expansion is well - behaved and shows small corrections for non - zero @xmath22 . @xmath18 can be converted to @xmath25 using the equality of @xmath18 and @xmath19 and the continuum relation between @xmath19 and @xmath25 . @xmath26\ ] ] figure 1 shows results for @xmath18 evolved to a common ( and arbitrarily chosen ) physical scale of 8.2 gev for various milc ensembles . notice the clear separation of results for @xmath27 , 2 and 3 . the figure shows results using either the @xmath6 orbital splitting or radial splitting to set the lattice spacing . the orbital splitting is calculated from the energy difference between the @xmath28 ( @xmath29 ) and @xmath30 ( @xmath6 ) states on the lattice . the experimental result for the @xmath28 mass is not known and so is taken as the spin - average of the masses of the @xmath31 ( @xmath32 ) states . the difference is expected to be very small . the radial splitting is the difference in energies of the @xmath33 ( @xmath34 ) and @xmath30 states , which is experimentally well - known . the value for @xmath35 agrees well between the two determinations on the 2 + 1 dynamical ensemble but not on the others . this reflects the fact that the lattice spacing will depend on the quantity chosen to fix it if the vacuum does not have dynamical content which is close enough to the ` real world ' . we can take the result for @xmath18 on the 2 + 1 dynamical ensemble , convert to @xmath25 and run to @xmath36 . the result for @xmath37 is then 0.121(3 ) . the dominant source of error ( 0.002 ) is the perturbative error from unknown higher orders in perturbation theory in eq . this result agrees with the world average and with our previous results using dynamical ( unimproved ) staggered quarks @xcite . other sources of systematic error give small effects . we can see from figure 1 that the @xmath22 = 3 results show very little dependence on dynamical light quark mass . extrapolating @xmath18 with @xmath38 would change the final result by less than 0.001 . the lattice spacing dependence of the result is expected to be very small but we currently have @xmath6 results only on the coarser of the milc set of ensembles . we can estimate the dependence from other results , however . figure 2 shows the dependence of @xmath18 in the quenched approximation on @xmath3 for improved and unimproved glue @xcite . here we have used @xmath39 to set the scale ( and quote @xmath18 at a scale of @xmath40 ) because it can be fixed very precisely . its disadvantage is that it is not clear how to convert it to a physical result . the figure shows that unimproved glue has very small dependence on @xmath41 and improved glue ( appropriate to the milc ensembles ) even less . we fold this lattice spacing dependence along with quark mass effects above and nrqcd systematics into a final error of 0.003 . the first calculation of @xmath10 with 2 + 1 flavours of dynamical quarks gives results in agreement with previous work , but without extrapolating in @xmath22 . further improvements , including the calculation of the next order in perturbation theory , are planned to reduce the error even further . 9 c. davies _ et al _ , phys . rev . d56 ( 1997 ) 2755 . milc collaboration , phys . rev . d64 ( 2001 ) 054506 . a. gray _ et al _ , these proceedings . g. p. lepage _ et al _ , phys . rev . d46 ( 1992 ) 4052 . we thank steve gottlieb for providing the unpublished milc results for this plot .	we describe the first lattice determination of the strong coupling constant with 3 flavors of dynamical quarks . the method follows previous analyses in using a perturbative expansion for the plaquette and upsilon spectroscopy to set the scale . using dynamical configurations from the milc collaboration with 2 + 1 flavors of dynamical quarks we are able to avoid previous problems of haivng to extrapolate to 3 light flavors from 0 and 2 . our result agrees with our previous work : @xmath0 .
we consider quantum group counterparts of several important structures manifested in logarithmic @xmath3 conformal field theory models examples of conformal field theory with nonsemisimple representation categories , which have been the subject of some attention from various standpoints @xcite . the logarithmic @xmath3 models can be quite interesting ( for the @xmath11 model , there is evidence of its relation to the percolation problem @xcite ) , but their attractiveness is matched or , rather , counterbalanced by the difficulty of studying them directly , which arises primarily because the symmetries of such models are not the virasoro algebra but its nonlinear extensions , some @xmath1-algebras ( and in the @xmath11 model , for example , the leading pole in the operator product expansion of two currents generating the @xmath1-algebra is already of order @xmath12 ) . this highlights the importance of `` indirect '' methods , which concentrate around the kazhdan lusztig correspondence @xcite . the origin of the kazhdan lusztig correspondence for logarithmic models can already be seen in their definition as _ kernels _ of screening operators acting in a free - field space ( we recall that the rational minimal models are defined as the _ cohomology _ of screenings @xcite ) . the kazhdan lusztig - dual quantum group @xmath13 is derived from the action of screenings that makes the space of states in conformal field theory into a _ bimodule_a module over the @xmath1-algebra and the quantum group @xmath13 . in the @xmath14 logarithmic models , this leads to the equivalence of braided quasitensor categories of the @xmath1-algebra representations and of the quantum group representations @xcite . quantum group counterparts can be found not only for irreducible @xmath1-algebra representations@xmath15characters but also for a larger space @xmath16 of torus amplitudes . we recall that in the nonsemisimple ( logarithmic ) case , @xmath16 is spanned not only by the characters of irreducible representations but also by `` generalized characters '' ( or `` pseudocharacters '' ) associated with some _ pseudotraces _ @xcite ( cf . @xcite ) . on the quantum group side , similarly , irreducible @xmath13-modules provide only a subspace in the space @xmath17 of @xmath18-characters ( @xmath19-invariant functionals on the quantum group ) ; spanning all of @xmath17 by elements associated with representations requires taking pseudotraces associated with some indecomposable @xmath13-modules . it can be expected in some generality that @xmath20 i.e. , the space of torus amplitudes and the space of @xmath18-characters of the kazhdan lusztig - dual quantum group are isomorphic.such an isomorphism becomes much more interesting if it extends to some structures defined on the respective spaces . the modular group action is a structure of major importance on @xmath16 . remarkably , it is nicely matched by the general theory of the modular group action on ribbon factorizable quantum groups @xcite ( also see @xcite for an application of this theory to the _ small _ quantum group ) . for factorizable quantum groups , there is an isomorphism ( actually , of associative algebras @xcite ) @xmath21 which allows replacing @xmath22 with a somewhat more `` tangible '' object , the quantum group center @xmath23 . for the @xmath3 models , the @xmath24 representation @xmath25 constructed on the center of the kazhdan lusztig - dual quantum group in accordance with the recipe in @xcite turns out to be equivalent to the @xmath24 representation on the torus amplitudes , as we show in what follows ( this has been known for the @xmath14 logarithmic models @xcite ) . to formulate the result , we use the notation @xmath26 and @xmath27 for a fixed pair of coprime positive integers and set - type '' notation , but it allows saving some space by `` @xmath28 '' formulas . ] @xmath29 _ on the conformal field theory side _ , the chiral algebra of the logarithmic @xmath30 conformal field theory is a certain `` triplet '' @xmath1-algebra @xmath31 defined in @xcite . the characters of its irreducible representations give rise to a @xmath32-dimensional representation @xmath33 of the modular group @xmath24 , evaluated as @xcite @xmath34 where @xmath35 is the standard two - dimensional representation , @xmath36 is its symmetrized square , @xmath37 is the @xmath38-dimensional @xmath24 representation on the characters of the @xmath30 virasoro minimal model , and @xmath39 , @xmath40 , and @xmath41 are certain @xmath24 representations of the respective dimensions @xmath42 , @xmath43 , and @xmath44 ( see @xcite for the precise formulas ) . _ on the quantum group side _ , the quantum group @xmath45 that is kazhdan lusztig - dual to the @xmath30 logarithmic conformal field theory is the quotient @xmath46 of the product of two restricted quantum groups @xmath47 at the roots of unity @xmath48 over the ( hopf ) ideal generated by the central element @xmath49 . [ thm : q - structure ] 1 . the center @xmath23 of @xmath45 is @xmath50-dimensional 2 . [ claim : factorization ] the @xmath24 representation @xmath25 on @xmath23 factors into the product of three @xmath24 representations , @xmath51 which pairwise commute i.e. , @xmath52 , @xmath53@xmath54@xmath55@xmath53 , and @xmath56 3 . [ claim : decomp]the center decomposes with respect to this product as @xmath57 where @xmath35 is the standard two - dimensional representation , @xmath58 is the trivial representation , @xmath37 is the @xmath59-dimensional @xmath24 representation on the characters of the @xmath30 virasoro minimal model , and @xmath39 , @xmath40 , and @xmath41 are those in 4 . [ claim : grring]the representation @xmath60 is the @xmath24 representation on the grothendieck ring of @xmath45 5 . [ claim : equiv ] moreover , the decomposition @xmath61 establishes an equivalence of the @xmath24 representation on @xmath23 to the representation on the space of generalized @xmath31-characters in . the @xmath45 quantum group has @xmath62 irreducible representations , which we label as @xmath63 with @xmath64 and @xmath65 . [ thm : gr - ring ] multiplication in the @xmath45 grothendieck ring is given by @xmath66 where @xmath67{l } 1{\,{\leqslant}\,}r{\,{\leqslant}\,}p_+,\\[-6pt ] 1{\,{\leqslant}\,}r'{\,{\leqslant}\,}p_- , \end{array } \\ { \mathsf{x}}^{\alpha}_{2p_+ - r , r ' } + 2{\mathsf{x}}^{-\alpha}_{r - p_+ , r ' } , & \begin{array}[t]{l } p_+\!+\!1{\,{\leqslant}\,}r{\,{\leqslant}\,}2 p_+\!-\!1,\\[-6pt ] 1{\,{\leqslant}\,}r'{\,{\leqslant}\,}p_- , \end{array } \\ { \mathsf{x}}^{\alpha}_{r,2p_- - r ' } + 2{\mathsf{x}}^{-\alpha}_{r , r ' - p_- } , & \begin{array}[t]{l } 1{\,{\leqslant}\,}r{\,{\leqslant}\,}p_+,\\[-6pt ] p_-\!+\!1{\,{\leqslant}\,}r'{\,{\leqslant}\,}2 p_-\!-\!1 , \end{array } \\ \mbox{}\kern-3pt\begin{aligned}[b ] & { \mathsf{x}}^{\alpha}_{2p_+ - r , 2 p_- - r ' } + 2{\mathsf{x}}^{-\alpha}_{2p_+ - r , r ' - p_-}\\[-3pt ] & { } + 2{\mathsf{x}}^{-\alpha}_{r - p_+ , 2 p_- - r ' } + 4{\mathsf{x}}^{\alpha}_{r - p_+ , r ' - p_- } , \end{aligned } & \begin{array}[t]{l } p_+\!+\!1{\,{\leqslant}\,}r{\,{\leqslant}\,}2 p_+\!-\!1,\\[-6pt ] p_-\!+\!1{\,{\leqslant}\,}r'{\,{\leqslant}\,}2 p_-\!-\!1 . \end{array } \end{cases}\ ] ] we also characterize this ring as a quotient of @xmath68 $ ] ( see * * ) . its interpretation as a @xmath31 fusion ring is discussed in @xcite . to derive the kazhdan lusztig - dual quantum group for the logarithmic @xmath30 conformal field theory models , we follow the strategy proposed in @xcite for @xmath14 models . the starting point is the hopf algebra @xmath69 of screening operators in a free - field space . in the @xmath14 case , @xmath69 is a taft hopf algebra at a certain root of unity , and in the @xmath30 case , @xmath69 is constructed from two taft algebras ( taken at different roots of unity ) . we then pass to the drinfeld double @xmath70 . from the conformal field theory standpoint , the double is a quantum group generated by the screening and contour - removal operators , the latter being just the dual to the former in drinfeld s construction . but the doubling procedure also yields the dual @xmath71 to the cartan element @xmath72 in @xmath69 , and it is to be `` eliminated '' as @xmath73 ; this amounts to taking a quotient @xmath74 of @xmath70 . as any drinfeld double , @xmath70 is quasitriangular , i.e. , has a universal @xmath75-matrix @xmath76 . it follows that @xmath74 is still quasitriangular , but the @xmath77-matrix @xmath78 turns out to be an element not of @xmath79 but of @xmath80 , where ( for the @xmath30 model specifically ) @xmath81 is the subalgebra in @xmath74 generated by @xmath82 and the other @xmath74 generators . this @xmath81 can be alternatively described as in . this kazhdan - lusztig - dual quantum group @xmath81 is _ not _ quasitriangular , but we have a pair @xmath83 of quantum groups , one of which has an @xmath77-matrix but not an @xmath75-matrix , and the other has an @xmath75-matrix but not a ( nondegenerate ) @xmath77-matrix . abusing the terminology , we still refer to the @xmath81 quantum group as factorizable ( although it , e.g. , is not unimodular ) . such a `` deficient factorizability '' suffices for constructing a modular group action on the @xmath81 center @xmath23 . in accordance with @xcite , the action of @xmath84 on @xmath23 is constructed in terms of the drinfeld and radford maps @xmath85 and @xmath86 and is given by composing one of these with the inverse of the other , @xmath87_{{\boldsymbol{\chi } } } \ar[dr]^{{\widehat{\pmb{\phi } } { } } } \\ { \mathsf{z}}&{}&{\mathsf{z}}\ar@/^/[ll]^{{\mathscr{s}}^{-1 } } } \ ] ] ( the radford map @xmath86 is the inverse compared with the standard definition ) . next , @xmath88 is represented on @xmath23 essentially by ( multiplication with ) the ribbon element , @xmath89 here , too , the properties of @xmath83 are such that the ribbon element @xmath90 actually belongs to @xmath81 ( not surprisingly , in view of the relation between the @xmath77-matrix and the ribbon element ) . we prove theorem * * rather directly , by an explicit calculation relying on the existence of two special bases in the center , associated with the drinfeld and radford maps . some `` obvious '' central elements that are related by @xmath91 are given by the radford and drinfeld images of the ( `` quantum '' ) traces over irreducible representations . map of the basis of characters and pseudocharacters in @xmath92 yields the ishibashi states , i.e. , states with a certain `` momentum . '' the @xmath85 map of the same elements in @xmath92 yields the localized , i.e. , cardy states @xcite . ] adding some pseudotraces yields the `` radford '' and `` drinfeld '' bases in @xmath23 , with @xmath91 mapping between them . in @xcite and in @xcite . ] the theorem eventually follows from studying how the ribbon element acts on the two bases . the relation between the @xmath24 representations in claims [ claim : decomp ] and [ claim : grring ] can be stated in terms of an automorphy factor , similarly to @xcite . the factorization itself is related to the multiplicative jordan decomposition of the ribbon element but goes somewhat further : the ribbon element factors as @xmath93 the first factor is the semisimple part , constructed from the primitive idempotents @xmath94 ( weighted by exponentials of the conformal dimensions of @xmath31 primary fields in the @xmath30 logarithmic conformal field theory model ) ; in the other two factors , which are unipotent , @xmath95 and @xmath96 are elements in the radical given by drinfeld - map images of certain pseudotraces associated with indecomposable @xmath45 representations in the full subcategory containing the trivial representation . the rest of the paper is not difficult to describe , despite its length . we recall how the quantum group is derived from conformal field theory data in * * and describe its representations in * * ( and appendix [ app : priojective ] ) , @xmath18-characters in * * , and the center in * * . we then travel along the diagram in from top down , considering the @xmath86 arrow in sec . [ sec : radford ] and the @xmath85 arrow in sec . [ sec : drinfeld ] . the radford and drinfeld maps @xmath86 and @xmath85 give two `` @xmath91-adapted '' bases in the center . it then remains to study how @xmath97 acts on these bases . this amounts to studying the action of the ribbon element derived in * * . we prove theorem * * in sec . [ sec : sl2z ] . the basic ingredients for the @xmath86 map ( and its inverse ) , the integral and cointegral , are found in * * ; for the @xmath85 map , the required @xmath77-matrix is derived from the drinfeld double in * * . the analysis of the @xmath18-characters @xmath92 and the center @xmath23 relies on the study of representations , with an important role played by projective and some other indecomposable modules . these are studied in appendix [ app : priojective ] . in appendix [ app : the - center ] , we detail the construction of the @xmath45 center and collect some explicit calculations . in @xmath98 $ ] , we consider the @xmath18-integers and @xmath18-binomial coefficients @xmath99}_q = { \mbox{\footnotesize$\displaystyle\frac{q^{n}-q^{-n}}{q - q^{-1}}$}},\quad { { \mathchoice { { { \mbox{\footnotesize$\displaystyle \genfrac{[}{]}{0pt}{}{m}{n}$}}}}{{\genfrac{[}{]}{0pt}{}{m}{n } } } { { \genfrac{[}{]}{0pt}{}{m}{n}}}{{\genfrac{[}{]}{0pt}{}{m}{n}}}}}_q = { \mbox{\footnotesize$\displaystyle\frac{{[m]}_q!}{{[n]}_q!\,{[m - n]}_q!}$}},\quad { [ n]}_q ! = { [ 1]}_q{[2]}_q\dots { [ n]}_q.\ ] ] we then let @xmath100}_+={[m]}_{{\mathfrak{q}_+}^{p_-}},\quad { [ m]}_-={[m]}_{{\mathfrak{q}_-}^{p_+}},\quad { { \mathchoice { { { \mbox{\footnotesize$\displaystyle \genfrac{[}{]}{0pt}{}{m}{n}$}}}}{{\genfrac{[}{]}{0pt}{}{m}{n } } } { { \genfrac{[}{]}{0pt}{}{m}{n}}}{{\genfrac{[}{]}{0pt}{}{m}{n}}}}}_+={{\mathchoice { { { \mbox{\footnotesize$\displaystyle \genfrac{[}{]}{0pt}{}{m}{n}$}}}}{{\genfrac{[}{]}{0pt}{}{m}{n } } } { { \genfrac{[}{]}{0pt}{}{m}{n}}}{{\genfrac{[}{]}{0pt}{}{m}{n}}}}}_{{\mathfrak{q}_+}^{p_-}},\quad { { \mathchoice { { { \mbox{\footnotesize$\displaystyle \genfrac{[}{]}{0pt}{}{m}{n}$}}}}{{\genfrac{[}{]}{0pt}{}{m}{n } } } { { \genfrac{[}{]}{0pt}{}{m}{n}}}{{\genfrac{[}{]}{0pt}{}{m}{n}}}}}_-={{\mathchoice { { { \mbox{\footnotesize$\displaystyle \genfrac{[}{]}{0pt}{}{m}{n}$}}}}{{\genfrac{[}{]}{0pt}{}{m}{n } } } { { \genfrac{[}{]}{0pt}{}{m}{n}}}{{\genfrac{[}{]}{0pt}{}{m}{n}}}}}_{{\mathfrak{q}_-}^{p_+}}\ ] ] denote their respective specializations at @xmath101 ( see ) . the fixed coprime pair @xmath102 and @xmath103 determines the four sets of pairs @xmath104 and @xmath105 with @xmath106 . the sets @xmath107 , @xmath108 , and @xmath109 are subsets of the extended kac table , a @xmath110 rectangle , and @xmath111 additionally contains the box conventionally labeled by @xmath112 ( it can be considered a ` second copy' of the steinberg box @xmath30 ) . for @xmath113 and @xmath114 , these are illustrated in fig . [ fig : kac - table ] . the quantum group corresponding to the logarithmic @xmath30 conformal field theory model , @xmath45 at an appropriate root of unity , is introduced in * * . we then consider its representations in * * , describe the space of @xmath18-characters in * * and give the structure of the @xmath45 center in * * . we refer the reader to @xcite for the fundamental facts regarding hopf algebra structures . a `` gateway '' between the direct study of the logarithmic @xmath30 model and its quantum group counterpart is provided by the screening operators @xmath116 and @xmath117 and a `` cartan '' operator @xmath72 constructed from the free - field zero mode @xcite . they act on the chiral sector of the space of states in the logarithmic model such that the relations @xmath118 are satisfied . let @xmath69 denote the hopf algebra with these generators and relations ; its comultiplication , antipode , and counit ( obtained by naturally combining those for the two taft hopf algebras generated by @xmath119 and @xmath120 ) are given in @xcite . the pbw basis in @xmath69 is @xmath121 we next construct the drinfeld double @xmath70 of @xmath69 . this involves introducing generators @xmath71 , @xmath122 , and @xmath123 , `` dual '' to @xmath72 , @xmath116 , and @xmath117 , by the relations @xmath124 further , we eliminate @xmath71 , which is extraneous from the conformal field theory standpoint , by taking the quotient @xmath125 over the _ hopf _ ideal generated by the central element @xmath126 . we then let @xmath45 be the subalgebra in @xmath127 generated by ( the images of ) @xmath116 , @xmath122 , @xmath123 , @xmath117 , and @xmath128 . it is shown in @xcite that this it can be equivalently described as follows . [ xxx - def ] @xmath45 is the hopf algebra generated by @xmath116 , @xmath122 , @xmath123 , @xmath117 , and @xmath129 with the relations @xmath130 = { \mbox{\footnotesize$\displaystyle\frac{k^{p_- } - k^{-p_-}}{{\mathfrak{q}_+}^{p_-}\!- { \mathfrak{q}_+}^{-p_-}}$}},\qquad { [ } { e_-},{f_- } ] = { \mbox{\footnotesize$\displaystyle\frac{k^{p_+ } - k^{-p_+}}{{\mathfrak{q}_-}^{p_+}\!- { \mathfrak{q}_-}^{-p_+}}$}}\\ \intertext{and the hopf algebra structure } \delta(k)= k{\otimes}k , \quad \delta({e_+})={e_+}{\otimes}{\boldsymbol{1}}+ k^{p_-}{\otimes}{e_+ } , \quad \delta({f_-})={f_-}{\otimes}{\boldsymbol{1}}+ k^{-p_+}{\otimes}{f_- } , \\ \delta({f_+})={f_+}{\otimes}k^{-p_-}+{\boldsymbol{1}}{\otimes}{f_+ } , \quad \delta({e_-})={e_-}{\otimes}k^{p_+}+{\boldsymbol{1}}{\otimes}{e_- } , \\ s(k)= k^{-1},\quad s({e_+ } ) = -k^{-p_-}{e_+ } , \quad s({f_-})=-k^{p_+}{f_-},\\ s({f_+})=-{f_+}k^{p_- } , \quad s({e_-})=-{e_-}k^{-p_+},\\ \epsilon({e_\pm})=\epsilon({f_\pm})=0,\quad\epsilon(k)=1 . \end{gathered}\ ] ] [ [ section ] ] furthermore , @xmath45 can be described as the quotient of the product of two restricted quantum groups @xmath47 with @xmath48 over the hopf ideal generated by the central element @xmath49 . by @xmath131 at @xmath132 , we mean the quantum group with the relations @xmath133 and @xmath134 , the standard relations @xmath135 = { \mbox{\footnotesize$\displaystyle\frac{k - k^{-1}}{q\!- { \mathfrak{q}}^{-1}}$ } } , \\ \intertext{and the hopf algebra structure given by } \delta(k)= k{\otimes}k , \quad \delta ( e)= e { \otimes}{\boldsymbol{1}}+ k{\otimes}e , \quad \delta(f)=f{\otimes}k^{-1 } + { \boldsymbol{1}}{\otimes}f , \\ s(k)= k^{-1},\quad s(e ) = -k^{-1 } e,\quad s(f)=- f k. \end{gathered}\ ] ] indeed , because @xmath102 and @xmath103 are coprime , the @xmath45 algebra in * * is equivalently generated by @xmath116 , @xmath122 , @xmath123 , @xmath117 , and the elements @xmath136 from the formulas in * * , we then have @xmath137 = { \mbox{\footnotesize$\displaystyle\frac{k_{\pm } - k_{\pm}^{-1 } } { { \mathfrak{q}}_{\pm}^{p_{\mp}}\!- { \mathfrak{q}}_{\pm}^{-p_{\mp}}}$}},\end{gathered}\ ] ] and it follows in addition that @xmath138 and @xmath139 . the formulas for @xmath140 and @xmath141 also `` separate , '' e.g. , @xmath142 . the two @xmath143 algebras are therefore @xmath144 and @xmath145 . clearly , the relations @xmath146 are satisfied by @xmath147 in . in what follows , we need the irreducible and projective @xmath45-modules ; on the way from the former to the latter , useful intermediate objects are the verma modules . it is easy to see that there are @xmath148 irreducible finite - dimensional @xmath45-modules . we label them as @xmath149 , where @xmath64 and @xmath65 , with the highest - weight vector @xmath150 of @xmath149 defined by the relations @xmath151 it follows that @xmath152 . @xmath153 is the trivial module . the module @xmath154 is linearly spanned by elements @xmath155 , @xmath156 and @xmath157 ( with @xmath158 ) , with the @xmath45-action given by @xmath159}_+{[r\!-\!n]}_+{\|r , r',n\!-\!1,n'\rangle}^{\alpha } , \notag\\ { e_-}\ , { \|r , r',n , n'\rangle}^{\alpha}&=\alpha^{p_+}(-1)^{r-1 } { [ n']}_-{[r'\!-\!n']}_-{\|r , r',n , n'\!-\!1\rangle}^{\alpha } , \label{on - vectors}\\ { f_+}\ , { \|r , r',n , n'\rangle}^{\alpha}&={\|r , r',n\!+\!1,n'\rangle}^{\alpha } , \notag\\ { f_-}\ , { \|r , r',n , n'\rangle}^{\alpha}&={\|r , r',n , n'\!+\!1\rangle}^{\alpha } , \notag\end{aligned}\ ] ] where we set @xmath160 . there are @xmath148 verma modules , labeled as @xmath161 with @xmath162 and @xmath163 . first , these are the two steinberg modules @xmath164 next , for each @xmath165 , @xmath166 , and @xmath167 , the verma modules @xmath168 and @xmath169 can be described as the respective extensions @xmath170 for consistency with the notation used in more complicated extensions below , we also write these extensions as @xmath171 with the convention that arrows are directed toward _ _ sub__modules . next , for @xmath172 and @xmath173 , the verma module @xmath174 can be described as a second extension with the following structure of subquotients :- action on each irreducible representation is changed in agreement with the arrows connecting a given subquotient with others . for example , the lowest - weight vector of @xmath175 is annihilated by @xmath117 in the _ irreducible _ representation , but is mapped by @xmath117 into the highest - weight vector in @xmath176 in . with some more work , the remaining details of the @xmath45-action can then be reconstructed . ] @xmath177_{{f_- } } \ar@/_/[dr]^{{f_+}}&\\ & \stackrel{{\mathsf{x}}^{-a}_{r , p_--r ' } } { \diamondsuit}\ar@/^/[dr]_{{f_+ } } & & \stackrel{{\mathsf{x}}^{-a}_{p_+-r , r'}}{\heartsuit } \ar@/_/[dl]^{{f_-}}\\ & & \stackrel{{\mathsf{x}}^{a}_{p_+-r , p_--r'}}{\spadesuit } & } \ ] ] ( a `` four - suit '' module ) . we note that @xmath178 . the @xmath179 projective modules @xmath180 , with @xmath64 and @xmath65 , are constructed in the standard way , as a projective cover of each irreducible module . [ prop : proj - struc ] the subquotient structure of the projective @xmath45-modules is as follows . 1 . @xmath181 are irreducible modules . 2 . for @xmath172 , the projective module @xmath182 admits a filtration @xmath183 where @xmath184 , @xmath185 , and @xmath186 . 3 . for @xmath173 , the projective module @xmath187 admits a filtration @xmath188 where @xmath189 , @xmath190 , and @xmath191 . 4 . for @xmath172 and @xmath173 , the projective module @xmath180 admits a filtration @xmath192 where @xmath193 , @xmath194 , @xmath195 , @xmath196 , and @xmath197 . the structure of subquotients and the left @xmath45-action on @xmath180 can be visualized with the aid of the diagram in fig . [ app : proj - pic - plus ] . their definition in terms of bases , whence * * follows , is given in * * . closely following the strategy in @xcite , we obtain theorem * * describing the grothendieck ring structure of @xmath45 . we do not repeat the standard steps leading to this statement , the derivation being totally similar to the one in @xcite ; it relies on a property of tensor products with verma modules and simple explicit calculations for two - dimensional representations . the @xmath45 grothendieck ring is generated by @xmath198 and @xmath199 . in * * , it is further characterized as a polynomial quotient ring . we now use irreducible and projective modules to introduce a basis in the space @xmath200 of @xmath18-characters of @xmath45 . in subsequent sections , this basis is mapped into the center , to produce two distinguished bases there . [ [ section-1 ] ] for a hopf algebra @xmath201 , the space @xmath202 of @xmath18-characters is defined as @xmath203 where the coadjoint action @xmath204 is @xmath205 , @xmath206 , @xmath207 . in what follows , we need the so - called _ balancing element _ @xmath208 that satisfies @xcite @xmath209 for all @xmath210 . for @xmath211 , @xmath212 , as we calculate in below . the space of @xmath18-characters contains a homomorphic image of the grothendieck ring under the @xmath18-trace : for any @xmath201-module @xmath213 , @xmath214 where @xmath215 is the balancing element @xcite . for @xmath211 , we thus have a @xmath148-dimensional subspace in @xmath92 spanned by @xmath18-traces over irreducible modules , i.e. , by @xmath216 the space of @xmath18-characters @xmath217 is not spanned by @xmath18-traces over irreducible modules ; it also contains `` pseudotraces '' associated with projective modules ( in the kazhdan lusztig context , they can be considered quantum group counterparts of pseudotraces @xcite ) . the strategy for constructing the pseudotraces is as follows . for any indecomposable module @xmath218 and a map @xmath219 , the functional @xmath220 is a @xmath18-character if and only if ( cf . ) @xmath221).\ ] ] we now find maps @xmath222 satisfying . first , to choose @xmath218 , we note that the space of @xmath18-characters @xmath217 is the center of the dual algebra @xmath223 . to describe @xmath217 , we can therefore use the isomorphism between the center of @xmath223 and the algebra of _ bimodule _ endomorphisms of the regular representation of @xmath223 . as a @xmath223-bimodule , the regular representation is contragredient to the representation described in * * . this means that @xmath18-characters are endomorphisms of the modules that are contragredient to projective @xmath45-modules . we therefore take @xmath218 to be the projective @xmath45-module in one full subcategory , that is , the direct sum @xmath224 ( for @xmath225 ) of the four indecomposable projective @xmath45-modules @xmath180 described in * * . we then consider the maps @xmath226 defined by its action on the corresponding basis vectors ( see ): in each of the four modules in , @xmath222 acts by zero on all elements in except @xmath227 with the coefficient @xmath228 , @xmath229 , @xmath230 , and @xmath231 chosen arbitrarily in each module . to distinguish the basis elements in each of the four modules in , we introduce the second superscript and write @xmath232 where @xmath233 and the symbols where @xmath234 `` takes the value '' @xmath235 correspond to the basis elements in @xmath236 , the elements where @xmath237 correspond to the basis elements in @xmath238 , those where @xmath239 to the basis elements in @xmath240 , and those where @xmath241 to the basis elements in @xmath242 . this @xmath222 map thus depends on ( @xmath243 and ) sixteen coefficients but , abusing the notation , we do not write these parameters explicitly . for any such @xmath222 , we now define a functional on @xmath45 as @xmath244 a lengthy but straightforward calculation shows the following proposition . [ prop : gammaupup ] for @xmath225 , @xmath245 if and only if @xmath246 the proof is outlined in * * . in what follows , we assume the relations in * * to be satisfied . if @xmath248 , then @xmath249 is a linear combination of the @xmath18-traces over irredicible modules introduced in * * . to construct the other linearly independent @xmath18-characters , we take one of the coefficients @xmath250 , @xmath251 , @xmath252 to be nonzero and thus define @xmath253_+}{{\mathfrak{q}_+}^{p_-}\!-\!{\mathfrak{q}_+}^{-p_-}}$ } } , \qquad ( r , r')\in{\mathcal{i}_{\boxslash}},\end{gathered}\ ] ] @xmath254_+}{{\mathfrak{q}_+}^{p_-}\!-\!{\mathfrak{q}_+}^{-p_-}}$ } } , \qquad ( r , r')\in{\mathcal{i}_{\boxslash}},\end{gathered}\ ] ] @xmath255_-}{{\mathfrak{q}_-}^{p_+}\!-\!{\mathfrak{q}_-}^{-p_+}}$ } } , \qquad ( r , r')\in{\mathcal{i}_{\boxbslash}},\end{gathered}\ ] ] @xmath256_-}{{\mathfrak{q}_-}^{p_+}\!-\!{\mathfrak{q}_-}^{-p_+}}$ } } , \qquad ( r , r')\in{\mathcal{i}_{\boxbslash}},\end{gathered}\ ] ] @xmath257 for convenience , we introduce the maps @xmath258 ( with @xmath259 ) , @xmath260 ( with @xmath261 ) , and @xmath262 ( with @xmath225 ) obtained by substituting the respective sets of coefficients in . we thus have @xmath263 @xmath18-characters in the space @xmath92 : @xmath264 \underset{\strut\mbox{\footnotesize$1{\,{\leqslant}\,}r{\,{\leqslant}\,}p_+\!-\!1 $ } } { \gamma^{+}(r , p_-),~~\gamma^{-}(p_+\!-\!r , p_-),}\quad \underset{\strut\mbox{\footnotesize$ \begin{array}[t]{c } 1{\,{\leqslant}\,}r{\,{\leqslant}\,}p_+\!-\!1,\\ 1{\,{\leqslant}\,}r'{\,{\leqslant}\,}p_-\!-\!1 \end{array}$ } } { \gamma^{{{\scriptscriptstyle\boldsymbol{\nearrow\!\!\!\swarrow}}}}(r , r'),~~\gamma^{{{\scriptscriptstyle\boldsymbol{\nwarrow\!\!\kern-1pt\searrow}}}}(r , r'),}\quad \underset{\strut\mbox{\footnotesize$1{\,{\leqslant}\,}r'{\,{\leqslant}\,}p_-\!-\!1 $ } } { \gamma^{+}(p_+ , r'),~\gamma^{-}(p_+ , p_-\!-\!r'),}\\[4pt ] \underset{\strut\mbox{\footnotesize$(r , r')\in{\mathcal{i}_1}$ } } { \gamma^{+}(r , r'),~\gamma^{-}(p_+\!-\!r , r'),~ \gamma^{-}(r , p_-\!-\!r'),~\gamma^{+}(p_+\!-\!r , p_-\!-\!r')}. \end{gathered}\kern-7pt\ ] ] there are @xmath265 linearly independent elements in @xmath92 thus obtained . in what follows , we also establish that @xmath266 , with the consequence that the @xmath247 listed above are a basis in @xmath92 . we call it _ the @xmath247-basis _ in what follows . we now describe the structure of the center of @xmath45 . [ prop - center ] the center @xmath23 of @xmath45 is @xmath267-dimensional and decomposes into a direct sum of associative algebras as @xmath268 where the dimension of each algebra is shown as a superscript . furthermore , * @xmath269 is spanned by a single idempotent @xmath270 , * @xmath271 is spanned by a single idempotent @xmath272 , * each @xmath273 is spanned by the primitive idempotent @xmath274 and the radical elements @xmath275 and @xmath276 , on all of which @xmath274 acts as identity and the other products vanish , * each @xmath277 is spanned by the primitive idempotent @xmath278 and the radical elements @xmath279 and @xmath280 , on all of which @xmath278 acts as identity and the other products vanish , * each @xmath281 is spanned by the primitive idempotent @xmath94 acting as identity on @xmath281 and the radical elements @xmath282 , @xmath283 , @xmath284 , @xmath285 , @xmath286 , @xmath287 , @xmath288 , @xmath289 , which have the nonzero products @xmath290 pairwise products between elements in the radical other than those in are identically zero . the theorem is proved in appendix * * . the @xmath267 elements @xmath291 , @xmath292 , and @xmath293 are called the _ canonical central elements_. a number of facts in what follows are established by decomposing various central elements with respect to the basis of the canonical central element . but because our next aim is to study the radford map , it proves convenient to use a different basis in the center , given by the radford - map images of the above @xmath294 and actually obtained by certain mixing of the @xmath291 , @xmath292 , and @xmath293 within each respective algebra in . the @xmath24-action on the center to be considered in sec . [ sec : sl2z ] involves the radford and drinfeld maps and a ribbon structure ; the radford map @xmath86 is studied in this section . we first find the necessary ingredients for @xmath86 and @xmath296 ( the cointegral and integral ) and then evaluate the radford images of the basis in @xmath92 constructed in the previous section . we recall that for a hopf algebra @xmath201 , a _ right integral _ @xmath297 is a linear functional on @xmath201 satisfying @xmath298 for all @xmath299 . if such a functional exists , it is unique up to multiplication by a nonzero constant . right , which is the case with @xmath211 . ] _ cointegral _ @xmath300 is an element in @xmath201 such that @xmath301 whenever it exists , this element is unique up to multiplication by a nonzero constant . we also note that the cointegral gives an embedding of the trivial representation of @xmath201 in the regular bimodule @xmath201 . we use the standard normalization @xmath302 . next , a _ comodulus _ @xmath303 is an element in @xmath201 such that @xmath304 whenever a square root of the comodulus exists , we have @xcite @xmath305 where @xmath208 is the balancing element . the radford map @xmath306 and its inverse @xmath307 are given by @xmath308 both @xmath86 and @xmath296 intertwine the left actions of @xmath201 on @xmath201 and @xmath309 , with the left @xmath201-module structures given by @xmath310 on @xmath309 and by the regular action on @xmath201 , and similarly for the right actions @xcite . we use the hat for notational consistency in what follows ( and swap @xmath86 and @xmath296 compared with the commonly used notation ) . in particular , @xmath86 gives an isomorphism of linear spaces @xmath311 . we now find the integral and cointegral and then evaluate @xmath86 on the @xmath18-characters constructed in * * . a simple calculation in @xmath45 shows that the right integral is given by @xmath312 next , as is easy to verify , the left right cointegral for @xmath45 is given by the product of the cointegrals for the two @xmath313 in , @xmath314 ( which is at the same time the product of the cointegrals for the two @xmath313 in , up to normalization ) . _ factorizable ribbon _ quantum groups offer a `` canonical '' normalization of the radford map ( i.e. , of the integral and cointegral ) up a power of @xmath315 , from the condition @xmath316 for @xmath91 in , and we therefore set @xmath317_+!\,[p_- - 1]_-!)^2={\sqrt{{\mbox{\footnotesize$\displaystyle\frac{p_+ p_-}{2}$}}}}.\ ] ] we further calculate @xmath318 which allows choosing the comodulus as @xmath319 . in accordance with , we then have the balancing element @xmath320 explicitly , the radford map @xmath311 of the irreducible representation characters is given by @xmath321{l } 1{\,{\leqslant}\,}r{\,{\leqslant}\,}p_+,\\ 1{\,{\leqslant}\,}r'{\,{\leqslant}\,}p_- . \end{array}\end{aligned}\ ] ] clearly , @xmath322 , in accordance with the fact that @xmath300 furnishes an embedding of the trivial representation @xmath323 into @xmath45 . because the radford map @xmath324 is a morphism of representations , it follows that the linear span of the @xmath325 is the annihilator of the radical in the center ( see formulas below in * * ) . the elements @xmath326 are mapped under @xmath86 as @xmath327 for all @xmath328 and @xmath65 . the elements @xmath329 are mapped under @xmath86 as @xmath330 for all @xmath64 and @xmath331 finally , the elements @xmath332 with @xmath225 are mapped under @xmath86 as @xmath333 [ prop : radford - decompose ] the radford - map images @xmath334 of the @xmath247-basis in are as follows . 1 . the radford images of the irreducible representation traces @xmath335 , @xmath64 , @xmath65 , are given by @xmath336 2 . the radford images of @xmath326 with @xmath337 , @xmath338 and of @xmath329 with @xmath64 , @xmath339 are given by @xmath340 @xmath341 @xmath342 @xmath343 3 . and the radford images of @xmath332 , @xmath225 , are @xmath344 where we use the notation @xmath345 the @xmath346 elements @xmath347 are a basis in @xmath23 , and therefore the @xmath348 are a basis in @xmath92 . this is proved in * * . [ [ section-2 ] ] we thus have a basis in the center @xmath23 of @xmath45 associated with the radford map : @xmath349 \underset{\strut\mbox{\footnotesize$1{\,{\leqslant}\,}r{\,{\leqslant}\,}p_+\!-\!1 $ } } { { \widehat{\pmb{\phi}}{}}^{+}(r , p_-),~~{\widehat{\pmb{\phi}}{}}^{-}(p_+\!-\!r , p_-),}\quad \underset{\strut\mbox{\footnotesize$ \begin{array}[t]{c } 1{\,{\leqslant}\,}r{\,{\leqslant}\,}p_+\!-\!1,\\ 1{\,{\leqslant}\,}r'{\,{\leqslant}\,}p_-\!-\!1 \end{array}$ } } { { { \widehat{\pmb{\phi}}{}}^{{{\scriptscriptstyle\boldsymbol{\nearrow\!\!\!\swarrow}}}}{}}(r , r'),~~{{\widehat{\pmb{\phi}}{}}^{{{\scriptscriptstyle\boldsymbol{\nwarrow\!\!\kern-1pt\searrow}}}}{}}(r , r'),}\quad \underset{\strut\mbox{\footnotesize$1{\,{\leqslant}\,}r'{\,{\leqslant}\,}p_-\!-\!1 $ } } { { \widehat{\pmb{\phi}}{}}^{+}(p_+ , r'),~{\widehat{\pmb{\phi}}{}}^{-}(p_+ , p_-\!-\!r'),}\\[4pt ] \underset{\strut\mbox{\footnotesize$(r , r')\in{\mathcal{i}_1}$ } } { { \widehat{\pmb{\phi}}{}}^{+}(r , r'),~{\widehat{\pmb{\phi}}{}}^{-}(p_+\!-\!r , r'),~ { \widehat{\pmb{\phi}}{}}^{-}(r , p_-\!-\!r'),~{\widehat{\pmb{\phi}}{}}^{+}(p_+\!-\!r , p_-\!-\!r')}. \end{gathered}\kern-7pt\ ] ] the first line comprises @xmath350 elements that are expressed as the corresponding primitive idempotents plus possible nilpotent elements . the second line comprises @xmath351 elements in the radical , and the bottom line @xmath352 elements in the radical with zero pairwise products . the elements of this basis are called the _ `` radford '' central elements _ in what follows . they are a `` mixing '' of the canonical central elements in * * within each subalgebra in , and are used in decompositions in what follows instead of the canonical central elements in * * . certain linear combinations of the radford central elements are extensively used in what follows . for the convenience of the reader , we summarize the notation for these combinations here . we define @xmath353 with @xmath354 as @xmath355 ( as we see in what follows , these form a basis in the linear span of the drinfeld - map image of projective - module characters . we also set @xmath356 next , we define @xmath357 with @xmath259 as @xmath358 we also define @xmath359 for @xmath259 as @xmath360 in this section , we study the drinfeld map and a ribbon structure for @xmath45 . we find the necessary ingredient for the drinfeld map ( the @xmath77-matrix ) in * * . as a preparation for the study of the @xmath24-action , we then evaluate the drinfeld images of the @xmath247 basis in @xmath92 . together with the `` radford '' basis constructed in sec . [ sec : radford ] , this gives two bases in the center that play a crucial role in the study of the modular group action in sec . [ sec : sl2z ] . we consider the drinfeld map in * * and find a ribbon element in * * . as a short detour in * * , we characterise the grothendieck ring in terms of chebyshev polynomials . for a quasitriangular hopf algebra @xmath201 with the universal @xmath75-matrix @xmath75 , the @xmath77-matrix is defined as @xmath361it satisfies the relations @xmath362 and @xmath363 @xmath364 . if in addition @xmath77 can be represented as@xmath365 where @xmath366 and @xmath367 are two _ bases _ in @xmath201 , the hopf algebra @xmath201 is called _ factorizable_. for a hopf algebra @xmath201 with an @xmath77-matrix , the drinfeld map @xmath368 is defined by @xmath369 in a factorizable hopf algebra @xmath201 , the restriction of the drinfeld map to the space @xmath92 of @xmath18-characters gives an isomorphism @xmath370 of associative algebras @xcite . as any drinfeld double , the double @xmath70 whose construction was outlined in * * ( see @xcite for the detailed calculation ) is endowed with a universal @xmath75-matrix . to evaluate it explicitly , we construct the elements of the basis dual to the pbw basis in @xmath69 , eqs . , in terms of the generators introduced in and thus find @xmath371_+![n]_-!}$}}\\[-3pt ] { } \times { \mathfrak{q}_+}^{-mj - p_-\frac{m(m-1)}{2 } } { \mathfrak{q}_-}^{nj - p_+\frac{n(n-1)}{2}}{\mathfrak{q}}^{-cj } \,{k}^j{e_+}^m{f_-}^n{\otimes}{f_+}^m{e_-}^n{\kappa}^c.\end{gathered}\ ] ] the @xmath77-matrix for @xmath70 is @xmath372 with this @xmath75 . by simple calculation , it then follows that the @xmath77-matrix @xmath373 for @xmath374 ( see ) is given by @xmath375_+![m']_-![n]_+![n']_-!}$}}\\[-3pt ] { } \times { \mathfrak{q}_+}^{(j - j ' ) m + p_-\frac{m(m+1)}{2 } - p_-\frac{n(n-1)}{2 } } { \mathfrak{q}_-}^{(j ' - j ) m ' + p_+\frac{m'(m'+1)}{2 } - p_+\frac{n'(n'-1)}{2 } } { \mathfrak{q}}^{2 j j'}\\[-3pt ] { } \times { f_+}^{n}{e_+}^{m}{e_-}^{n'}{f_-}^{m'}{k}^{2j}{\otimes}{e_+}^{n}{f_+}^m{f_-}^{n'}{e_-}^{m'}{k}^{2j'},\end{gathered}\ ] ] which involves only even powers of @xmath376 , and therefore @xmath377 . to repeat this in words , the attempted calculation of the @xmath77-matrix for @xmath374 has yielded the @xmath77-matrix for its @xmath45 subalgerba ( and hence @xmath374 is not a factorizable quantum group ; as noted in the introduction , the relevant object is the pair @xmath378 of quantum groups one of which has an @xmath77-matrix and the other an @xmath75-matrix).-matrix acting on two vertex operators describes their braiding , while the @xmath77-matrix describes its `` square , '' the operation of taking one of the operators _ around _ the other . the braiding requires choosing one of the two possible directions , which corresponds to choosing how @xmath379 acts on the representations . ] we next use @xmath373 to construct the drinfeld map for @xmath45 . in considering the drinfeld map , we first restrict it to the image of the grothendieck ring in the space @xmath92 of @xmath18-characters under the map @xmath380 ( see ) , that is , consider the map @xmath381 _ which is a homomorphism of associative commutative algebras _ for any factorizable ribbon hopf algebra @xmath201 . with the balancing element for @xmath211 given by , eq . becomes @xmath382 where @xmath64 , @xmath383 , and @xmath384 . clearly , @xmath385 . the other @xmath386 are evaluated as follows . [ prop - eval ] for @xmath64 and @xmath383 , @xmath387 where @xmath388{(-1)^{r-1}\sum_{a=0}^{r-1}\sum_{m=0}^{a } } ( { \mathfrak{q}}_{\pm}^{p_\mp}\!- { \mathfrak{q}}_{\pm}^{-p_\mp})^{2 m } { \mathfrak{q}}_{\pm}^{p_\mp m ( m + r - 2a ) + p_\mp(r-1 - 2a)}\\ * { } \times{{\mathchoice { { { \mbox{\footnotesize$\displaystyle \genfrac{[}{]}{0pt}{}{r\!-\!a\!+\!m\!-\!1}{m}$}}}}{{\genfrac{[}{]}{0pt}{}{r\!-\!a\!+\!m\!-\!1}{m } } } { { \genfrac{[}{]}{0pt}{}{r\!-\!a\!+\!m\!-\!1}{m}}}{{\genfrac{[}{]}{0pt}{}{r\!-\!a\!+\!m\!-\!1}{m}}}}}_\pm { { \mathchoice { { { \mbox{\footnotesize$\displaystyle \genfrac{[}{]}{0pt}{}{a}{m}$}}}}{{\genfrac{[}{]}{0pt}{}{a}{m } } } { { \genfrac{[}{]}{0pt}{}{a}{m}}}{{\genfrac{[}{]}{0pt}{}{a}{m}}}}}_\pm e_{\pm}^{m } f_{\pm}^m k_\pm^{-m - r+1 + 2a}. \end{gathered}\ ] ] moreover , for @xmath64 , @xmath383 , and @xmath384 , @xmath389 decomposes in terms of the radford central elements as recall the notation in @xmath390 where @xmath391 that is , @xmath392 for @xmath393 and @xmath394 for @xmath395 . because is an algebra homomorphism , the product @xmath396 is expressed through the @xmath397 just by the formulas in * * with @xmath398 and @xmath399 ( cf . @xcite ) . the proof of is by straightforward calculation based on and . using , we then find that @xmath400_+![m']_-!)^2}\\ * { } \times{{\mathchoice { { { \mbox{\footnotesize$\displaystyle \genfrac{[}{]}{0pt}{}{r\!-\!a\!+\!m\!-\!1}{m}$}}}}{{\genfrac{[}{]}{0pt}{}{r\!-\!a\!+\!m\!-\!1}{m } } } { { \genfrac{[}{]}{0pt}{}{r\!-\!a\!+\!m\!-\!1}{m}}}{{\genfrac{[}{]}{0pt}{}{r\!-\!a\!+\!m\!-\!1}{m}}}}}_+ { { \mathchoice { { { \mbox{\footnotesize$\displaystyle \genfrac{[}{]}{0pt}{}{a}{m}$}}}}{{\genfrac{[}{]}{0pt}{}{a}{m } } } { { \genfrac{[}{]}{0pt}{}{a}{m}}}{{\genfrac{[}{]}{0pt}{}{a}{m}}}}}_+ { { \mathchoice { { { \mbox{\footnotesize$\displaystyle \genfrac{[}{]}{0pt}{}{r'\!-\!a'\!-\!1}{m'}$}}}}{{\genfrac{[}{]}{0pt}{}{r'\!-\!a'\!-\!1}{m ' } } } { { \genfrac{[}{]}{0pt}{}{r'\!-\!a'\!-\!1}{m'}}}{{\genfrac{[}{]}{0pt}{}{r'\!-\!a'\!-\!1}{m'}}}}}_- { { \mathchoice { { { \mbox{\footnotesize$\displaystyle \genfrac{[}{]}{0pt}{}{a'\!+\!m'}{m'}$}}}}{{\genfrac{[}{]}{0pt}{}{a'\!+\!m'}{m ' } } } { { \genfrac{[}{]}{0pt}{}{a'\!+\!m'}{m'}}}{{\genfrac{[}{]}{0pt}{}{a'\!+\!m'}{m'}}}}}_- { \mathfrak{q}_+}^{(r-1 - 2a)(j + p_- ) } { \mathfrak{q}_-}^{(r'-1 - 2a')(j - p_+)},\end{gathered}\ ] ] which after some rearrangements gives in accordance with for the @xmath77-matrix in . the decomposition then follows by calculating the action on projective modules , as explained in * * . [ drinfeld - factors ] unlike the @xmath77-matrix , the drinfeld map in factors into a product of the drinfeld maps for the two @xmath47 quantum groups with @xmath48 . the corresponding multiplication formulas are known from @xcite : @xmath401{\sum_{\substack{r''=\|r - r'\| + 1\\ \mathrm{step}=2}}^{r + r ' - 1 } } { \widetilde}{\boldsymbol{\chi}}_{(\pm)}(r''),\ ] ] where @xmath402 we recall that @xmath403 in @xmath45 . naturally , relations become the corresponding formulas in @xcite in the cases where either @xmath404 or @xmath405 . we next evaluate the drinfeld map on the @xmath406 , @xmath407 , and @xmath408 in . this gives @xmath409 @xmath410 for all @xmath411 , @xmath339 , and @xmath412 for @xmath225 . [ prop : skew - drinfeld - decompose ] the central elements @xmath413 , @xmath414 , and @xmath415 decompose in terms of the radford central elements as @xmath416 @xmath417}_+ { \widehat{\pmb{\rho}}{}^{\,\boxslash}{}}(s , p_- ) \\ * { } + 2\!\!\!\sum_{(s , s')\in{\mathcal{i}_1}}\!\!\ ! ( -1)^{(r+p_+ ) s ' + p_- s } [ r s]_+ { \widehat{\pmb{\rho}}{}^{\,\boxslash}{}}(s , s ' ) \!\biggr ) , \end{gathered}\ ] ] @xmath418 @xmath419}_- { \widehat{\pmb{\rho}}{}^{\,\boxbslash}{}}(p_+ , s ' ) \\ * { } + 2\!\!\!\sum_{(s , s')\in{\mathcal{i}_1}}\!\!\ ! ( -1)^{(r'+p_- ) s + p_+ s ' } [ r ' s']_-\ , { \widehat{\pmb{\rho}}{}^{\,\boxbslash}{}}(s , s ' ) \!\biggr ) , \end{gathered}\ ] ] @xmath420 where we use the notation in * * and set @xmath421 for @xmath422 . a calculation similar to the one in the proof of * * shows that the @xmath413 , @xmath414 , and @xmath415 are explicitly given by @xmath423 for @xmath225 , where @xmath424 are defined in and @xmath425_{\pm } \sum_{a=0}^{r-1}\sum_{m=0}^{p_\pm-1 } { \mbox{\footnotesize$\displaystyle\frac{({\mathfrak{q}}_{\pm}^{p_\mp}\!- { \mathfrak{q}}_{\pm}^{-p_\mp})^{2m-1 } } { ( [ m]_\pm!)^2}$}}\ , { \mathfrak{q}}_{\pm}^{p_\mp m(m+r-2a ) + p_\mp(r-1 - 2a ) } \\ * { } \times [ x^1]\bigl({\mathscr{c}_{\pm , r , a}}^{m}(x)\bigr ) e_{\pm}^{m } f_{\pm}^m k^{-p_\mp ( m+r-1 - 2a ) } , \end{gathered}\ ] ] where the polynomials @xmath426 are defined in . the rest is shown by evaluation of the action of @xmath427 , @xmath428 , and @xmath429 ( that is , of @xmath430 and @xmath431 ) on projective modules , see * * . because of the drinfeld map factorization ( cf . * * ) , the demonstration largely reduces to separate calculations in each of the two @xmath47 quantum groups with @xmath48 , cf . @xcite . we next characterize the grothendieck ring as follows . let @xmath432 be the chebyshev polynomials of the second kind . they satisfy ( and are determined by ) the recursive relation @xmath433 with the initial data @xmath434 , @xmath435 . [ prop : cheb ] the grothendieck ring @xmath436 of @xmath45 is the quotient of @xmath68 $ ] over the ideal generated by the polynomials @xmath437 moreover , the basis of irreducible representations @xmath149 is given by the image under the quotient mapping of the respective polynomials @xmath438 , @xmath64 , @xmath65 , where @xmath439 to show * * , we first of all recall , once again , that the drinfeld map @xmath85 is an algebra isomorphism . next , it follows from that the grothendieck ring is _ generated _ by the elements @xmath440 ( see and ) , which , moreover , satisfy eqs . , where we can rewrite @xmath441 as @xmath442 and the equation @xmath443 ( where both sides are actually equal to @xmath444 ) . the statement in * * is now immediate . we recall that a _ ribbon hopf algebra _ @xcite is a quasitriangular hopf algebra equipped with an invertible central element @xmath90 , called the _ ribbon element _ , such that @xmath445 in a ribbon hopf algebra , @xmath446 and , in fact , @xmath447 where @xmath215 is the balancing element and @xmath448 ( where @xmath449 ) is the _ canonical element _ @xcite . [ prop : ribbon ] the @xmath45-ribbon element @xmath90 is given by @xmath450 where @xmath451 with @xmath452 and @xmath94 being the primitive idempotents in the center , and @xmath453 first , using the @xmath75-matrix in , we explicitly calculate the canonical element @xmath454 in @xmath374 as @xmath455_+ ! [ n]_-!}}$}}\ , { \mathfrak{q}_+}^{p_-\left(\!\frac{m(m+3)}{2 } - \frac{r^2}{2}\right ) } { \mathfrak{q}_-}^{p_+\left(\!\frac{n(n+3)}{2 } - \frac{s^2}{2}\right)}\\ * { } \times(-1)^{r s } \bigl({f_+}^{m } k_+^{r - m } { e_+}^{m}\bigr ) \bigl({\boldsymbol{1}}+ ( -1)^{p_+ s + p_- r}(-i)^{p_+ p_- } k^{p_+ p_-}\bigr ) \bigl({e_-}^{n } k_-^{s + n } { f_-}^{n}\bigr ) . \end{gathered}\ ] ] the calculation involved the simple gaussian sum @xmath456 the expression obtained shows , in particular , that @xmath454 is an element of @xmath45 , not of @xmath374 . we then find the ribbon element in accordance with and , as @xmath457 we next decompose the ribbon element @xmath90 with respect to the basis of the radford central elements . first , with the explicit expression for @xmath90 , we find its action on the irreducible representations @xmath458 . an elementary calculation shows that the action of the ribbon element on irreducible modules is given by @xmath459 thus associating @xmath460 with each pair @xmath354 . with some more work ( as for other decompositions obtained above ; also cf . @xcite ) , we calculate the nondiagonal part of the ribbon element action on projective modules and thus find the coefficients in the decomposition of the ribbon element with respect to the canonical central elements , as explained in * * . with the notation in * * , we have @xmath461 the formulas in * * of course allow expressing this through the radford - map images , similarly to other decompositions of central elments , but the above form involving primitive idempotents shows that @xmath90 factors into semisimple and unipotent ( one - plus - nilpotent ) parts as @xmath462 , with @xmath463 just as in the proposition and @xmath464 given by @xmath465 plus all the terms in except the first sum , with all the factors @xmath466 replaced by unity . the formulas in * * then make it easy to verify that @xmath467 whence the formula in the proposition follows immediately . the aim of this section is to prove theorem * * . we define the @xmath24-action on the @xmath45 center @xmath23 as @xmath468 which follows @xcite with insignificant variations . here , @xmath469 which is the central charge of the @xmath30 conformal field theory model . we have two `` distinguished '' bases in the center @xmath23 of @xmath45 , associated with the radford and the drinfeld maps . the `` radford '' basis is given in . the `` drinfeld '' basis consists of the @xmath470@xmath471 elements @xmath472{l } 1{\,{\leqslant}\,}r{\,{\leqslant}\,}p_+\!-\!1,\\ 1{\,{\leqslant}\,}r'{\,{\leqslant}\,}p_- , \end{array}\qquad { { \boldsymbol{\chi}}^{{{\scriptscriptstyle\boldsymbol{\nearrow\!\!\!\swarrow}}}}}(r , r'),\ \begin{array}[t]{l } 1{\,{\leqslant}\,}r{\,{\leqslant}\,}p_+,\\ 1{\,{\leqslant}\,}r'{\,{\leqslant}\,}p_-\!-\!1 , \end{array } \\ & { { \boldsymbol{\chi}}^{{{\scriptstyle\boldsymbol{\downdownarrows}}}}}(r , r'),\quad ( r , r')\in{\mathcal{i}_1}.\end{aligned}\ ] ] from the definitions , the @xmath91 mapping relates the two bases as @xmath473 in accordance with @xcite , @xmath474 which , clearly , requires a proper choice of normalization in our formulas , most easily checked by calculating @xmath475 . ( as we have noted , this fixes the value of @xmath476 in . ) we now prove the most tedious part of theorem * * , the structure of the @xmath24 representation on @xmath23 in claim [ claim : decomp ] . the proof amounts to a series of lemmas . [ lemma : r - min ] for @xmath225 , the elements @xmath477 see the notation in span a @xmath478-dimensional @xmath24-subrepresentation in @xmath23 . moreover , this representation is isomorphic to @xmath37 in . representation in the minimal @xmath30 conformal field theory model . ] first , the formulas @xmath479 ( which are a direct consequence of * * and ) show that @xmath480 second , we find @xmath481 from eqs . and , which imply @xmath482 where we also used . with a similar expression for @xmath483 , we have @xmath484 the lemma is proved . we once again recall the notation in and set @xmath485 [ lemma : triplet ] the elements @xmath486 , @xmath487 , and @xmath488 with @xmath225 span a @xmath489-dimensional @xmath24-subrepresentation @xmath490 in @xmath23 and , moreover , @xmath491 where @xmath36 is the symmetrized square of the standard two - dimensional representation and @xmath37 is the representation in . clearly , @xmath492 from and * * . we next evaluate the action of the ribbon element on @xmath486 using * * , which gives @xmath493 next , a similar calculation yields @xmath494 as regards the @xmath91 action , we recall and , which we rewrite as @xmath495 next , repeating the calculation in the proof of * * , we obtain @xmath496 this completes the proof . [ lemma : lsh ] for @xmath259 , the elements @xmath357 see and and @xmath359 see and span a @xmath497-dimensional @xmath24-subrepresentation @xmath498 in @xmath23 . moreover , its structure is @xmath499 where @xmath35 is the standard two - dimensional @xmath24 representation and @xmath41 is the representation in . we first calculate the ribbon action . from and * * , we have @xmath500 \intertext{for $ ( r , r')\in{\mathcal{i}_1}$ , and } { { \boldsymbol{v}}}{{{\widehat}{\pmb{\varphi}}{}}^{\boxslash}{}}(r , p_- ) & = e^{2i\pi\delta_{r , p_- } } { { { \widehat}{\pmb{\varphi}}{}}^{\boxslash}{}}(r , p_- ) + e^{2i\pi\delta_{r , p_-}}{\widehat{\pmb{\rho}}{}^{\,\boxslash}{}}(r , p_- ) . \end{aligned}\ ] ] from and , evidently , @xmath501 for all the @xmath502 elements in the lemma . next , the @xmath91-action is calculated as @xmath503 where @xmath225 and we used , and , similarly , @xmath504 conversely , from we have @xmath505 which completes the proof . totally similarly , we establish the following lemma . [ lemma : rsh ] for @xmath261 , the elements @xmath506 see and and @xmath507 see and span a @xmath508-dimensional @xmath24-subrepresentation @xmath509 in @xmath23 . moreover , its structure is @xmath510 where @xmath35 is the standard two - dimensional @xmath24 representation and @xmath40 is the representation in . to formulate our next ( and final ) lemma , we recall the notation in . [ lemma - proj ] the elements @xmath353 , @xmath354 , span a @xmath511-dimensional @xmath24-subrepresentation @xmath39 in @xmath23 . clearly , for @xmath354 , the ribbon element acts as @xmath512 as regards the @xmath91 action , we recall and , which we rewrite , for @xmath225 and using notation , as @xmath513 and similarly for @xmath514 , @xmath515 , and @xmath516 . the lemma is proved . this completes the proof of claim [ claim : decomp ] in * * . [ [ section-3 ] ] claim [ claim : grring ] is now obvious if we recall from that the drinfeld - map image of the grothendieck ring in @xmath23 is spanned by @xmath517 claim [ claim : equiv ] in * * is established by direct comparison with formulas in ( * ? ? ? * lemmas 4.3 and 4.4 ) . first , with * * , we define @xmath518 then _ the @xmath519 transform under @xmath91 and @xmath520 in as the functions @xmath521 in @xcite transform under @xmath522 and @xmath523 . _ ( in particular , @xmath524 . ) next , with * * , we define @xmath525 where @xmath225 . then _ @xmath526 , @xmath527 , and @xmath528 transform under @xmath91 and @xmath520 in as the respective functions @xmath529 , @xmath530 , and @xmath531 in @xcite transform under @xmath522 and @xmath523 . _ ( in particular , @xmath532 and @xmath533 . ) similarly , with * * , we set @xmath534 then _ the @xmath535 and @xmath536 thus defined transform under @xmath91 and @xmath520 in as the respective functions @xmath537 and @xmath538 in @xcite transform under @xmath522 and @xmath523 . _ we give some more details here , mainly because of a minor difference between the conventions in this paper and in @xcite . first , from the formulas in the proof of * * , it follows that @xmath539 for all the @xmath540 defined above , which coincides with the @xmath91-transformation formula for @xmath541 found in @xcite . next , from the last formula in the proof of * * and formulas , it follows that @xmath542 for @xmath225 , which is also the formula for @xmath543 in @xcite . finally , it follows from the definitions ( see ) that @xmath544 . using now gives @xmath545 which readily shows that @xmath546 is given by the above formula for @xmath547 at @xmath548 ( as in @xcite ) . it next follows from the ribbon element action in * * that @xmath549 for all @xmath550 defined above ( for @xmath551 , the simple derivation involves the identity @xmath552 , see ) . equally easily , * * also implies that @xmath553 for all @xmath554 defined above . the correspondence involving the central elements in * * is totally similar . finally , for the central elements in * * , the correspondence is also established straightforwardly , and we omit the details . we next consider factorization claim [ claim : factorization ] in * * . as in a simpler case studied in @xcite , this factorization is related to the factorization of the ribbon element in . we first note that with the @xmath77-matrix expressed as @xmath555 , it readily follows @xcite that @xmath556 we now define a map @xmath557 , which is an isomorphism of vector spaces and intertwines the coadjoint and adjoint actions of @xmath45 , as @xmath558 where @xmath559 from the @xmath91-transformation formulas in * * , we immediately obtain @xmath560 the maps @xmath561 given by @xmath562 and @xmath563 then provide a factorization of @xmath564 , @xmath565 the @xmath24-representation on the center is thus factored into two representations , generated by @xmath566 and @xmath567 . the representation generated by @xmath566 can be further factored in accordance with the factorization of @xmath464 as @xmath568 where @xmath569 { \boldsymbol{1}}+ { \mbox{\footnotesize$\displaystyle\frac{1}{p_-}$}}\ , { { \boldsymbol{\chi}}^{{{\scriptscriptstyle\boldsymbol{\nearrow\!\!\!\swarrow}}}}}(1 , 1 ) \end{cases}\\ = { } & { \boldsymbol{1}}- \sum_{m=1}^{p_\pm - 1}\sum_{a = m-1}^{p_\pm - 1}\ ! ( -1)^{m } { \mbox{\footnotesize$\displaystyle\frac{({\mathfrak{q}}_{\pm}^{p_\mp}\!- { \mathfrak{q}}_{\pm}^{-p_\mp})^{2m-1}}{p_\pm [ m]_\pm}$}}\ , { \mathfrak{q}}_{\pm}^{p_\mp m(m - 1 - 2a ) - p_\mp ( 2 + 2a ) } \\ * & \qquad\qquad\qquad{}\times { { \mathchoice { { { \mbox{\footnotesize$\displaystyle \genfrac{[}{]}{0pt}{}{a}{m\!-\!1}$}}}}{{\genfrac{[}{]}{0pt}{}{a}{m\!-\!1 } } } { { \genfrac{[}{]}{0pt}{}{a}{m\!-\!1}}}{{\genfrac{[}{]}{0pt}{}{a}{m\!-\!1}}}}}_\pm^2 e_{\pm}^{m } f_{\pm}^m k^{-p_\mp ( m - 2 - 2a)}\end{aligned}\ ] ] and we calculated @xmath570 and @xmath571 from and using and . this leads to a further factorization of the @xmath24-action on the center , constructed similarly to . a remarkable correspondence between modular group actions in nonsemisimple ( logarithmic ) conformal field theory models and on quantum groups deserves further study . another quite interesting problem is to extend the category of finite - dimensional @xmath45-rerpesentations such that it becomes _ equivalent _ to the category of the @xmath31 algebra representations realized in logarithmic @xmath30 conformal field theory models . we are grateful to a. belavin and s. parkhomenko for the useful discussions . we thank g. mutafian for help at the early stages of this work . this paper was supported in part by the rfbr grant 04 - 01 - 00303 and the rfbr jsps grant 05 - 01 - 02934yaf_a . ams was also supported in part by the rfbr grant 05 - 01 - 00996 . part of the paper was written when the authors were visiting kyoto university , and we are grateful to t. miwa for hospitality . as a preparation for the description of projective modules , we introduce modules @xmath573 and @xmath574 , where @xmath384 , with the respective structure of subquotients described as @xmath575{1pproj.eps } \ ] ] for @xmath172 and @xmath65 , the @xmath576 module has the basis @xmath577 where @xmath578 is the basis corresponding to the top @xmath149 module in , @xmath579 to the bottom @xmath149 , @xmath580 to the right @xmath581 , and @xmath582 to the left @xmath581 module . we fix the arbitrariness involved in choosing @xmath583 , @xmath584 , and @xmath585 by specifying the @xmath45-action in @xmath573 as @xmath586 @xmath587 @xmath588}_+{[r\!-\!n]}_+{\mathsf{t}^{\scriptstyle\uparrow}}_{n-1,n'}+{\mathsf{t}^{\scriptstyle\downarrow}}_{n-1,n ' } , & 1{\,{\leqslant}\,}n{\,{\leqslant}\,}r-1,\\ { \mathsf{t}^{\scriptstyle\leftarrow}}_{p_+\!-\!r\!-\!1,n ' } , & n=0,\\ \end{cases } \notag\\ { e_+}{\mathsf{t}^{\scriptstyle\rightarrow}}_{k , k'}&= \begin{cases } ( -\alpha)^{p_-}(-1)^{r'-1 } { [ k]}_+{[p_+\!-\!r\!-\!k]}_+{\mathsf{t}^{\scriptstyle\rightarrow}}_{k-1,k ' } , & 1{\,{\leqslant}\,}k{\,{\leqslant}\,}p_+\!-\!r\!-\!1,\\ { \mathsf{t}^{\scriptstyle\downarrow}}_{r-1,k ' } , & k=0,\\ \end{cases } \notag\\ { e_+}{\mathsf{t}^{\scriptstyle\leftarrow}}_{k , k ' } & = ( -\alpha)^{p_-}(-1)^{r'-1}{[k]}_+{[p_+\!-\!r\!-\!k]}_+ { \mathsf{t}^{\scriptstyle\leftarrow}}_{k-1,k ' } , \ ; \begin{array}[t]{ll } 0{\,{\leqslant}\,}k{\,{\leqslant}\,}p_+\!-\!r\!-\!1\\ ( \text{with}\;{\mathsf{t}^{\scriptstyle\leftarrow}}_{-1,k'}\equiv0 ) , \end{array } \\ { e_+}{\mathsf{t}^{\scriptstyle\downarrow}}_{n , n ' } & = ( \alpha)^{p_-}(-1)^{r'-1}{[n]}_+{[r\!-\!n]}_+ { \mathsf{t}^{\scriptstyle\downarrow}}_{n-1,n ' } , \quad 0{\,{\leqslant}\,}n{\,{\leqslant}\,}r\!-\!1 \quad(\text{with}\;{\mathsf{t}^{\scriptstyle\downarrow}}_{-1,n'}\equiv0),\notag\\ { f_+}{\mathsf{t}^{\scriptstyle\uparrow}}_{n , n'}&= \begin{cases } { \mathsf{t}^{\scriptstyle\uparrow}}_{n+1,n ' } , & 0{\,{\leqslant}\,}n{\,{\leqslant}\,}r\!-\!2,\\ { \mathsf{t}^{\scriptstyle\rightarrow}}_{0,n ' } , & n = r\!-\!1 , \end{cases}\quad { f_+}{\mathsf{t}^{\scriptstyle\leftarrow}}_{k , k'}= \begin{cases } { \mathsf{t}^{\scriptstyle\leftarrow}}_{k+1,k ' } , & 0{\,{\leqslant}\,}k{\,{\leqslant}\,}p_+\!-\!r\!-\!2,\\ { \mathsf{t}^{\scriptstyle\downarrow}}_{0,k ' } , & k = p_+\!-\!r\!-\!1,\\ \end{cases } \notag\\ { f_+}{\mathsf{t}^{\scriptstyle\rightarrow}}_{k , k'}&={\mathsf{t}^{\scriptstyle\rightarrow}}_{k+1,k ' } , \quad 0 { \,{\leqslant}\,}k{\,{\leqslant}\,}p_+\!-\!r\!-\!1 \quad(\text{with}\;{\mathsf{t}^{\scriptstyle\rightarrow}}_{p_+-r , k'}\equiv0),\notag\\ { f_+}{\mathsf{t}^{\scriptstyle\downarrow}}_{n , n'}&={\mathsf{t}^{\scriptstyle\downarrow}}_{n+1,n ' } , \quad 0{\,{\leqslant}\,}n{\,{\leqslant}\,}r\!-\!1\quad(\text{with}\;{\mathsf{t}^{\scriptstyle\downarrow}}_{r , n'}\equiv0),\notag\end{aligned}\ ] ] where @xmath589 , and @xmath590}_- { [ r'\!-\!n']}_-{\mathsf{t}^{\scriptstyle\uparrow}}_{n , n'-1 } \quad(\text{with}\;{\mathsf{t}^{\scriptstyle\uparrow}}_{n,-1}\equiv0),&\\ { e_-}{\mathsf{t}^{\scriptstyle\rightarrow}}_{k , k ' } & = ( -\alpha)^{p_+}(-1)^{p_+-r-1 } { [ k']}_-{[r'\!-\!k']}_-{\mathsf{t}^{\scriptstyle\rightarrow}}_{k , k'-1 } \quad(\text{with}\;{\mathsf{t}^{\scriptstyle\rightarrow}}_{k,-1}\equiv0),&\\ { e_-}{\mathsf{t}^{\scriptstyle\leftarrow}}_{k , k ' } & = ( -\alpha)^{p_+}(-1)^{p_+-r-1}{[k']}_- { [ r'\!-\!k']}_-{\mathsf{t}^{\scriptstyle\leftarrow}}_{k , k'-1 } \quad(\text{with}\;{\mathsf{t}^{\scriptstyle\leftarrow}}_{k,-1}\equiv0),&\\ { e_-}{\mathsf{t}^{\scriptstyle\downarrow}}_{n , n ' } & = ( \alpha)^{p_+}(-1)^{r-1}{[n']}_- { [ r'\!-\!n']}_-{\mathsf{t}^{\scriptstyle\downarrow}}_{n , n'-1 } \quad(\text{with}\;{\mathsf{t}^{\scriptstyle\downarrow}}_{n,-1}\equiv0),&\\ { f_-}{\mathsf{t}^{\scriptstyle\uparrow}}_{n , n ' } & = { \mathsf{t}^{\scriptstyle\uparrow}}_{n , n'+1}\quad(\text{with}\;{\mathsf{t}^{\scriptstyle\uparrow}}_{n , r'}\equiv0),\quad { f_-}{\mathsf{t}^{\scriptstyle\rightarrow}}_{k , k ' } = { \mathsf{t}^{\scriptstyle\rightarrow}}_{k , k'+1}\quad(\text{with}\;{\mathsf{t}^{\scriptstyle\rightarrow}}_{k , r'}\equiv0),\\ { f_-}{\mathsf{t}^{\scriptstyle\leftarrow}}_{k , k ' } & = { \mathsf{t}^{\scriptstyle\leftarrow}}_{k , k'+1}\quad(\text{with}\;{\mathsf{t}^{\scriptstyle\leftarrow}}_{k , r'}\equiv0 ) , \quad { f_-}{\mathsf{t}^{\scriptstyle\downarrow}}_{n , n ' } = { \mathsf{t}^{\scriptstyle\downarrow}}_{n , n'+1}\quad(\text{with}\;{\mathsf{t}^{\scriptstyle\downarrow}}_{n , r'}\equiv0),\end{aligned}\ ] ] where @xmath591 , @xmath592 , and @xmath589 . for @xmath64 and @xmath339 , we similarly define the @xmath593 module . we describe the projective modules @xmath180 for @xmath64 and @xmath65 . first , for @xmath411 and @xmath339 , the projective module @xmath180 is a second extension of the verma module @xmath161 , @xmath594_{{e_- } } \ar@/_/[dr]^{{e_+}}&\\ & { \mathsf{v}}^{\mp}_{r , p_--r ' } \ar@/^/[dr]_{{e_+ } } & & { \mathsf{v}}^{\mp}_{p_+-r , r ' } \ar@/_/[dl]^{{e_-}}\\ & & { \mathsf{v}}^{\pm}_{p_+-r , p_--r ' } & } \ ] ] and is therefore a `` deck '' of sixteen subquotients shown in fig . [ app : proj - pic - plus ] . we note that @xmath595 for @xmath411 and @xmath339 . second , @xmath596 for @xmath411 and @xmath597 for @xmath339 , where the modules @xmath598 and @xmath599 are defined above in * * . third , @xmath600 are irreducible . the module @xmath180 has the basis @xmath601 where @xmath602 , @xmath603 , @xmath604 , @xmath605 if @xmath606 , + @xmath607 if @xmath608 , and @xmath609 is the basis corresponding to the top module @xmath576 , @xmath610 to the bottom @xmath576 , @xmath611 to the right @xmath612 , and @xmath613 to the left module @xmath612 in the diagram @xmath614_{{e_- } } \ar@/_/[dr]^{{f_-}}&\\ & { { \mathscr{p}}^{\mp,+}_{r , p_--r ' } } \ar@/^/[dr]_{{f_- } } & & { { \mathscr{p}}^{\mp,+}_{r , p_--r ' } } \ar@/_/[dl]^{{e_-}}\\ & & { \mathscr{p}}^{\pm,+}_{r , r ' } & } \ ] ] the action of the generators @xmath129 , @xmath116 , and @xmath122 on @xmath609 is defined in and the action of @xmath123 and @xmath117 is given by @xmath615}_- { [ r'\!-\!n']}_-{\mathsf{t}^{\scriptstyle\bullet}}_{n , n'-1}+{\mathsf{b}^{\scriptstyle\bullet}}_{n , n'-1 } , & 1 { \,{\leqslant}\,}n'{\,{\leqslant}\,}r'\!-\!1,\\ { \mathsf{l}^{\scriptstyle\bullet}}_{n , p_--r'-1 } , & n'=0,\\ \end{cases } \\ { f_-}{\mathsf{t}^{\scriptstyle\bullet}}_{n , n'}&= \begin{cases } { \mathsf{t}^{\scriptstyle\bullet}}_{n , n'+1 } , & 0{\,{\leqslant}\,}n'{\,{\leqslant}\,}r'\!-\!2,\\ { \mathsf{r}^{\scriptstyle\bullet}}_{n,0 } , & n'=r'\!-\!1 . \end{cases } \\ \intertext{where $ { { \scriptstyle\bullet}}\in\{{{\scriptstyle\uparrow}},{{\scriptstyle\downarrow}}\}$ and $ 0{\,{\leqslant}\,}n{\,{\leqslant}\,}r-1 $ , } { e_-}{\mathsf{t}^{\scriptstyle\bullet}}_{n , n'}&= \begin{cases } ( \mp1)^{p_+}(-1)^{p_+-r-1}{[n']}_- { [ r'\!-\!n']}_-{\mathsf{t}^{\scriptstyle\bullet}}_{n , n'-1}+{\mathsf{b}^{\scriptstyle\bullet}}_{n , n'-1 } , & 1{\,{\leqslant}\,}n'{\,{\leqslant}\,}r'\!-\!1,\\ { \mathsf{l}^{\scriptstyle\bullet}}_{n , p_--r'-1 } , & n'=0,\\ \end{cases } \\ { f_-}{\mathsf{t}^{\scriptstyle\bullet}}_{n , n'}&= \begin{cases } { \mathsf{t}^{\scriptstyle\bullet}}_{n , n'+1 } , & 0{\,{\leqslant}\,}n'{\,{\leqslant}\,}r'{-}2,\\ { \mathsf{r}^{\scriptstyle\bullet}}_{n,0 } , & n'=r'{-}1 . \end{cases}\end{aligned}\ ] ] where @xmath616 and @xmath617 . the action of the generators @xmath129 , @xmath116 , @xmath122 , and @xmath117 on @xmath618 coincides with the action on @xmath612 ( see * * ) . the action of @xmath123 is given by @xmath619}_-{[p_-\!-\!r'\!-\!k']}_-{\mathsf{r}^{\scriptstyle\bullet}}_{k , k'-1 } , & 1{\,{\leqslant}\,}k'{\,{\leqslant}\,}p_-\!-\!r'\!-\!1,\\ { \mathsf{b}^{\scriptstyle\bullet}}_{k , r'-1 } , & k'=0,\\ \end{cases } \\ \intertext{where $ { { \scriptstyle\bullet}}\in\{{{\scriptstyle\uparrow}},{{\scriptstyle\downarrow}}\}$ and $ 0{\,{\leqslant}\,}k{\,{\leqslant}\,}r-1 $ , } { e_-}{\mathsf{r}^{\scriptstyle\bullet}}_{k , k'}&= \begin{cases } ( \pm1)^{p_+}(-1)^{p_+-r-1 } { [ k']}_-{[p_-\!-\!r'\!-\!k']}_-{\mathsf{r}^{\scriptstyle\bullet}}_{k , k'-1 } , & 1{\,{\leqslant}\,}k'{\,{\leqslant}\,}p_-\!-\!r'\!-\!1,\\ { \mathsf{b}^{\scriptstyle\bullet}}_{k , r'-1 } , & k'=0,\\ \end{cases}\end{aligned}\ ] ] where @xmath616 and @xmath620 . the action of the generators @xmath129 , @xmath116 , @xmath122 , and @xmath123 on @xmath621 coincides with the action on @xmath612 ( see * * ) . the action of @xmath117 is given by @xmath622 where @xmath616 and @xmath592 . the @xmath45-action on @xmath610 is defined in * * . here , we find the center of @xmath45 using the isomorphism between the center and the algebra of _ bimodule _ endomorphisms of the regular representation . in * * , we study decomposition of the regular representation as a @xmath45-bimodule . in * * , we describe the algebra of bimodule endomorphisms of the regular representation . in * * , the canonical central elements are also constructed explicitly in terms of the algebra generators . in * * , we collect several calculations needed in the proofs in this paper . we study the decomposition of the regular representation as a @xmath45-bimodule . for this , we closely follow the strategy proposed in @xcite for @xmath14 models . the starting point is to introduce indecomposable @xmath45-bimodules @xmath623 defined as follows . for @xmath172 and @xmath65 , @xmath624 is composed of eight subquotients with the left @xmath45-action represented as @xmath625{wzhzh1.eps } \ ] ] and the right action as @xmath625{wzhzh2.eps } \ ] ] in these diagrams , as we have noted , it is understood that the @xmath45-action on each subquotient is changed compared with the action on the corresponding irreducible representation in agreement with the arrows connecting a given subquotient with others . for example , @xmath116 maps the highest - weight vector of @xmath626 into the lowest - weight vector in @xmath627 . as a @xmath45-bimodule , the regular representation decomposes as @xmath632{\bigoplus_{(r , r')\in{\mathcal{i}_1}}{{\mathscr}{q}}(r , r')\oplus \bigoplus_{r=1}^{p_+-1}{{\mathscr}{q}}(r , p_-)\oplus \bigoplus_{r'=1}^{p_--1}{{\mathscr}{q}}(p_+,r ' ) \oplus{{\mathscr}{q}}(p_+,p_-)\oplus{{\mathscr}{q}}(0,p_-)},\ ] ] where 1 . @xmath633 and @xmath634 are simple bimodules ; 2 . @xmath635 and @xmath636 for @xmath172 and @xmath173 ; 3 . the bimodules @xmath637 , @xmath172 and @xmath173 , are indecomposable and admit two length-@xmath638 filtrations : @xmath639 where the structure of subquotients is given by @xmath640 and @xmath641 where the structure of subquotients is given by @xmath642 here , we study the algebra of bimodule endomorphisms of the regular representation @xmath643 . simultaneously , we derive the structure of the @xmath45 center because bimodule endomorphisms of the regular representation are in a one - to - one correspondence with elements in the center . clearly , the algebra of bimodule endomorphisms of @xmath643 decomposes , in agreement with , as @xmath644 \bigoplus_{r'=1}^{p_--1}{\mathrm{end}}\bigl({{\mathscr}{q}}(p_+,r')\bigr)\oplus { \mathrm{end}}\bigl({{\mathscr}{q}}(p_+,p_-)\bigr)\oplus{\mathrm{end}}\bigl({{\mathscr}{q}}(0,p_-)\bigr).\end{gathered}\ ] ] for each @xmath637 , there is a bimodule endomorphism @xmath645 that acts as identity on @xmath637 and is zero on @xmath646 unless @xmath647 and @xmath648 . these endomorphisms give rise to @xmath511 primitive idempotents ( also denoted by @xmath94 ) in the center of @xmath45 . 1 . there is a homomorphism @xmath649 ( defined up to a nonzero factor ) whose image is @xmath650 ; in other words , @xmath282 sends the quotient @xmath650 ( see ) into the subbimodule @xmath650 at the bottom of @xmath637 and is zero on @xmath646 unless @xmath647 and @xmath648 . similarly , there is a central element associated with the homomorphism @xmath651 with the image @xmath652 , i.e. , sending the quotient @xmath652 ( see ) into the subbimodule @xmath652 ( and zero on @xmath646 unless @xmath647 and @xmath648 ) . a central element is associated with the homomorphism @xmath653 with the image @xmath654 , i.e. , sending the quotient @xmath654 ( see ) into the subbimodule @xmath654 ( and zero on @xmath646 unless @xmath647 and @xmath648 ) . 4 . similarly , there is a central element associated with the homomorphism @xmath655 with the image @xmath656 , i.e. , sending the quotient @xmath656 ( see ) into the subbimodule @xmath656 ( and zero on @xmath646 unless @xmath647 and @xmath648 ) . there is also a homomorphism @xmath657 ( defined up to a nonzero factor ) whose image is @xmath658 ; in other words , @xmath659 sends the quotient @xmath658 into the subbimodule @xmath658 at the bottom of @xmath637 and is zero on @xmath646 unless @xmath647 and @xmath660 . 6 . similarly , there is a homomorphism @xmath661 whose image is @xmath662 ; in other words , @xmath287 sends the quotient @xmath662 into the subbimodule @xmath662 at the bottom of @xmath637 and is zero on @xmath646 unless @xmath647 and @xmath660 . there is also a homomorphism @xmath663 whose image is @xmath664 ; in other words , @xmath289 sends the quotient @xmath664 into the subbimodule @xmath664 at the bottom of @xmath637 and is zero on @xmath646 unless @xmath647 and @xmath660 . 8 . similarly , there is also a homomorphism @xmath665 whose image is @xmath666 ; in other words , @xmath288 sends the quotient @xmath666 into the subbimodule @xmath666 at the bottom of @xmath637 and is zero on @xmath646 unless @xmath647 and @xmath648 . these maps give @xmath667 nilpotent elements @xmath289 , @xmath659 , @xmath287 , @xmath288 , @xmath284 , @xmath282 , @xmath285 , @xmath668 . from the structure of the maps , it is obvious that relations are satisfied and that all other compositions are zero . next , for each @xmath669 with @xmath172 , there is a homomorphism @xmath670 with the image @xmath671 ; in other words , @xmath275 sends the quotient @xmath671 into the subbimodule @xmath671 at the bottom of @xmath669 and is zero on @xmath672 with . similarly , for each @xmath673 , there is a central element associated with the homomorphism @xmath674 with the image @xmath675 , i.e. , sending the quotient @xmath675 into the subbimodule @xmath675 ( and zero on @xmath672 with @xmath676 ) . next , for each @xmath677 with @xmath678 , there are similar homomorphisms @xmath679 and @xmath680 . any central element @xmath201 can be decomposed in the above central elements as @xmath683 it immediately follows from * * that the coefficient @xmath684 is the eigenvalue of @xmath201 in the irreducible representation @xmath626 . to determine the @xmath685 , @xmath686 , @xmath687 , and @xmath688 coefficients similarly , we fix the normalizations such that in terms of the respective bases described in * * , @xmath689 , @xmath690 , @xmath691 , and @xmath692 act as @xmath693 then the coefficients in are determined from the relation @xmath694 these formulas are used many times in explicit calculations in the text . to explicitly construct the canonical central elements in * * in terms of the @xmath45 generators , we proceed similarly to @xcite . let @xmath354 ( see ) . we set @xmath695 these are roots of the respective polynomials @xmath696 and @xmath697 ( see * * ) . moreover , _ as @xmath243 range over the set @xmath111 , all roots of @xmath696 and @xmath697 occur among the values taken by @xmath698 and @xmath699 respectively _ ( not necessarily once ) . we then define @xmath700 { \mbox{\footnotesize$\displaystyle\frac{\psi_\pm(x)}{(x - \beta_\pm(r , r'))^2}$}}&\text{otherwise } \end{cases}\ ] ] and set @xmath701 and @xmath702 where @xmath703 . then @xmath704 are the @xmath511 primitive idempotents in the center . the constants @xmath705 involved in the normalization factors are easily found explicitly . for this , we recall that the chebyshev polynomials are eigenfunctions of a second - order differential operator , @xmath706 whenever @xmath707 and @xmath708 , i.e. , @xmath709 is a multiplicity-@xmath710 root of @xmath711 in * * , we use and to find @xmath712 and hence @xmath713 also , if @xmath714 , then @xmath715 and if @xmath716 , then @xmath717 to construct the canonical nilpotent elements , we next introduce the operators operators project on the weights ( i.e. , the eigenvalues of @xmath129 ) occurring in the respective irreducible modules.]@xmath718 these are pairwise orthogonal projection operators . we also note that @xmath719 and @xmath720 for @xmath225 ( see ) , we use the above projectors to define the central elements @xmath721 clearly , their nonzero products are those in , with @xmath722 the idempotent @xmath94 acts as identity on each of these @xmath723 elements . this gives @xmath724 central elements , labeled by @xmath225 . in addition , for @xmath411 , there are the central radical elements @xmath725 on which @xmath274 acts as identity ( the total of @xmath726 central elements ) , and for @xmath339 , there are the central radical elements @xmath727 on which @xmath278 acts as identity ( the total of @xmath728 central elements ) . here , we prove about a quarter of the relations in * * by establishing a similar statement for @xmath729 with @xmath730 , in which case the trace actually restricts to the submodule @xmath731 ( see the definition of the modules @xmath732 in * * ) , and which eventually ( after we ensure that @xmath733 ) gives @xmath326 in . we thus temporarily write @xmath326 for @xmath729 as defined in in the case where @xmath734 . it is a @xmath18-character if and only if ( see ) @xmath735),\ ] ] where , as we have just noted , @xmath736 ) = { \mathrm{tr}^{\vphantom{y}}}_{{\mathscr{p}}^{+,+}_{r , r'}\oplus{\mathscr{p}}^{-,+}_{p_+-r , r ' } } ( { { \boldsymbol{g}}}^{-1}x[y,\sigma^{{{\scriptscriptstyle\boldsymbol{\nearrow\!\!\!\swarrow}}}}])\ ] ] it now follows from the formulas in * * that @xmath737=0 $ ] and the other commutation relations in @xmath738 are given by @xmath739{\mathsf{b}^{\scriptstyle\downarrow,{{\scriptstyle\uparrow}}}}_{n , n ' } & = \begin{cases } { { \alpha^{\scriptstyle\uparrow,{{\scriptstyle\uparrow}}}}}{\mathsf{b}^{\scriptstyle\downarrow,{{\scriptstyle\uparrow}}}}_{n-1,n ' } , & 1{\,{\leqslant}\,}n{\,{\leqslant}\,}r\!-\!1,\\ { { \alpha^{\scriptstyle\uparrow,{{\scriptstyle\uparrow}}}}}{\mathsf{b}^{\scriptstyle\leftarrow,{{\scriptstyle\uparrow}}}}_{p_+-r-1,n ' } , & n=0 , \end{cases } \\ [ { e_+},\sigma^{{{\scriptscriptstyle\boldsymbol{\nearrow\!\!\!\swarrow}}}}(r , r')]{\mathsf{b}^{\scriptstyle\uparrow,{{\scriptstyle\uparrow}}}}_{n , n ' } & = \begin{cases } - { { \alpha^{\scriptstyle\uparrow,{{\scriptstyle\uparrow}}}}}{\mathsf{b}^{\scriptstyle\uparrow,{{\scriptstyle\uparrow}}}}_{n-1,n ' } , & 1{\,{\leqslant}\,}n{\,{\leqslant}\,}r\!-\!1,\\ 0,&n=0 , \end{cases } \\ [ { e_+},\sigma^{{{\scriptscriptstyle\boldsymbol{\nearrow\!\!\!\swarrow}}}}(r , r')]{\mathsf{b}^{\scriptstyle\rightarrow,{{\scriptstyle\uparrow}}}}_{k , k ' } & = \begin{cases } 0 , & 1{\,{\leqslant}\,}k{\,{\leqslant}\,}p_+\!-\!r\!-\!1,\\ - { { \alpha^{\scriptstyle\uparrow,{{\scriptstyle\uparrow}}}}}{\mathsf{b}^{\scriptstyle\uparrow,{{\scriptstyle\uparrow}}}}_{r-1,k ' } , & k=0 , \end{cases } \\ [ { e_+},\sigma^{{{\scriptscriptstyle\boldsymbol{\nearrow\!\!\!\swarrow}}}}(r , r')]{\mathsf{b}^{\scriptstyle\leftarrow,{{\scriptstyle\uparrow}}}}_{k , k ' } & = 0 , \quad 1{\,{\leqslant}\,}k{\,{\leqslant}\,}p_+\!-\!r\!-\!1.\end{aligned}\ ] ] writing the trace operation as @xmath740 we then have @xmath741)= { } { { \alpha^{\scriptstyle\uparrow,{{\scriptstyle\uparrow}}}}}\sum_{n'=0}^{r'-1 } { \langle{\mathsf{b}^{\scriptstyle\downarrow,{{\scriptstyle\uparrow}}}}_{0,n'}\|}{{\boldsymbol{g}}}^{-1}x { \|{\mathsf{b}^{\scriptstyle\leftarrow,{{\scriptstyle\uparrow}}}}_{p_+-r-1,n'}\rangle } -{{\alpha^{\scriptstyle\uparrow,{{\scriptstyle\uparrow}}}}}\sum_{n'=0}^{r'-1 } { \langle{\mathsf{b}^{\scriptstyle\rightarrow,{{\scriptstyle\uparrow}}}}_{0,n'}\|}{{\boldsymbol{g}}}^{-1}x { \|{\mathsf{b}^{\scriptstyle\uparrow,{{\scriptstyle\uparrow}}}}_{r-1,n'}\rangle}\ ] ] and , similarly , @xmath742 ) = { } \\ { } = { { \alpha^{\scriptstyle\uparrow,{{\scriptstyle\rightarrow}}}}}\sum_{n'=0}^{r'-1 } { \langle{\mathsf{b}^{\scriptstyle\downarrow,{{\scriptstyle\rightarrow}}}}_{0,n'}\|}{{\boldsymbol{g}}}^{-1}x { \|{\mathsf{b}^{\scriptstyle\leftarrow,{{\scriptstyle\rightarrow}}}}_{r-1,n'}\rangle } -{{\alpha^{\scriptstyle\uparrow,{{\scriptstyle\rightarrow}}}}}\sum_{n'=0}^{r'-1 } { \langle{\mathsf{b}^{\scriptstyle\rightarrow,{{\scriptstyle\rightarrow}}}}_{0,n'}\|}{{\boldsymbol{g}}}^{-1}x { \|{\mathsf{b}^{\scriptstyle\uparrow,{{\scriptstyle\rightarrow}}}}_{p_+-r-1,n'}\rangle}.\end{gathered}\ ] ] summing these two contributions of the modules @xmath738 and @xmath743 and using the obvious correspondence between the bases in the respective composition factors in @xmath738 and @xmath743 , we immediately obtain that the trace in with @xmath744 vanishes if and only if @xmath745 . similar calculations for the other generators ( @xmath122 , @xmath123 , and @xmath117 ) give no more conditions for @xmath326 to be a @xmath18-character . this completes the proof in the simpler case of @xmath406 . the case of @xmath408 is equally uneventful but somewhat lengthier . in the calculations , we often use the well - known quantum-@xmath746 identity ( see , e.g. , @xcite ) @xmath747 where the element @xmath748 is central in @xmath45 and therefore acts by the same eigenvalue on all vectors in each irreducible representation : using , we readily calculate @xmath749 the elements @xmath753 satisfy the equations @xmath754 where @xmath755 ( another relation between @xmath753 and @xmath756 , characteristic of the @xmath45 quantum group rather than of the quantum @xmath746 , is given in ) . we note that @xmath757 ( and similarly @xmath758 ) has two roots of multiplicity @xmath58 , which are @xmath759 , and all the other roots are of multiplicity @xmath710 . in finding decompositions of central elements as explained in * * , we also need the action of @xmath760 on some indecomposable modules . for example , the action of @xmath753 on @xmath573 ( see * * ) can be written as @xmath761 in terms of the bases introduced in * * . it now follows from and that @xmath762_+\,[a - s]_+\bigr)\ , { \mathsf{t}^{\scriptstyle\uparrow}}_{a , a ' } , \quad m < p_+.\ ] ] writing @xmath763 , we see that @xmath764\bigl({\mathscr{c}_{+,r , a}}^{m}(x)\bigr),\ ] ] where we define the polynomials @xmath765_\pm [ a - s]_\pm)\ ] ] and use @xmath766(f(x))$ ] to denote the coefficient at @xmath767 . we note that @xmath768\bigl({\mathscr{c}_{\pm , r , a}}^m(x)\bigr ) = ( [ m]_\pm!)^2{{\mathchoice { { { \mbox{\footnotesize$\displaystyle \genfrac{[}{]}{0pt}{}{a}{m}$}}}}{{\genfrac{[}{]}{0pt}{}{a}{m } } } { { \genfrac{[}{]}{0pt}{}{a}{m}}}{{\genfrac{[}{]}{0pt}{}{a}{m}}}}}_\pm{{\mathchoice { { { \mbox{\footnotesize$\displaystyle \genfrac{[}{]}{0pt}{}{r\!-\!a\!+\!m\!-\!1}{m}$}}}}{{\genfrac{[}{]}{0pt}{}{r\!-\!a\!+\!m\!-\!1}{m } } } { { \genfrac{[}{]}{0pt}{}{r\!-\!a\!+\!m\!-\!1}{m}}}{{\genfrac{[}{]}{0pt}{}{r\!-\!a\!+\!m\!-\!1}{m}}}}}_\pm\ ] ] and @xmath769\bigl({\mathscr{c}_{+,1,a}}^{m}(x)\bigr ) & = ( -1)^{m + 1}[m]_+ ! [ m\!-\!1]_+!{{\mathchoice { { { \mbox{\footnotesize$\displaystyle \genfrac{[}{]}{0pt}{}{a\!-\!1}{a\!-\!m}$}}}}{{\genfrac{[}{]}{0pt}{}{a\!-\!1}{a\!-\!m } } } { { \genfrac{[}{]}{0pt}{}{a\!-\!1}{a\!-\!m}}}{{\genfrac{[}{]}{0pt}{}{a\!-\!1}{a\!-\!m}}}}}_+^2 , \\ \intertext{with $ [ x^1]\bigl({\mathscr{c}_{+,1,0}}^{m}(x)\bigr ) = ( -1)^{m + 1}[m]_+ ! [ m\!-\!1]_+!$ for $ m{\,{\geqslant}\,}1 $ , and } \label{c - top } [ x^1]\bigl({\mathscr{c}_{+,p_+ - 1,a}}^{m}(x)\bigr ) & = ( -1)^{(p_- + 1)(m + 1 ) } [ m\!-\!1]_+ ! [ m]_+ ! { { \mathchoice { { { \mbox{\footnotesize$\displaystyle \genfrac{[}{]}{0pt}{}{a}{m\!-\!1}$}}}}{{\genfrac{[}{]}{0pt}{}{a}{m\!-\!1 } } } { { \genfrac{[}{]}{0pt}{}{a}{m\!-\!1}}}{{\genfrac{[}{]}{0pt}{}{a}{m\!-\!1}}}}}_+^2.\end{aligned}\ ] ] we begin with calculating the comultiplication of the cointegral as @xmath770 and then evaluate the traces in the definition of @xmath325 using . this gives @xmath771 where @xmath64 and @xmath383 , and @xmath772_+!)^2 \\ { } \times { \mathfrak{q}_+}^{p_- ( r - 1 - 2a)(b - m ) } { { \mathchoice { { { \mbox{\footnotesize$\displaystyle \genfrac{[}{]}{0pt}{}{r\!-\!a\!+\!m\!-\!1}{m}$}}}}{{\genfrac{[}{]}{0pt}{}{r\!-\!a\!+\!m\!-\!1}{m } } } { { \genfrac{[}{]}{0pt}{}{r\!-\!a\!+\!m\!-\!1}{m}}}{{\genfrac{[}{]}{0pt}{}{r\!-\!a\!+\!m\!-\!1}{m}}}}}_+ { { \mathchoice { { { \mbox{\footnotesize$\displaystyle \genfrac{[}{]}{0pt}{}{a}{m}$}}}}{{\genfrac{[}{]}{0pt}{}{a}{m } } } { { \genfrac{[}{]}{0pt}{}{a}{m}}}{{\genfrac{[}{]}{0pt}{}{a}{m}}}}}_+ { f_+}^{p_+ - m - 1 } { e_+}^{p_+ - m - 1}k_+^{b - m}\end{gathered}\ ] ] and @xmath773_-!)^2\\ { } \times { \mathfrak{q}_-}^{p_+(r ' - 1 - 2a ' - 2m')(b ' + m ' ) } { { \mathchoice { { { \mbox{\footnotesize$\displaystyle \genfrac{[}{]}{0pt}{}{r'\!-\!a'\!-\!1}{m'}$}}}}{{\genfrac{[}{]}{0pt}{}{r'\!-\!a'\!-\!1}{m ' } } } { { \genfrac{[}{]}{0pt}{}{r'\!-\!a'\!-\!1}{m'}}}{{\genfrac{[}{]}{0pt}{}{r'\!-\!a'\!-\!1}{m'}}}}}_- { { \mathchoice { { { \mbox{\footnotesize$\displaystyle \genfrac{[}{]}{0pt}{}{a'\!+\!m'}{m'}$}}}}{{\genfrac{[}{]}{0pt}{}{a'\!+\!m'}{m ' } } } { { \genfrac{[}{]}{0pt}{}{a'\!+\!m'}{m'}}}{{\genfrac{[}{]}{0pt}{}{a'\!+\!m'}{m'}}}}}_- { f_-}^{p_- - m ' - 1 } { e_-}^{p_- - m ' - 1}k_-^{b ' + m'}.\end{gathered}\ ] ] as regards the central elements @xmath774 and @xmath775 , we recall the nondiagonal action of @xmath760 on indecomposable modules , see * * . straightforward calculations then give the following formulas : @xmath776 for @xmath64 and @xmath173 , and @xmath777 for @xmath225 , where @xmath778_+}{{\mathfrak{q}_+}^{p_-}\!- { \mathfrak{q}_+}^{-p_-}}$ } } \sum_{m=0}^{p_+ - 1}\sum_{m'=0}^{p_- - 1 } \sum_{a=0}^{r - 1}\sum_{a'=m'}^{r ' - 1 } \sum_{\ell=0}^{2 p_+ p_- - 1}\!\!\ ! \alpha^{\ell - p_- } ( -1)^{r ' - 1}\\ \qquad{}\times{\mathfrak{q}_+}^{(\ell - p_- m)(r - 1 - 2a ) } { \mathfrak{q}_-}^{(\ell + p_+ m')(r ' - 1 - 2a ' ) } [ x^1]\bigl({\mathscr{c}_{+,r , a}}^{m}(x)\bigr)\ , [ y^0]\bigl({\mathscr{c}_{-,r',a'}}^{m'}(y)\bigr)\\ { } \times{f_+}^{p_+ - 1 - m } { e_+}^{p_+ - 1 - m } { f_-}^{p_- - 1 - m ' } { e_-}^{p_- - 1 - m ' } k^{\ell - p_- m + p_+ m'},\end{gathered}\ ] ] @xmath779_-}{{\mathfrak{q}_-}^{p_+}\!- { \mathfrak{q}_-}^{-p_+}}$}}\sum _ { m'=0}^{p_- - 1}\sum_{m=0}^{p_+ - 1 } \sum_{\ell=0}^{2 p_+ p_- - 1 } \sum_{a'=0}^{r ' - 1 } \sum_{a = m}^{r - 1 } \alpha^{\ell - p_+ } ( -1)^{r - 1}\\ * \qquad{}\times{\mathfrak{q}_-}^{(\ell - p_+ m')(r ' - 1 - 2a ' ) } { \mathfrak{q}_+}^{(\ell + p_- m)(r - 1 - 2a ) } [ x^0]\bigl({\mathscr{c}_{+,r , a}}^{m}(x)\bigr)\ , [ y^1]\bigl({\mathscr{c}_{-,r',a'}}^{m'}(y)\bigr)\\ * { } \times{f_-}^{p_- - 1 - m ' } { e_-}^{p_- - 1 - m ' } { f_+}^{p_+ - 1 - m } { e_+}^{p_+ - 1 - m } k^{\ell - p_+ m ' + p_- m},\end{gathered}\ ] ] and @xmath780_+\,[r']_-}{({\mathfrak{q}_+}^{p_-}\!-\!{\mathfrak{q}_+}^{-p_- } ) ( { \mathfrak{q}_-}^{p_+}\!-\!{\mathfrak{q}_-}^{-p_+})}$ } } \sum_{m=0}^{p_+ - 1}\sum_{m'=0}^{p_- - 1 } \sum_{a=0}^{r - 1}\sum_{a'=0}^{r ' - 1 } \sum_{\ell=0}^{2 p_+ p_- - 1}\!\!\ ! \alpha^{\ell + p_- + p_+ } ( -1)^{r ' + r}\\ * { } \times{\mathfrak{q}_+}^{(\ell - p_- m)(r - 1 - 2a ) } { \mathfrak{q}_-}^{(\ell - p_+ m')(r ' - 1 - 2a ' ) } [ x^1]\bigl({\mathscr{c}_{+,r , a}}^{m}(x)\bigr)\ , [ y^1]\bigl({\mathscr{c}_{+,r',a'}}^{m'}(y)\bigr)\\ * { } \times{f_+}^{p_+ - 1 - m } { e_+}^{p_+ - 1 - m } { f_-}^{p_- - 1 - m ' } { e_-}^{p_- - 1 - m ' } k^{\ell - p_- m - p_+ m'}.\end{gathered}\ ] ] the rest of the proof ( establishing the decomposition in terms of the canonical central elements ) is by direct evaluation of the action of @xmath325 , @xmath774 , @xmath775 , and @xmath486 on the projective modules , as described in * * . the nondiagonal part of the action on the projective modules described in * * is related to the coefficients at the radical elements entering the decomposition with respect to the canonical central elements . the calculation yielding , , , , and is very similar to the one given in @xcite , and we therefore omit it . feigin , a.m. gainutdinov , a.m. semikhatov , and i.yu . tipunin , _ modular group representations and fusion in logarithmic conformal field theories and in the quantum group center _ , commun . math . 265 ( 2006 ) 4793 [ hep - th@xmath150504093 ] . feigin , a.m. gainutdinov , a.m. semikhatov , and i.yu . tipunin , _ kazhdan lusztig correspondence for the representation category of the triplet @xmath1-algebra in logarithmic cft _ , math.qa@xmath150512621 . d. kazhdan and g. lusztig , _ tensor structures arising from affine lie algebras , _ i , j. amer . soc . 6 ( 1993 ) 905947 ; ii , j. amer . ( 1993 ) 9491011 ; iii , j. amer . math . soc . 7 ( 1994 ) 335381 ; iv , j. amer . math . soc . 7 ( 1994 ) 383453 . g. felder , _ brst approach to minimal models _ , b317 ( 1989 ) 215236 . v. lyubashenko , _ invariants of @xmath638-manifolds and projective representations of mapping class groups via quantum groups at roots of unity _ , commun . ( 1995 ) 467516 [ hep - th@xmath159405167 ] ; _ modular properties of ribbon abelian categories _ , symposia gaussiana , proc . of the 2nd gauss symposium , munich , 1993 , conf . a ( berlin , new york ) , walter de gruyter , ( 1995 ) 529579 [ hep - th@xmath159405168 ] ; _ modular transformations for tensor categories _ , j. pure applied algebra 98 ( 1995 ) 279327 . feigin , a.m. gainutdinov , a.m. semikhatov , and i.yu . tipunin , _ logarithmic extensions of minimal models : characters and modular transformations _ , nucl . phys . b 757 ( 2006 ) 303343 [ hep - th@xmath150606196 ] .	we derive and study a quantum group @xmath0 that is kazhdan lusztig - dual to the @xmath1-algebra @xmath2 of the logarithmic @xmath3 conformal field theory model . the algebra @xmath2 is generated by two currents @xmath4 and @xmath5 of dimension @xmath6 , and the energy momentum tensor @xmath7 . the two currents generate a vertex - operator ideal @xmath8 with the property that the quotient @xmath9 is the vertex - operator algebra of the @xmath3 virasoro minimal model . the number ( @xmath10 ) of irreducible @xmath0 representations is the same as the number of irreducible @xmath2-representations on which @xmath8 acts nontrivially . we find the center of @xmath0 and show that the modular group representation on it is equivalent to the modular group representation on the @xmath2 characters and `` pseudocharacters . '' the factorization of the @xmath0 ribbon element leads to a factorization of the modular group representation on the center . we also find the @xmath0 grothendieck ring , which is presumably the `` logarithmic '' fusion of the @xmath3 model .
this work has been supported by the miur project firb - lichis- rbfr10yq3h . mgap thanks f. ticozzi and p. villoresi for discussions . 99 c. w. helstrom , _ quantum detection and estimation theory _ ( new york : academic , 1976 ) . j. a. bergou , u. herzog and m. hillery , lect . not * 649 * , 415 ( 2004 ) a. chefles , lect . not * 649 * , 465 ( 2004 ) j. a. bergou , j. mod . opt . * 57 * , 160 ( 2010 ) . m. osaki , m. ban and o. hirota , phys . a * 54 * , 1691 ( 1996 ) . k. kato , m. osaki , m. sasaki and o. hirota , ieee trans . comm . * 47 * , 248 ( 1999 ) . lau , v. a. vilnrotter , s. dolinar , j. m. geremia and h. mabuchi , ipn progr . rep . * 42 - 165 * , 1 ( 2006 ) . r. s. kennedy , mit rle quart . progr . rep . * 108 * , 1973 , p. 219 . m. takeoka , m. ban and m. sasaki , phys . a * 313 * , 16 ( 2003 ) . s. dolinar , mit rle quart . rep . * 111 * , 1973 , p. 115 . a. assalini , n. dalla pozza and g. pierobon , phys . a * 84 * , 022342 ( 2011 ) . v. a. vilnrotter , ipn progress report * 42 - 189 * , 1 ( 2012 ) . r. l. cook , p. j. martin and j. m. geremia , nature * 466 * , 774 ( 2007 ) . m. gregory , f. heine , h. kampfner , r. lange , m. lutzer and r. meyer , opt . * 51 * , 031202 ( 2012 ) . m. ban , j. mod . opt . * 44 * , 1175 ( 1997 ) . s. olivares and m. g. a. paris , j. opt . b : quantum semiclass . opt . * 6 * , 69 ( 2004 ) . g. cariolaro and g. pierobon , ieee trans . comm * 58 * , 623 ( 2010 ) . c. wittmann , u. l. andersen , m. takeoka , d. sych and g. leuchs , phys . a * 81 * , 062338 ( 2010 ) . m. takeoka and m. sasaki , phys . a * 78 * 022320 ( 2008 ) . f. e. becerra , j. fan , g. baumgartner , s. v. polyakov , j. goldhar , j. t. kosloski and a. migdall , phys . a * 84 * , 062324 ( 2011 ) . c. r. muller , m. a. usuga , c. wittmann , m. takeoka , ch . marquardt , u. l. andersen and g. leuchs , new j. phys . * 14 * , 083009 ( 2012 ) . s. izumi , m. takeoka , m. fujiwara , n. dalla pozza , a. assalini , k. ema and m. sasaki , phys . a * 86 * , 042328 ( 2012 ) . r. nair , b. j. yen , s. guha , j. h. shapiro and s. pirandola , phys . rev . a * 86 * , 022306 ( 2012 ) . c. gardiner and p. zoller , _ quantum noise _ ( springer , 2004 ) m. g. genoni , s. olivares and m. g. a. paris , phys . rev . lett . * 106 * , 153603 ( 2011 ) . m. g. genoni , s. olivares , d. brivio , s. cialdi , d. cipriani , a. santamato , s. vezzoli and m. g. a. paris , phys . a * 85 * , 043817 ( 2012 ) . j. a. lopez leya , a. arvizu mondragon , e. garcia , f. j. mendieta , e. alvarez guzman and p. gallion , opt . eng . * 51 * , 105002 ( 2012 ) .	we address binary optical communication channels based on phase - shift keyed coherent signals in the presence of phase diffusion . we prove theoretically and demonstrate experimentally that a discrimination strategy based on homodyne detection is robust against this kind of noise for any value of the channel energy . moreover , we find that homodyne receiver beats the performance of kennedy receiver as the signal energy increases , and achieves the helstrom bound in the limit of large noise . _ introduction _ in a binary quantum communication channel the sender encodes the logical symbols `` 1 '' and `` 0 '' on two states of a physical system described by the density operators @xmath0 and @xmath1 , respectively . in order to retrieve the logical information , the receiver should discriminate between the two signals which , in general , may be not orthogonal , due to the encoding process itself ( @xmath0 and @xmath1 may refer to non - orthogonal states ) or because of the noise during the propagation stage @xcite . in this case , an unavoidable probability of error @xmath2 $ ] appears , where @xmath3 is the probability of inferring the symbol @xmath4 when the signal sent over the channel was meant for @xmath5 , and where we assumed that the two states are sent with equal probability . the minimum error probability allowed by quantum mechanics is given by the helstrom bound @xcite @xmath6 , where @xmath7 . optimal discrimination of non orthogonal states is a crucial topic for the effective implementation of quantum communication channels and , as a consequence , different strategies has been employed to attack the problem in different situations @xcite . in the following , we address binary communication channels with phase - shift keyed ( psk ) signals @xcite , i.e. , channels where the information is encoded on two coherent states @xmath8 and @xmath9 ( without lack of generality we can assume @xmath10 , @xmath11 ) . in this case , the helstrom bound rewrites as : @xmath12 $ ] , i.e. @xmath13,\end{aligned}\ ] ] where @xmath14 is the average number of photons contained in each signal , and it will be referred to as the _ signal energy _ throughout the paper . a detection strategy achieving this level of error probability is said to be an _ optimum receiver_. quantum state discrimination strategies for psk signals is an active field of research since it encompasses the read - out problem , and it is an important tool to assess the performances of quantum limited measurements in optical communication , e.g. , for deep - space missions @xcite . on the one hand , in the high energy regime , a near - optimum receiver based on photodetection , the so - called kennedy receiver , has been proposed long ago for coherent signals propagating in an ideal channel @xcite . kennedy receiver has been also extended @xcite , and , in particular , an optimum receiver based on an adaptive scheme has been recently experimentally realized @xcite . on the other hand , strategies based on homodyne detection have great practical advantages @xcite , and for this reason their performances have been analyzed in realistic situations . in particular , receivers based on homodyne detection has been theoretically @xcite and experimentally investigated @xcite in the presence of losses and phase insensitive thermal noise . more generally , these receivers have been proved to represent the optimum gaussian strategy for the discrimination of the psk coherent signals @xcite , and it has been also demonstrated that it possible to emulate adaptive processes by means of postprocessing and bayesian analyses @xcite . hybrid @xcite and displacement - based @xcite detectors have been also implemented to decrease the error probability , with application to @xmath15-ary communication channels @xcite . communication schemes based on coherent signals may be useful in scenarios in which quantum resources ( as single photons and entanglement ) can not be fully exploited , as in free - space communication . in this paper we analyze a source of noise , namely , phase diffusion , which is detrimental for coherent - signal based channels . at the same time , this provides a relevant example of non - gaussian noise . we demonstrate experimentally the robustness of the homodyne receiver against phase noise . moreover , we find that in the presence of phase noise , homodyne receiver beats the performances of kennedy receiver as the signal energy increases and/or the noise is larger than an intensity dependent threshold value . finally , we show that homodyne detection achieves the helstrom bound in the limit of large noise . _ phase diffusion _ any phase diffusion process coming from a non - dissipative interaction with a bosonic environment may be described by a suitable phase - diffusion master equation @xcite . the overall effect of phase diffusion on a coherent state @xmath16 is described by the map @xcite : @xmath17 is a normal distribution of the variable @xmath18 with zero mean and standard deviation @xmath19 . at the end of a channel affected by phase diffusion the receiver is faced with the problem of discriminating between @xmath20 and @xmath21 . in order to calculate the corresponding helstrom bound , one has to diagonalize the traceless operator @xmath22 whose expansion in the photon - number basis reads : @xmath23\ , numerically after a suitable truncation of the hilbert space , which is determined by the average number of photons of the two signals . the helstrom bound for different values of the signal energy @xmath24 and the diffusion coefficient @xmath19 is reported in fig . [ f : exp ] , together with the experimental data for the homodyne receiver . for small values of the coherent amplitude we may truncate @xmath25 at low dimension achieving the analytic expression @xmath26 whereas , in the limit @xmath27 of large noise we have @xmath28 where @xmath29 is a decreasing function of the amplitude . _ kennedy receiver in the presence of phase noise _ in the ideal case , a near - optimum strategy is provided by the kennedy receiver @xcite based on photodetection : the signal mode @xmath30 excited either in @xmath31 or @xmath32 , interferes with a reference mode @xmath33 excited in the coherent state @xmath34 at a beam splitter ( bs ) with transmissivity @xmath35 and a photodetector detects the light at one output . the overall evolved state reads @xmath36 is the evolution operator associated with the bs . if we choose @xmath37 , take @xmath38 and consider only the first output port , then we have the following input - output relation @xmath39 . therefore , a natural discrimination strategy is to associate the absence of light ( the vacuum ) with the symbol `` 0 '' , i.e. , @xmath32 , and the detection of any number of photons with the symbol is `` 1 '' , i.e. , @xmath31 . this strategy also implies a nonzero probability of inferring the wrong symbol , which is determined by the conditional probabilities @xcite : [ p : cond : dd ] @xmath40 = e^{-4n } , \\ p(1\|0 ) & = \lim_{\tau\to 1 } { \rm tr}[\hu_{\tau}\hvarrho_{-\alpha}\otimes \hvarrho_{\beta}\hu_{\tau}^{\dag } \ , \hpi_1\otimes\hat{\mathbbm i } ] = 0,\end{aligned}\ ] ] where we introduced the positive - operator valued measure @xmath41 and @xmath42 , describing an on - off detector detecting the absence or the presence of photons , respectively . the overall probability of error ( still assuming the two signals sent with the same probability ) is : @xmath43 it is worth noting that , in the limit @xmath44 , @xmath45 , i.e. , the kennedy receiver based on photodetection is nearly optimal . in the presence of phase diffusion , eqs . ( [ p : cond : dd ] ) become : [ p : cond : dd : phn ] @xmath46,\\ p_\delta ( 1\|0 ) & = 1- \int_{\mathbbm r } d\phi\ , g(\phi,\delta)\ , \exp\left [ -4\alpha^2\sin^2(\phi/2 ) \right],\end{aligned}\ ] ] and @xmath47\,,\ ] ] which can be easily evaluated numerically . for small values of the coherent amplitude we have the analytic expression @xmath48 whereas , in the limit @xmath27 of large noise we have @xmath49 where @xmath50 is a decreasing function of the amplitude . _ homodyne receiver in the presence of phase noise _ let us now focus on a different strategy based on homodyne detection : the receiver measures the quadrature @xmath51 , where @xmath52 $ ] ( in our case @xmath53 or @xmath54 ) and associates the symbol `` 1 '' ( `` 0 '' ) to a positive ( `` negative '' ) outcome . in the presence of losses and thermal noise but without phase noise , this strategy has been proven to beat the photodetection strategy either in the low or in the high energy regime @xcite . in the presence of phase diffusion we have the following conditional probabilities ( the noiseless case is recovered in the limit @xmath55 ) : @xmath56 where : @xmath57 is the homodyne probability , namely , the probability of obtaining as outcome @xmath58 addressing the quadrature @xmath59 given the input @xmath60 . the overall probability of error thus writes : @xmath61\,,\ ] ] and can be easily evaluated numerically . for small values of the coherent amplitude we have the analytic expression @xmath62 whereas , in the limit @xmath27 of large noise we have @xmath63 where @xmath64 is a decreasing function of the amplitude with @xmath65 and with @xmath66 approaching @xmath29 for increasing amplitude . ] _ experimentals _ the experimental apparatus we used to investigate the performance of the homodyne receiver in the presence of phase diffusion is sketched in fig . [ f : scheme ] . the output beam of a he : ne laser is split into two paths , the local oscillator ( lo in fig . [ f : scheme ] ) and the signal , by means of a suitable combination of a half wave plate ( hwp ) and a polarizing beam splitter ( pbs ) . the phase modulation ( pm ) is obtained by a piezo connected to the computer ( pc ) , which sends a voltage signal with a frequency of 1 khz corresponding to the phase diffusion . the amplitude modulation ( am ) of the coherent signals is controlled by a suitable combination of a kdp crystal and a pbs @xcite . the actual homodyne receiver is composed by a 50:50 beam splitter ( bs ) and a balanced amplifier detector based on a silicon photodiodes ( d1 and d2 ) . in order to experimentally determine the probability of error , for a fixed value of the signal energy @xmath67 , we performed 10 runs with @xmath68 shots each , randomly chosen between @xmath69 and @xmath70 . , of the coherent signals @xmath69 ( red ) and @xmath71 ( blue ) for increasing noise ( from top to bottom , @xmath72 rad ) in the case of @xmath73 . right panels : homodyne traces for the quadrature @xmath59 versus the detection time . the horizontal line refers to the threshold for the discrimination strategy : if the dots fall above ( below ) that line , the signal is chosen to be `` 1 '' ( `` 0 '' ) . [ f : traces ] ] _ data analysis and discussion _ in fig . [ f : traces ] we report the data of our experiments for different values of the noise parameter . in the left panels we show the homodyne traces for the quadrature @xmath74 , of the coherent signals @xmath69 ( red ) and @xmath70 ( blue ) as a function of @xmath18 and for increasing level of noise . in the right panels we report homodyne data for the optimal quadrature @xmath59 versus the detection time . the horizontal line refers to the threshold for the discrimination strategy : if the dots ( i.e. , the homodyne outcomes ) fall above ( below ) that line , the signal is inferred to be `` 1 '' ( `` 0 '' ) . in fig . [ f : exp ] we report the behavior of the probabilities of error @xmath75 , @xmath76 and @xmath77 as functions of the phase diffusion coefficient @xmath19 and for different values of the input energy @xmath24 . as one may expect , the presence of the phase noise dramatically affects the performance of the kennedy receiver : it is worth noting that as the input energy increases , a small amount of phase noise in enough to increase of orders of magnitude the probability of error @xmath78 . on the other hand , the strategy based on the homodyne detection is quite robust with respect to the phase noise . as it is apparent from the plots , for @xmath79 the value of @xmath80 is almost constant . in addition , @xmath80 approaches the quantum mechanical limit given by the helstrom bound @xmath81 as @xmath19 increases , as anticipated by the asymptotic expansions in eqs . ( [ asyqd ] ) and ( [ asyhd ] ) . in general , as the signal energy and the noise increase , the homodyne receiver becomes more effective than kennedy receiver . for different values of the signal energy @xmath24 . each point corresponds to the average of ten acquisitions of @xmath68 values with a phase diffusion frequency of 1 khz . the errors bars corresponds to the standard deviation of the mean . the theoretical predictions @xmath82 ( solid red ) is reported for comparison . we also show the error probability @xmath83 of a kennedy receiver ( dashed blue ) in the same experimental conditions and the helstrom bound @xmath75 ( dotted green ) . [ f : exp],title="fig : " ] for different values of the signal energy @xmath24 . each point corresponds to the average of ten acquisitions of @xmath68 values with a phase diffusion frequency of 1 khz . the errors bars corresponds to the standard deviation of the mean . the theoretical predictions @xmath82 ( solid red ) is reported for comparison . we also show the error probability @xmath83 of a kennedy receiver ( dashed blue ) in the same experimental conditions and the helstrom bound @xmath75 ( dotted green ) . [ f : exp],title="fig : " ] for different values of the signal energy @xmath24 . each point corresponds to the average of ten acquisitions of @xmath68 values with a phase diffusion frequency of 1 khz . the errors bars corresponds to the standard deviation of the mean . the theoretical predictions @xmath82 ( solid red ) is reported for comparison . we also show the error probability @xmath83 of a kennedy receiver ( dashed blue ) in the same experimental conditions and the helstrom bound @xmath75 ( dotted green ) . [ f : exp],title="fig : " ] for different values of the signal energy @xmath24 . each point corresponds to the average of ten acquisitions of @xmath68 values with a phase diffusion frequency of 1 khz . the errors bars corresponds to the standard deviation of the mean . the theoretical predictions @xmath82 ( solid red ) is reported for comparison . we also show the error probability @xmath83 of a kennedy receiver ( dashed blue ) in the same experimental conditions and the helstrom bound @xmath75 ( dotted green ) . [ f : exp],title="fig : " ] this behavior is illustrated in fig . [ f : delta : th ] , where we plot the threshold @xmath84 on the diffusion coefficient @xmath19 as a function of the the signal energy @xmath24 : at any fixed value of @xmath24 if @xmath85 then homodyne receiver exhibits a smaller error probability than kennedy receiver in the same experimental conditions , in agreement with eqs . ( [ asykd ] ) and ( [ asyhd ] ) . _ conclusions _ in conclusion , we have addressed psk binary optical communication in the presence of phase diffusion , i.e. , a detrimental noise for schemes based on coherent signals . we demonstrated experimentally that a discrimination strategy based on homodyne detection is robust against this kind of noise . in addition , we have also demonstrated that homodyne receivers beat the performances of kennedy receivers as far as the noise is larger than an energy - dependent threshold . finally , we have shown that homodyne receivers achieve the helstrom bound on the error probability in the limit of large noise and for any value of the signal energy . our results help to clarify the fundamental limits of quantum communications , and show that receivers that perform near the quantum limit in noisy conditions are realizable with current technology @xcite . of the diffusion coefficient @xmath19 as a function of the signal energy @xmath24 . the meaning of the threshold value is the following : if @xmath86 , then homodyne receiver shows a smaller error probability than kennedy receiver ( gray region ) , the opposite otherwise ( white region ) . [ f : delta : th ] ]
coronal mass ejections ( cmes ) are observed with white - light coronagraphs as significant changes in coronal structure moving outward ( skirgiello , 2003 ) . they are believed to be the prime driver of most space weather events , e.g. , interplanetary shocks , high - energy particles and geomagnetic storms ( gosling , 1993 ) . recently many studies have been carried out on the properties of cme source regions in order to gain insight into the initiation mechanism of such phenomenon . subramanian and dere ( 2001 ) found that 41% of cme - related transients observed are associated with active regions and have no filament eruptions , 44% are associated with eruptions of filaments embedded in active regions , and 15% are associated with eruptions of filaments outside active regions . zhou et al . ( 2003 ) found that 88% of the earth - directed halo cmes are associated with flares and 94% are associated with eruptive filaments . with regard to the locations of cme source regions , they found that there are 79% cmes coming from active regions and 21% originating outside active regions . in this letter we concentrate on the latitudinal variations of cme source regions also with a view of providing clues about the cme triggering processes . skirgiello ( 2003 ) proposed a mathematical tool using the apparent latitude distribution to deduce the real latitude distribution . by referring to this idea , a much simpler and easier new method with the same function is designed in order to obtain the evolution of real cme latitudes during solar cycle 23 . cmes are selected from the catalog on the website of _ http://cdaw.gsfc.nasa.gov/cme_list _ in the interval 1996 january to 2006 december , as observed by soho / lasco . the soho / lasco coronagraph images solar corona continuously since 1996 covering a field of view starting from about 1.5 @xmath4 to 32 @xmath4 ( gopalswamy et al . 2003 , @xmath4 is the sun s radius ) . by performing preliminary estimation , st . cyr et al . ( 2000 ) found that its detection of cmes possessed an efficiency not less than 95% . the cmes are observed on the plane of the sky , therefore all apparent spatial parameters are a projection of the real values onto that plane . the apparent latitude @xmath5 comes from the central position angle ( @xmath6 ) , which is defined and listed in the cme catalog by yashiro et al . when @xmath7 , @xmath8 and when @xmath9 , @xmath10 . the complete halo events with an angular width of @xmath11 are excluded from the statistics because of difficulty in determining @xmath5 . in consideration of the cme number and the temporal precision , we divide the whole interval into 22 parts , as shown in figure 1 . if the time of a part is set very short , the cme number will get too small to be able to represent the real distribution of @xmath5 ; if very long , the temporal precision will then get so bad that it is difficult to show the solar cycle variation clearly . for the time interval of every part , we compute the occurrence percentages of cmes corresponding to @xmath12 , @xmath13 , @xmath14 , and @xmath15 , respectively . the bin width of @xmath5 is @xmath16 by following that of the @xmath6 . hence we can define a vector @xmath17^t \label{1}\ ] ] to give the distribution of @xmath5 , where @xmath18 means the percentage at @xmath19 . in the present study , @xmath20 is smoothed with a window of @xmath21 , see the thin solid lines in figure 1 . furthermore , the distributions for the northern ( @xmath22 ) and southern ( @xmath23 ) hemisphere are summed without the consideration of the north - south asymmetry of solar activities , e.g. , the sunspots ( oliver & ballester , 1994 ) , the filaments ( hansen & hansen , 1975 ) , the x - ray flares ( li et al . 1998 ) and the high - latitude cmes ( gopalswamy et al . the proposed method is based on the assumption of radial motion and uniform longitudinal distribution of cmes . as shown in figure 2 , the point @xmath24 indicates the cme source region , @xmath25 is the direction of cme motion , @xmath26 is @xmath25 s projection onto the plane of the sky . thereupon @xmath27 , @xmath28 , and @xmath29 are , namely , the apparent latitude @xmath5 , the real latitude @xmath1 and the longitude @xmath30 . by a simple deduction ( @xmath31 , @xmath32 , @xmath33 ) we have @xmath34 set the discrete variable @xmath35 $ ] . for each @xmath1 fixed , we can obtain 91 @xmath5 values by using equation 2 if given @xmath36 , @xmath37 . we let all @xmath5 equal to the expression @xmath38 converted to integer type and then we can define a vector @xmath39^t$ ] , where @xmath40 means the percentage at @xmath19 in the obtained 91 @xmath5 values . some special conditions are dealt with as follows . when @xmath41 , @xmath42^t$ ] ; when @xmath43 , @xmath44^t$ ] ; when @xmath36 and @xmath45 , @xmath15 . combining 91 vectors @xmath46 we get a @xmath47 matrix @xmath48 , \label{1}\ ] ] where the subscript @xmath49 means the case @xmath50 . figure 3 shows the contour of matrix @xmath51 . if we use vector @xmath52^t$ ] to indicate the real distribution of @xmath1 , we have @xmath53 and @xmath54 the final results of @xmath55 , also smoothed with a window of @xmath21 , are plotted as 22 thick solid lines in figure 1 . in comparison with @xmath20 , most high - latitude cmes are substituted with events coming from low latitudes . the reason of this phenomenon is that by projection the cmes remote from solar limb can be observed at any @xmath56 . for a direct demonstration of the solar cycle variation of @xmath1 , we combine all vectors @xmath55 in time order and give a filled contour plot in figure 4a . figure 4b then shows the variation of the actual ( not the percentage ) cme occurrence rate @xmath57 , where @xmath58 is the cme number , @xmath59 is the time length ( in @xmath60 ) , both given in figure 1 . from figure 4 , we find that most high - latitude cmes occur during the period from 1999 to the middle of 2001 . gopalswamy et al . ( 2003 ) found a general spreading of cme latitudes attaining @xmath61 by 1999 . the northern high - latitude cmes became nonexist beyond october 2000 . however the southern ones continued to occur until the first quarter of 2002 . our time interval of most high - latitude cmes occurrence is approximately consistent with the above gopalswamy et al.s . as shown by white dotted lines , we find a poleward motion with a speed of 12.5 @xmath62 and an equatorward motion with a speed of 25.4 @xmath62 taking place during years 1999 - 2001 and during 2000 to the middle of 2001 , respectively . for the poleward motion , years 1999 and 2001 are the polarity reversal time at @xmath63 and the epoch when the polar regions area is of minimum ( song & wang , 2006 ) . therefore we suggest this motion come from the collision between the unipolar poleward meridional flows ( song & wang , 2006 ) and the polar regions of an opposite polarity . such collision hastens the formation of polar crown filaments whose eruptions have a close correspondence with cmes ( gopalswamy et al . the much lower speed than that of the meridional flows ( @xmath64 @xmath62 ) can be explained by the polar regions blocking the way . the equatorward motion may have the same mechanism because during its interval , we can see that the tilt angle , defined as the maximum extent of the heliospheric current sheet ( hcs ) , also declines sharply with a similar speed ( see figure 8 in gopalswamy et al . the high - latitude cmes constitute 5 - 10% during the time when the polar magnetic fields reverse sign and 3% of all cmes during the whole solar cycle . in the range @xmath65 $ ] , we only find three prominent peaks located in 2000 , near the middle of 2002 and in 2005 , respectively , which seem to have a period of 2.5 @xmath66 . by comparing these peak positions with the variation of sunspot number ( see the pink solid lines in figure 4 ) , we find no any close correlations . therefore this phenomenon is very puzzling indeed . wang et al . ( 1989 ) found many poleward surges of alternating polarities in the active belts . do they initiate the middle - latitude cmes ? we have no definite answer at present . such cmes constitute 6 - 12% during their main phases and 4% of all cmes . 93% cmes occur in the range @xmath67 . from figure 4 , we find that the highest occurrence probability ( @xmath55 ) of these low - latitude cmes is at the minimum and during the declining phase of solar cycle . however the highest occurrence rate ( @xmath68 ) is at the maximum and during the declining phase of solar cycle . as shown by green solid lines , at the minimum and during the rising phase , cmes average latitude has a linear increase . then at the maximum and during the declining phase , cmes average latitude is in a nearly steady state , which keeps around @xmath69 . it is obvious that it does not follow the sprer sunspot law . this means that there are a large quantity of cmes originating outside active regions . zhou et al . ( 2006 ) classified cme - associated large - scale structures into four different categories : extended bipolar regions ( ebrs ) , transequatorial magnetic loops , transequatorial filaments and long filaments along the boundaries of ebrs ( maybe corresponding to the high - latitude cmes ) . the latitudes of former three categories are higher or lower than active regions , which result in a much wider distribution of cme latitudes . using the proposed method , we have studied the solar cycle variation of real cme latitudes during cycle 23 and found three main features . ( 1 ) the latitudinal drift of the high - latitude cmes ( @xmath0 ) was correlated with that of the hcs . ( 2 ) the mid - latitude cmes ( @xmath65 $ ] ) occurred primarily in 2000 , near the middle of 2002 and in 2005 , respectively , forming a prominent three - peak structure . ( 3 ) the latitudinal evolution of low - latitude cmes ( @xmath70 ) did not follow the sprer sunspot law , which suggests that many cmes originated outside of active regions . we believe that such features can provide hints for understanding the origin of cmes . at present , solar terrestrial relations observatory ( stereo ) , a very nice mission of nasa and launched successfully last october , is observing solar corona in 3-d instead of on the plane of the sky . with it the finding of cme source regions is no longer a difficult thing and then real cme latitude distribution can be better determined . this cme catalog is generated and maintained at the cdaw data center by nasa and the catholic university of america in cooperation with the naval research laboratory . soho is a project of international cooperation between esa and nasa . this work is jointly supported by national natural science foundation of china ( 40621003 , 40536029 , and 40604019 ) , the 973 project under grant 2006cb806304 , and the cas international partnership program for creative research teams . gopalswamy , n. , lara , a. , yashiro , s. , et al . 2003 , international solar cycle studies ( iscs ) symposium , solar variability as an input to the earth s environment , 403 gosling , j.t . 1993 , jgr , 98 , 18937 hansen , r. , & hansen , s. 1975 , sol . phys . , 44 , 225 li , k.j . , schmieder , b. , & li , q.s . 1998 , a&as , 131 , 99 oliver , r. , & ballester , j.l . 1994 , sol . phys . , 152 , 481 skirgiello , m. 2003 , grl , 30 , ssc1 song , w.b . , & wang , j.x . 2006 , apj , 643 , l69 st . cyr , o.c . , howard , r.a . , sheeley , n.r . , 2000 , jgr , 105 , 8169 subramanian , p. & dere , k.p . 2001 , apj , 561 , 372 wang , y .- m . , nash , a.g . , & sheeley , n.r . 1989 , science , 245 , 712 zhou , g.p . , wang , j.x . , & cao , z.l . 2003 , a&a , 397 , 1057 zhou , g.p . , wang , j.x . , & zhang , j. 2006 , a&a , 445 , 1133	with the assumption of radial motion and uniform longitudinal distribution of coronal mass ejections ( cmes ) , we propose a method to eliminate projection effects from the apparent observed cme latitude distribution . this method has been applied to soho lasco data from 1996 january to 2006 december . as a result , we find that the real cme latitude distribution had the following characteristics : ( 1 ) high - latitude cmes ( @xmath0 where @xmath1 is the latitude ) constituted 3% of all cmes and mainly occurred during the time when the polar magnetic fields reversed sign . the latitudinal drift of the high - latitude cmes was correlated with that of the heliospheric current sheet . ( 2 ) 4% of all cmes occurred in the range @xmath2 . these mid - latitude cmes occurred primarily in 2000 , near the middle of 2002 and in 2005 , respectively , forming a prominent three - peak structure ; ( 3 ) the highest occurrence probability of low - latitude ( @xmath3 ) cmes was at the minimum and during the declining phase of the solar cycle . however , the highest occurrence rate of low - latitude cmes was at the maximum and during the declining phase of the solar cycle . the latitudinal evolution of low - latitude cmes did not follow the sprer sunspot law , which suggests that many cmes originated outside of active regions .
many studies have shown that massive galaxies with low star formation rates were remarkably compact at @xmath4 ( e.g. , daddi et al . 2005 ; trujillo et al . 2006 ; van dokkum et al . 2008 ; damjanov et al . 2011 ; conselice 2014 ) . at fixed stellar mass of @xmath13 , quiescent galaxies are a factor of @xmath14 smaller at @xmath15 than at @xmath16 ( e.g. , van der wel et al . 2014b ) . as the stellar mass of the galaxies also evolves , the inferred size growth of individual galaxies is even larger ( van dokkum et al . 2010 ; patel et al . it is unlikely that all massive galaxies in the present - day universe had a compact progenitor ( van dokkum et al . 2008 , 2014 ; franx et al . 2008 ; newman et al . 2012 ; poggianti et al . 2013 ; belli , newman , & ellis 2014a ) ; however , the vast majority of compact , massive galaxies that are observed at @xmath15 ended up in the center of a much larger galaxy today ( belli et al . 2014a ; van dokkum et al . their size growth after @xmath15 is probably dominated by minor mergers : such mergers are expected , and other mechanisms can not easily produce the observed @xmath17 scaling between size growth and mass growth ( bezanson et al . 2009 ; naab , johansson , & ostriker 2009 ; hopkins et al . 2010 ; trujillo et al . 2011 ; hilz , naab , & ostriker 2013 ) . it is not yet clear how these massive , extremely compact galaxies were formed , and this question has significance well beyond the somewhat narrow context of the size evolution of quiescent galaxies . the dense centers of massive galaxies today are home to the most massive black holes in the universe ( magorrian et al . 1998 ) ; have an enrichment history that is very different from that of the milky way ( worthey , faber , & gonzalez 1992 ) ; and probably had a bottom - heavy stellar initial mass function ( imf ) ( conroy & van dokkum 2012 ) . all these characteristics are the product of processes that took place in the star forming progenitors of @xmath18 massive quiescent galaxies . furthermore , stars in very dense regions represent only a very small fraction ( @xmath19% ) of the stellar mass in the universe today , but their contribution rises sharply with redshift : depending on the imf , stars inside dense cores with @xmath20 may contribute 10% 20% of the stellar mass density at @xmath21 ( van dokkum et al . 2014 ) . the formation of compact massive galaxies requires large amounts of gas to be funneled in a region that is only 12 kpc in diameter , while preventing significant star formation at larger radii . galaxy formation models have been able to reproduce the broad characteristics of compact massive galaxies , either by mergers that are accompanied by a strong central star burst ( e.g. , hopkins et al . 2009b ; wuyts et al . 2010 ; wellons et al . 2015 ) , by in - situ formation from highly efficient gas cooling ( naab et al . 2009 ; wellons et al . 2015 ) , or by contraction ( `` compaction '' ) of star forming gas disks ( dekel & burkert 2014 ; zolotov et al . these scenarios have testable predictions : for example , if compact massive galaxies formed in mergers then they may be expected to show tidal features . furthermore , the star formation rates of galaxies , and their evolution in the size - mass plane , can be compared to observations . observationally , the challenge is to identify these star forming progenitors of compact massive galaxies . once they are found they can be studied , to measure the physical conditions inside them and to test proposed mechanisms for their formation ( see barro et al . 2013 , 2014b ; nelson et al . 2014 ; williams et al . 2014 , 2015 , for examples of such studies ) . the main observational complication is that typical quiescent galaxies at @xmath4 are structurally very different from typical star forming galaxies ( see , e.g. , franx et al . 2008 ) . at fixed mass , star forming galaxies are larger , have a lower sersic ( 1968 ) index and , as a result , a much lower central density ( e.g. , franx et al . 2008 ; kriek et al . 2009a ; van der wel et al . 2014b ; van dokkum et al . it may be that a subset of the star forming galaxies decrease their size through mergers or `` compaction '' , but it would be difficult to pinpoint which among the many large , star forming galaxies are destined to go through these phases . a similar problem arises when linking compact , quiescent descendants at @xmath15 to ( lower mass ) star forming galaxies at much higher redshift ( williams et al . 2014 , 2015 ) : although there may be progenitors of massive quiescent galaxies among small , blue , low mass star forming galaxies at @xmath22 , most of those galaxies will likely follow different paths . barro et al . ( 2013 , 2014b ) and nelson et al . ( 2014 ) use a relatively model - independent and straightforward way to identify plausible progenitors : they select massive star forming galaxies at @xmath4 with the same small sizes as quiescent galaxies . these objects form the compact tail of the size distribution of star forming galaxies : for every massive star forming galaxy at @xmath23 that is compact , there are several that are not ( see sect . [ candidates.sec ] , and van der wel et al . it seems plausible that star forming galaxies with the same structure as quiescent galaxies are the direct ancestors of these galaxies , and there may be physical reasons why the most compact star forming galaxies are the most likely to shut off : many proposed quenching and maintenance mechanisms operate most effectively when a significant bulge ( and associated black hole ) has formed ( croton et al . 2006 ; hopkins et al . 2008 ; johansson , naab , & ostriker 2009 ; conroy , van dokkum , & kravtsov 2015 ) . in this paper we build on previous studies by identifying a sample of massive , compact , star forming galaxies at @xmath23 in the 3d - hst survey ( van dokkum et al . 2011 ; brammer et al . 2012b ; skelton et al we study all five 3d - hst / candels fields in a homogeneous way , providing improved measurements of the number density of candidate compact galaxies in formation . we present extensive keck spectroscopy of a subset of these candidates , and measure redshifts , emission line widths , and emission line ratios . the h@xmath8 line profile and spatial extent is used to probe the potential beyond the stellar effective radius , allowing us to reconstruct the average rotation curve of this class of objects . in the second part of the paper we discuss a framework for the formation and evolution of massive galaxies that places the results of the keck spectroscopy in context . we show that , even though individual galaxies likely have complex formation histories , the evolution of the _ population _ of massive galaxies can be described with a simple model in which galaxies follow parallel tracks in the size - mass plane . for consistency with previous studies we assume @xmath24 , @xmath25 , and @xmath26mpc@xmath27 . we use data from the 3d - hst project ( van dokkum et al . 2011 ; brammer et al . 2012b ) to identify candidate compact massive galaxies . the 3d - hst catalogs ( skelton et al . 2014 ) provide multi - band photometry for objects in the five extra - galactic fields of the candels survey ( grogin et al . 2011 ; koekemoer et al . objects were selected using a signal - to - noise ( s / n ) optimized combination of the wfc3 @xmath28 , @xmath29 , and @xmath30 images . the catalogs encompass nearly all publicly available data in the candels fields , including deep irac data , as well as medium - band imaging in the optical and the near - ir . stars were excluded , as well as objects that have ( see skelton et al . 2014 ) . the imaging data are combined with 3d - hst wfc3 g141 grism spectroscopy , which together with data from program go-11600 covers @xmath31% of the candels photometric area ( see brammer et al . 2012b ) . the analysis of the combined photometric and spectroscopic dataset will be described in detail in i. momcheva et al . , in preparation . briefly , the photometric data from skelton et al . ( 2014 ) and the 2d grism data were fit simultaneously with a modified version of the eazy code ( brammer , van dokkum , & coppi 2008 ) to measure redshifts , rest - frame colors , and the strengths of emission lines ( brammer et al . if there are no significant emission or absorption features in the grism spectrum , or if no grism spectrum is available , the fit is similar to a standard photometric redshift analysis . in version 4.1.4 of our data release spectra are extracted only to @xmath32 ( and obviously only in the area covered by the grism observations ) . in addition to the skelton et al . photometric information and the grism spectroscopy we use spitzer mips 24@xmath33 m data to estimate total ir luminosities and star formation rates , as described in whitaker et al . ( 2012 , 2014 ) . these ir luminosities and star formation rates are consistent ( within a factor of @xmath9 ) with those derived from the full mid- and far - ir seds , at least for the ir - luminous galaxies that have reliable far - ir photometry ( see , e.g. , muzzin et al . 2010 ; elbaz et al . 2011 ; wuyts et al . 2011 ; utomo et al . 2014 ) . structural parameters of galaxies in the skelton et al.catalogs were measured by van der wel et al . ( 2014b ) , using the methodology described in van der wel et al . sizes , total luminosities , and ellipticities were measured from the wfc3 imaging using the galapagos implementation ( barden et al . 2012 ) of galfit ( peng et al . 2002 ) . in sect . 7.2 we show with a stacking analysis that the structural parameters in the van der wel et al . ( 2014b ) catalogs are reliable for the compact , massive galaxies studied in this paper . the catalog contains a small number of `` catastrophic '' failures . to identify these , we compared the total galaxy fluxes from the galfit fit to the total fluxes in the skelton et al . catalogs . galaxies were excluded from the analysis if the absolute difference between these two measurements exceeds 0.5 magnitudes . in this paper we use circularized half - light radii throughout , defined as @xmath34 with @xmath35 the half - light radius along the major axis and @xmath36 the axis ratio of the galaxy . the sizes are determined from data in the @xmath30 band , which corresponds to rest - frame @xmath37 at @xmath38 . finally , stellar masses were determined from fits of stellar population synthesis models to the 0.3@xmath33 m 8@xmath33 m photometry , as described in skelton et al . the fits were done with the fast code ( kriek et al . 2009b ) , using a chabrier ( 2003 ) imf , the calzetti et al . ( 2000 ) attenuation law , and exponentially - declining star formation histories . these parameters were chosen for consistency with previous studies ; small changes such as using `` delayed @xmath39 '' models do not change the masses significantly . in this paper we do not use the best - fitting star formation rates , ages , or extinction from these fits , as they tend to be less robust than the stellar masses ( see , e.g. , kriek et al . 2009b ; muzzin et al . a small ( typically @xmath40% ) correction was applied to each galaxy to make its half - light radius and stellar mass self - consistent : @xmath41 with @xmath42 the total @xmath43 band luminosity as implied by the galfit fit and @xmath44 the total @xmath43 band luminosity in the skelton et al . catalog ( see taylor et al . 2010a ; van dokkum et al . 2014 ) . in this paper we use the rest - frame colors of galaxies to separate ( candidate ) star forming galaxies from quiescent galaxies . as shown by labb et al . ( 2005 ) , wuyts et al . ( 2007 ) , whitaker et al . ( 2011 ) , and many others , galaxies occupy distinct regions in the space spanned by the rest - frame @xmath45 and @xmath46 colors , depending on their specific star formation rate . the reason is that dust and age have a subtly different effect on the spectral energy distributions ( seds ) of galaxies : galaxies that are young and dusty are red in both @xmath45 and @xmath46 , whereas galaxies that are old and dust - free are red in @xmath45 but ( relatively ) blue in @xmath46 . with high quality redshifts and photometry it has been demonstrated that there is a gap between the ( age-)sequence of quiescent galaxies and the ( dust-)sequence of star forming galaxies in the @xmath47 plane ( whitaker et al . 2011 ; brammer et al . 2011 ) , leading to a relatively unambiguous separation of the two galaxy classes . = 8.5 cm the distribution of galaxies with @xmath48 and @xmath6 in the @xmath47 plane is shown in fig . [ uvj.fig ] . the quiescent box is indicated with the black lines ; galaxies inside this box satisfy the equations @xmath49 galaxies are color - coded by their specific star formation rates , defined as ssfr = sfr/@xmath50 , with sfr the star formation rate derived from their uv+ir emission ( see whitaker et al . 2014 , and references therein ) . as can be seen in fig . 1 the @xmath47 selection corresponds very well to a selection on specific star formation rate . this was expected from previous studies ( e.g. , wuyts et al . 2011 ) ; nevertheless , the correspondence is striking as the mips 24@xmath33 m measurements ( which dominate the star formation rates in this stellar mass range ) are entirely independent from the @xmath45 and @xmath46 colors . we note that a subset of quiescent galaxies has high ssfrs in fig . [ uvj.fig ] ; these are galaxies whose rest - frame optical / near - ir seds show no signs of star formation even though they have high mips 24@xmath33 m fluxes . these galaxies are difficult to interpret : they may be quiescent galaxies with an active nucleus , or their star formation is so obscured that the young stars do not contribute significantly to the sed . fumagalli et al . ( 2014 ) show that the optical / near - ir seds of these galaxies are very similar to the ones that have no mips detection . approximately 20% of galaxies in the barro et al . ( 2013 ) sample fall in this category . of 582 galaxies with @xmath48 and @xmath6 , 185 ( 32% ) are quiescent and 397 ( 68% ) are star forming . the total area of the five fields is 896arcmin@xmath51 , and the number densities of massive quiescent galaxies and massive star forming galaxies are @xmath52mpc@xmath53 and @xmath54mpc@xmath53 respectively . these numbers are consistent with previous measurements from other datasets ( e.g. , marchesini et al . 2009 ; brammer et al . 2011 ; muzzin et al . 2013 ) . the size - mass relation for galaxies in the 3d - hst survey with @xmath6 is shown in fig . [ masssize.fig ] . quiescent and star forming galaxies , identified using eq . [ uvj.eq ] , are indicated with red and blue points respectively . as is well known , star forming galaxies are larger than quiescent galaxies at fixed mass ( e.g. , franx et al . 2008 ; williams et al . 2010 ; van der wel et al . note that the galaxy distribution in fig . [ masssize.fig ] is displaced with respect to that in fig . 5 of van der wel et al . ( 2014b ) , as we use circularized half - light radii and van der wel et al . use half - light radii along the major axis . = 8.5 cm compact massive galaxies ( cmgs ) are in the lower right portion of the size - mass diagram . barro et al . ( 2013 ) use the criterion @xmath55 to isolate compact galaxies ( dashed line in fig . [ masssize.fig ] ) . however , at masses of @xmath56 this selection does not produce a sample of compact star forming galaxies that is directly comparable to compact quiescent galaxies . the median size of quiescent galaxies with @xmath57 that satisfy the barro et al . compactness criterion is @xmath58kpc . the median size of star forming galaxies with @xmath57 that satisfy this criterion is @xmath59kpc . for comparison , the median size of the full sample of star forming galaxies with @xmath57 is @xmath60kpc . that is , at high masses , the barro et al . criterion selects star forming galaxies whose sizes are closer to those of the full sample of star forming galaxies than to those of compact quiescent galaxies . the reason is that the barro et al . `` compactness '' criterion is not very restrictive at high masses , as it selects 60% of all star forming galaxies that have @xmath57 . as our goal is to select plausible progenitors of massive , compact quiescent galaxies we adopt a slightly more restrictive criterion : @xmath61 with @xmath50 in units of m@xmath62 and @xmath63 in units of kpc . this limit is indicated by the solid diagonal line in fig.[masssize.fig ] . thirty - nine percent of star forming galaxies with @xmath57 satisfy this criterion and their median size is @xmath64kpc . as we discuss below , the slope of unity of our compactness criterion can be readily interpreted in terms of a physical parameter , namely the velocity dispersion . the slope of @xmath65 used by barro et al . ( 2013 ) was chosen to be consistent with the slope of the size - mass relation of quiescent galaxies as found by newman et al . we note that van der wel et al . ( 2014b ) find a slightly steeper slope than newman et al . ( 2012 ) at @xmath66 ( @xmath67 versus @xmath68 ) . in addition to their compactness criterion barro et al . apply a mass limit of @xmath69 . this relatively low limit is also used for their comparison samples of quiescent galaxies and spatially - extended star forming galaxies . however , very few galaxies that have @xmath70 at @xmath15 will grow into @xmath0 galaxies by @xmath16 ( e.g. , van dokkum et al . 2010 ; leja , van dokkum , & franx 2013a ; behroozi et al . we therefore apply a mass limit that is higher by a factor of 4 : @xmath48 . this selection produces homogeneous samples of _ massive _ compact galaxies . another consideration when choosing this mass limit is that sizes are uncertain when the effective radius is significantly smaller than the pixel size ( the drizzled pixel size is @xmath71 , corresponding to 0.5kpc at @xmath15 ) . in the remainder of the paper we will use `` cmg '' , for `` compact massive galaxy '' , to denote objects with @xmath48 and @xmath72 . based on their location in the @xmath47 diagram we distinguish `` qcmg '' , for quiescent compact massive galaxy , and `` scmg '' , for star forming compact galaxy . there are 112 scmgs at @xmath6 in the five 3d - hst / candels fields . five of these have effective radii @xmath73kpc ; when calculating dynamical masses and expected velocity dispersions of these galaxies we use 0.5kpc instead of their best - fitting radius . it should be noted that many of the star forming progenitors of @xmath74 qcmgs are expected to be at higher redshift than @xmath75 ; we discuss the evolution of scmgs and qcmgs in sections 7 and 8 . we quantify the compactness of galaxies by their expected galaxy - integrated velocity dispersion , as this quantity follows directly from our size - mass selection and can be compared to observations ( see sect . [ sigcomp.sec ] ) . for simplicity , we use the following relation : @xmath76 with @xmath77 the predicted velocity dispersion in , @xmath50 in units of m@xmath62 , and @xmath63 in units of kpc ( franx et al . 2008 ; van dokkum , kriek , & franx 2009 ) . this relation has been shown to reasonably predict the observed stellar velocity dispersions of both quiescent galaxies and star forming galaxies , at least in the regime where this has been tested : out to @xmath78 for massive star forming galaxies ( taylor et al . 2010a ; bezanson , franx , & van dokkum 2015 ) and out to @xmath18 for massive quiescent galaxies ( bezanson et al . 2013 ; van de sande et al . 2013 ; belli et al . 2014a ) . our compactness criterion ( eq . [ selcompact.eq ] ) corresponds to @xmath79 , or @xmath80 . the distributions of predicted dispersions of scmgs and qcmgs are shown by the histograms in fig.[sigpred.fig ] . the median expected dispersions of the two populations are similar but not identical : @xmath81 for quiescent galaxies and @xmath82 for star forming galaxies . the reason for this difference is that the size distribution of quiescent galaxies is different from that of star forming galaxies . for star forming galaxies we select the tail of the distribution , with the largest number of galaxies close to the compactness cutoff , whereas for quiescent galaxies we select the bulk of the population ( see van der wel et al . 2014b , for a discussion of the form of the size distributions of quiescent and star forming galaxies ) . phrased differently , irrespective of the exact compactness criterion , the smallest galaxies tend to be quiescent . we will return to this in sect . [ simplemod.sec ] , where we define a `` quenching line '' just inside the compact massive galaxy box . as shown in taylor et al . ( 2010a ) , the residuals between expected and observed dispersions correlate with the sersic index . the lines in fig . [ sigpred.fig ] show the distributions when the sersic index of the galaxies is taken into account , using @xmath83 with @xmath84 ( cappellari et al . 2006 ) . here @xmath85 is the sersic index and @xmath86 when @xmath50 is in units of , @xmath87 is in kpc , and @xmath77 is in kms@xmath27 . s have a slightly smaller median sersic index ( @xmath88 ) than s ( @xmath89 ) . for quiescent galaxies the line and histogram are nearly the same , but for star forming galaxies the sersic - dependent dispersions are on average @xmath90% lower than those calculated with eq . [ sigpred.eq ] . = 8.5 cm the number density of s and s is the same , @xmath91mpc@xmath53 ( for reference , the number density of the full population of quiescent galaxies with @xmath48 is @xmath52mpc@xmath53 ; see sect . [ uvj.sec ] ) . this result is consistent with previous studies that noted the overlap of the compact tail of star forming galaxies and the bulk of the quiescent population ( barro et al . 2013 ; van der wel et al . we therefore confirm that a population of star forming galaxies can be identified at @xmath6 that has a median mass , median size , and number density similar to the population of massive quiescent galaxies at the same redshifts . if all these compact star forming galaxies quench in the near future , the number density of massive quiescent galaxies will increase by 70% , and the number density of s will double . lcccc aegis_9163 & 14@xmath9221@xmath930368 & 5304373 & 25.8 & 23.2 + aegis_26952 & 14@xmath9220@xmath934081 & 5304519 & 25.2 & 22.2 + aegis_41114 & 14@xmath9218@xmath933292 & 5246067 & 25.1 & 22.7 + cosmos_163 & 10@xmath9200@xmath932501 & 210441 & 25.9 & 23.2 + cosmos_1014 & 10@xmath9200@xmath933592 & 211278 & 23.1 & 21.5 + cosmos_11363 & 10@xmath9200@xmath932871 & 217454 & 24.2 & 21.3 + cosmos_12020 & 10@xmath9200@xmath931791 & 218072 & 25.8 & 22.0 + cosmos_22995 & 10@xmath9200@xmath931715 & 224523 & 24.6 & 22.1 + cosmos_27289 & 10@xmath9200@xmath934158 & 227515 & & 22.1 + goods - n_774 & 12@xmath9236@xmath932773 & 6207128 & 27.1 & 23.0 + goods - n_6215 & 12@xmath9236@xmath930686 & 6210214 & 25.2 & 21.5 + goods - n_13616 & 12@xmath9236@xmath930633 & 6212329 & 25.9 & 22.8 + goods - n_14283 & 12@xmath9237@xmath930260 & 6212440 & 25.0 & 22.9 + goods - n_22548 & 12@xmath9237@xmath930046 & 6215089 & 25.5 & 22.5 + goods - s_5981 & 3@xmath9232@xmath931455 & @xmath9452565 & 24.9 & 22.4 + goods - s_30274 & 3@xmath9232@xmath933146 & @xmath9446232 & 23.5 & 21.3 + goods - s_37745 & 3@xmath9232@xmath934388 & @xmath9444057 & 24.1 & 22.0 + goods - s_45068 & 3@xmath9232@xmath933302 & @xmath9442004 & 25.0 & 22.5 + goods - s_45188 & 3@xmath9232@xmath931518 & @xmath9441587 & 25.4 & 22.9 + uds_16442 & 2@xmath9217@xmath932080 & @xmath9513160 & 27.4 & 23.4 + uds_25893 & 2@xmath9218@xmath930297 & @xmath9511213 & & 23.1 + uds_26012 & 2@xmath9217@xmath930366 & @xmath9511222 & 25.4 & 22.4 + uds_33334 & 2@xmath9216@xmath935501 & @xmath9509528 & 26.2 & 23.3 + uds_35673 & 2@xmath9217@xmath930533 & @xmath9509257 & 25.1 & 22.4 + uds_42571 & 2@xmath9217@xmath934395 & @xmath9507513 & 27.0 & 22.8 we observed candidate scmgs with the near - ir spectrographs mosfire ( mclean et al . 2012 ) and nirspec ( mclean et al . 1998 ) on keck in 2014 and 2015 . the resulting spectra provide spectroscopic redshifts ( measured from and at @xmath96 ) , which can be used to verify that a population of scmgs exists at these redshifts . furthermore , the spectroscopic observations provide galaxy - integrated kinematics of the ionized gas : if compact star forming galaxies are in the process of forming the stars that are later in compact quiescent galaxies , their gas kinematics should be similar to the stellar kinematics of quiescent galaxies . in addition to redshifts and kinematics the spectra provide star formation rates and strong line ratios ; these are important for understanding the physical processes that take place in these galaxies , although their interpretation is often not unique . the mosfire spectra were obtained in three separate observing runs : january 11 , 12 2014 ; april 18 , 23 , 25 2014 ; and dec 12 , 13 , 15 2014 . the january run suffered from clouds and poor seeing ; conditions were generally good during the other two runs . compact , massive star forming galaxies were not always the main targets , and were not always selected using the criteria of sect . 2.2 . one target from the april run , a galaxy at @xmath97 , is described in oesch et al . the december run gave higher priority to galaxies at @xmath98 than to galaxies at lower redshift . in this paper we will limit the discussion to star forming galaxies at @xmath74 that satisfy the criteria of sect . [ candidates.sec ] . the observations were all taken in the @xmath5-band , using a standard abab dither pattern . the exposure times varied from @xmath99hr to @xmath14hrs , depending on conditions and the requirements imposed by the primary targets in the masks . one of the slits in each mask was devoted to a relatively bright , relatively blue star . this has four important functions : the s / n ratio of the star is used to weight individual exposures in the reduction ; the @xmath100position of the star is used to correct the data for small vertical drifts of the mask relative to the sky ( see kriek et al . 2015 ) ; the extracted spectrum is used to identify regions of strong sky absorption ; and the width of the 2d stellar spectrum in the spatial direction provides us with a model of the point spread function ( psf ) that is otherwise very difficult to construct ( see sect . [ morphline.sec ] ) . the data reduction used the standard mosfire pipeline drp , with small modifications ( see oesch et al . individual sequences were reduced and shifted to a common reference frame before stacking . one - dimensional spectra were obtained from the 2d spectra by summing rows , as dictated by the observed spatial extent of the galaxies . for each mask an empirical noise spectrum was created by removing all rows with signal , and determining the width of the pixel distribution of the remaining rows for each pixel in the wavelength direction . the width was measured by removing the lowest and highest 16% of values , and is therefore equivalent to the @xmath101 width of a gaussian . for each individual galaxy in a mask the noise spectrum was multiplied by the square root of the number of rows that was summed to create the 1d spectrum of that object . the nirspec data were obtained in two runs , january 10 , 13 , 14 2014 and january 25 , 26 2015 . conditions were poor in the 2014 run and the only object in our final sample that came from it is goods - n_774 , which was published in nelson et al . conditions in 2015 were excellent , with the seeing ranging from @xmath102 during both nights . the selection for the nirspec runs was very similar to that described in sect . [ candidates.sec ] ; within these criteria priority was generally given to galaxies with higher star formation rates ( and with good blind offset stars ; see below ) . we followed standard observing procedures for nirspec spectroscopy of faint targets ( see , e.g. , erb et al . 2003 ; van dokkum et al . target aquisition was done with blind offsets from nearby stars , as the galaxies are not detected in the scam slit - viewing camera . the n6 filter was used for goods - n_774 ; all data in the 2015 run were taken with the n7 filter . a typical observing sequence consisted of four 900s exposures in an abba pattern with @xmath103 offsets between nods . the data were continuously inspected as objects sometimes drift out of the slit . the data reduction followed standard procedures for near - ir , single slit data ( see , e.g. , van dokkum et al . the data were initially reduced in pairs , using the sky of the a frame for the b frame and vice versa . this method yields relatively clean , photon noise - dominated spectra , at the expense of reducing the s / n in the final frames by @xmath104 ( see , e.g. , kriek et al . wavelength calibration was done using sky lines , which were also used to determine the spectral resolution of the data ( see sect . [ fitting.sec ] ) . the slit is not long enough to obtain an accurate noise spectrum from empty regions ; therefore , we calculate the noise spectrum from the sky spectrum and the noise in the darks . an analysis of the residuals from fits to the emission lines shows that this is sufficient for our purposes ( see sect . [ fitting.sec ] ) . we identify the redshifted and emission lines in 20 out of 24 compact , massive star forming galaxies with expected redshifts in the range @xmath6 . this success rate of 86% is encouraging , but it should be noted that our selection at the telescope was somewhat subjective , particularly in the nirspec runs . as an example , if there were two plausible targets and one showed a hint of an h@xmath8 contribution to the broad band flux we would generally give it preference . additionally , there are five non - overlapping galaxies in barro et al . ( 2014b ) that satisfy our criteria ( see sect . 3.5 ) ; the total sample of massive compact star forming galaxies with measurements is therefore 25 ( table 1 ) . the properties of the galaxies in the spectroscopic sample are compared to the parent sample in fig . [ select.fig ] . the median size and mass are @xmath58kpc and @xmath105respectively , close to the medians of the parent sample . the spread is somewhat smaller ; 24 out of 25 galaxies are in the mass range @xmath106 . the galaxies have bluer @xmath45 colors and slightly higher uv+ir star formation rates than the parent sample . this is by selection : galaxies with specific star formation rates @xmath107yr@xmath27 were given lower priority . despite the lack of galaxies with low star formation rates in the spectroscopic sample , the median ssfr is only 0.1dex higher than that of the parent sample ( @xmath108yr@xmath27 compared to @xmath109yr@xmath27 for the parent sample ) . both medians are close to the whitaker et al . ( 2014 ) main sequence for this redshift ( dark grey line in fig . [ select.fig]c ) . panel d of fig . [ select.fig ] shows the dust content of the galaxies , as parameterized by both the ratio of the ir and uv luminosities and the rest - frame @xmath46 color . galaxies in the upper right part of this panel are very dusty , with the re - radiated ir emission exceeding the uv emission by a factor of @xmath110 . the median @xmath111 ratio of the parent sample is @xmath112 . the median ratio for the galaxies in the spectroscopic sample is slightly lower , at 42 . we only have a few spectroscopic objects in this part of the diagram , and all four spectroscopic failures are located here . we infer that the most likely explanation for the failures is that the h@xmath8 emission in these galaxies is too obscured for a detection in our current observations . the keck spectra of the 20 galaxies that we observed are shown in fig . [ spectra.fig ] . the galaxies are ordered by the measured velocity dispersion ( see below ) . we include the five objects from barro et al . ( 2014b ) that satisfy our selection criteria ; as we can not show the spectra of these objects in fig . [ spectra.fig ] , we instead show models that are based on their published best - fitting parameters . figures [ hstim.fig ] and [ seds.fig ] show the hst images and the rest - frame uv near - ir spectral energy distributions ( seds ) of the 25 galaxies of fig . [ spectra.fig ] . the @xmath30 images are shown separately at high dynamic range in appendix a. the seds range from relatively unobscured ( cosmos_1014 ) to extremely dusty ( e.g. , goods - n_774 ) . some have excess emission in the irac bands ( uds_42571 ; see , e.g. , mentuch et al . 2009 ) two galaxies show clear signs of merging : cosmos_11363 is an ongoing merger between two compact massive galaxies that are only @xmath113 apart , and goods - s_30274 is probably a merger remnant ( see sect . [ sb.sec ] ) . interestingly , there is no clear relation between the measured velocity dispersion and either the morphology or the sed . phrased differently , it is not possible to predict the h@xmath8 line width based on the information shown in figs . [ hstim.fig ] and [ seds.fig ] . the spectra were fitted with a model that has the redshift , the continuum level , the and h@xmath8 line fluxes , and the line width as free parameters . the instrumental resolution is explicitly taken into account . the model has the following form : @xmath114 with @xmath115 the model for the line emission , @xmath116 the instrumental resolution , @xmath117 the continuum level , and @xmath118 denoting convolution . the instrumental resolution is modeled with a gaussian : @xmath119 with @xmath120 measured from sky lines in the vicinity of the redshifted h@xmath8 line , @xmath121 the pixel size in , and @xmath122 the center of the fitting range . expressed as a velocity , the resolution of the mosfire spectra is @xmath123 , and the resolution of the nirspec data is @xmath31 . the lines are parameterized as follows : @xmath124 } \left ( l_{6584}(\lambda ) + \frac{1}{3}l_{6548(\lambda ) } , \right)\ ] ] with @xmath125 here @xmath126 is the line strength , @xmath127 is the galaxy - integrated line - of - sight velocity dispersion , @xmath128 is the rest - frame wavelength of the line ( with @xmath129 and @xmath130 , @xmath131 for h@xmath8 and the two [ nii ] lines respectively ) , and @xmath132 is the redshift . some galaxies show evidence for multiple velocity components ( e.g. , cosmos_1014 ) . we do not attempt to separately fit broad and narrow velocity components to these galaxies ( as was done by , e.g. , frster schreiber et al . as discussed later , broad components could indicate the presence of winds but could also indicate rapidly rotating gas at small radii in the galaxies . in the absense of high spatial resolution data , it is difficult to distinguish these possibilities ; we therefore simply interpret the h@xmath8-luminosity - weighted velocity profiles in this paper . it should be noted that the formal uncertainties underestimate the error in the velocity dispersion if the velocity distribution is not gaussian . this is particularly important for galaxies with a high s / n ratio , such as cosmos_12020 . the emcee mcmc algorithm ( foreman - mackey et al . 2013 ) was used to fit this model to the galaxy spectra . the fit was done over the wavelength region @xmath133 ; the results are not dependent on the choice of fitting region as long as the continuum is reasonably well covered . priors are top hats with boundaries that comfortably encompass the fitting results . that is , the bayesian aspects of emcee were essentially turned off . we used 100 walkers and generated 500 chains in each fit . burn - in was typically fast , but we removed the first 200 chains when calculating errors . for each fit parameter the best fit is defined as the median of the 300 remaining samples . errors were determined from the 16@xmath134 and 84@xmath134 percentiles ( see foreman - mackey et al . 2013 , for details ) . the best fit models are shown by red lines in fig.[spectra.fig ] . residuals from the fits are shown in fig . [ spectra_noise.fig ] . as discussed in appendix [ noise.sec ] the residuals are consistent with the expected noise in almost all cases . the redshifts and velocity dispersions follow directly from the mcmc fit , but the line fluxes , equivalent widths , and line ratios need to be calibrated or corrected . the continuum is detected for every galaxy , which makes it possible to calculate equivalent widths directly from the spectra . the equivalent widths , in turn , enable us to calibrate the line fluxes using the known @xmath5-band magnitudes of the galaxies . the equivalent width of h@xmath8 in the observed frame is given by @xmath135 the second term is a correction for the underlying stellar continuum absorption , which has a non - negligible effect on the measured equivalent widths and line ratios in our sample . we adopt @xmath136(moustakas & kennicutt 2006 ; alonso - herrero et al . the relation between rest - frame equivalent width and the observed equivalent width is @xmath137 the mean rest - frame equivalent width in our sample is @xmath138 , consistent with the general population of ( detected ) massive star forming galaxies at these redshifts ( fumagalli et al . 2012 ) . the ratio , corrected for absorption , is @xmath139}{{\rm h}\alpha } = \frac{f_{[{\rm n}{\textsc{ii}}]}}{f_{{\rm h}\alpha } } \times \frac{{\rm ew}_{{\rm h}\alpha}^0 - { \rm ew}_{{\rm h}\alpha,\,{\rm abs } } } { { \rm ew}_{{\rm h}\alpha}^0},\ ] ] with @xmath126 taken from the mcmc fit . note that we use positive values for both absorption equivalent widths and emission equivalent widths in these expressions , as `` absorption '' here is more accurately described as `` emission that is filling in the underlying absorption line '' . the line flux is calculated from the observed equivalent width and the @xmath5 magnitude using @xmath140 with @xmath141 the ab magnitude of the object and @xmath142 in units of ergss@xmath27@xmath143 . this expression ignores small differences between the filters used in each field as well as the detailed shape of the continuum within the @xmath141 filter . we verified that the transmission at the observed wavelenghts of the lines is within @xmath40% of the central transmission of the filter in all cases . finally , the line luminosity is calculated using @xmath144 with @xmath145 the luminosity distance in mpc and @xmath146 in ergss@xmath27 . the results for all galaxies are listed in table 2 . the error bars reflect the ( propagated ) mcmc errors ; no additional calibration uncertainty was included in the error budget . there are seven galaxies in the barro et al . ( 2014b ) sample that satisfy our more restrictive selection criteria . two of these seven galaxies , cosmos_12020 and goods - s_37745 , are also in our sample : cosmos_12020 was observed with nirspec and goods - s_37745 with mosfire . for cosmos_12020 we find @xmath147 and @xmath148 , whereas barro et al . have @xmath149 and @xmath150 . the kinematics of this galaxy are very complex , and a gaussian is a poor fit ( see fig . [ spectra.fig ] , and sect . [ complex.sec ] ) ; this probably explains the differences between the two measurements and the large uncertainty in the barro et al . values . as noted in sect . 3.4.1 the formal uncertainty in our measurement of this galaxy is smaller than the true uncertainty , as it does not take deviations from a gaussian into account . given that a gaussian is clearly a poor fit , the velocity dispersion of this galaxy is not well determined . for goods - s_37745 we find @xmath151 and @xmath152 , compared to @xmath153and @xmath154 in barro et al . ( 2014b ) . these values are in agreement within the ( relatively large ) @xmath155 uncertainties . for the two galaxies that overlap we use our own measurements . the other five galaxies from barro et al . are added to our sample ( see tables 1 and 2 ) . we do not have measurements of the line flux or spatial extent of the emission line gas for these objects , but they are included in the analysis whenever only the redshift , velocity dispersion , or line ratio is needed . they are shown in fig.[spectra.fig ] by their best - fitting models . the total number of scmgs at @xmath6 that are studied in this paper is 25 . lccccccccccccc aegis_9163 & 2.445 & 0.8 & 0.9 & 5.4 & 0.72 & 131 & 1.81 & & nirs & @xmath156 & @xmath157 & @xmath158 & @xmath159 + aegis_26952 & 2.097 & 1.1 & 1.8 & 3.6 & 0.64 & 148 & 1.62 & yes & nirs & @xmath160 & @xmath161 & @xmath162 & @xmath163 + aegis_41114 & 2.332 & 0.5 & 0.2 & 8.0 & 0.62 & 95 & 1.38 & & nirs & @xmath164 & @xmath165 & @xmath166 & @xmath167 + cosmos_163 & 2.312 & 0.8 & 1.1 & 2.5 & 0.60 & 336 & 2.25 & yes & mosf & @xmath168 & @xmath169 & @xmath170 & @xmath171 + cosmos_1014 & 2.100 & 0.5 & 0.7 & 8.0 & 0.79 & 150 & 0.93 & & nirs & @xmath172 & @xmath173 & @xmath174 & @xmath175 + cosmos_11363 & 2.096 & 1.1 & 2.1 & 5.2 & 0.76 & 169 & 1.31 & yes & nirs & @xmath176 & @xmath177 & @xmath178 & @xmath179 + cosmos_12020 & 2.094 & 2.0 & 2.1 & 5.7 & 0.57 & 185 & 1.96 & yes & nirs & @xmath180 & @xmath181 & @xmath182 & @xmath183 + cosmos_22995 & 2.469 & 1.2 & 1.1 & 2.8 & 0.67 & 188 & 1.41 & yes & nirs & @xmath184 & @xmath185 & @xmath186 & @xmath187 + cosmos_27289 & 2.234 & 1.3 & 2.3 & 3.3 & 0.81 & 398 & 2.02 & & nirs & @xmath188 & @xmath189 & @xmath190 & @xmath191 + goods - n_774 & 2.301 & 1.0 & 1.0 & 2.9 & 0.59 & 150 & 2.07 & & nirs & @xmath192 & @xmath193 & @xmath194 & @xmath195 + goods - n_6215 & 2.321 & 1.8 & 1.8 & 2.6 & 0.72 & 110 & 1.28 & yes & mosf & & & @xmath196 & @xmath197 + goods - n_13616 & 2.487 & 1.1 & 1.9 & 5.6 & 0.97 & 130 & 1.79 & & mosf & & & @xmath198 & @xmath199 + goods - n_14283 & 2.420 & 0.9 & 1.2 & 2.7 & 0.86 & 111 & 1.43 & yes & mosf & & & @xmath200 & @xmath201 + goods - n_22548 & 2.330 & 1.0 & 1.7 & 5.9 & 0.78 & 120 & 1.53 & yes & mosf & & & @xmath202 & @xmath203 + goods - s_5981 & 2.253 & 0.8 & 0.8 & 4.4 & 0.85 & 206 & 1.75 & & mosf & @xmath204 & @xmath205 & @xmath206 & @xmath207 + goods - s_30274 & 2.226 & 1.4 & 2.5 & 8.0 & 0.46 & 404 & 1.47 & yes & mosf & @xmath208 & @xmath209 & @xmath210 & @xmath211 + goods - s_37745 & 2.432 & 0.9 & 0.6 & 3.6 & 0.94 & 118 & 1.04 & & mosf & @xmath212 & @xmath213 & @xmath214 & @xmath215 + goods - s_45068 & 2.453 & 1.1 & 1.3 & 4.9 & 0.97 & 139 & 1.57 & & mosf & & & @xmath216 & @xmath217 + goods - s_45188 & 2.407 & 0.7 & 1.4 & 4.3 & 0.90 & 134 & 1.66 & yes & nirs & @xmath218 & @xmath219 & @xmath220 & @xmath221 + uds_16442 & 2.218 & 1.7 & 3.3 & 1.6 & 0.52 & 176 & 2.36 & & mosf & @xmath222 & @xmath223 & @xmath224 & @xmath225 + uds_25893 & 2.304 & 0.6 & 0.2 & 8.0 & 0.92 & 73 & 1.88 & yes & mosf & @xmath226 & @xmath227 & @xmath228 & @xmath229 + uds_26012 & 2.321 & 1.3 & 2.6 & 3.5 & 0.73 & 109 & 1.47 & & mosf & @xmath230 & @xmath231 & @xmath232 & @xmath233 + uds_33334 & 2.290 & 0.7 & 1.4 & 2.4 & 0.56 & 13 & 1.01 & & mosf & @xmath234 & @xmath235 & @xmath236 & @xmath237 + uds_35673 & 2.182 & 0.9 & 0.7 & 6.4 & 0.75 & 492 & 2.18 & & mosf & @xmath238 & @xmath239 & @xmath240 & @xmath241 + uds_42571 & 2.292 & 1.6 & 2.3 & 1.9 & 0.82 & 388 & 2.39 & yes & nirs & @xmath242 & @xmath243 & @xmath244 & @xmath245 considering that the 25 s of fig . [ spectra.fig ] were selected in a very restricted region of parameter space , their emission lines show a surprisingly large range of properties . the velocity dispersions range from 50to @xmath246 , the ratios from 0.2 to @xmath247 , and the h@xmath8 line luminosities from @xmath248l@xmath62 to @xmath249l@xmath62 . two of these parameters , the ratio and the velocity dispersion , show a significant correlation : as shown in fig . [ lrsig.fig ] , galaxies with the highest velocity dispersions tend to have the highest line ratios . the correlation has a formal significance of @xmath250% . the broken line is the best fit relation , which has the form @xmath251}{{\rm h}\alpha } = ( -0.51 \pm 0.08 ) + ( 1.0 \pm 0.2 ) \log \left ( \frac{\sigma_{\rm gas}}{100}\right).\ ] ] the canonical high - metallicity saturation value for in low redshift star forming galaxies is @xmath252 ( e.g. , baldwin , phillips , & terlevich 1981 ; denicol ' o , terlevich , & terlevich 2002 ; pettini & pagel 2004 ; kewley et al . although this limit is observed to be higher at @xmath4 ( e.g. , brinchmann , pettini , & charlot 2008 ; steidel et al . 2014 ; shapley et al . 2015 ) , values of @xmath253 are extreme at any redshift ( see , e.g. , leja et al . 2013b ; shapley et al . 2015 ) . a likely explanation for the highest @xmath127 , highest galaxies in fig . [ lrsig.fig ] is that shocks ( dopita & sutherland 1995 ) and/or emission from agns ( kewley et al . 2013 ) are responsible for the line ratios . = 8.5 cm this is supported by the x - ray luminosities of the objects , obtained from all public catalogs in the candels fields . twelve of the 25 s ( 48% ) have @xmath254ergss@xmath27 and are classified as agn . the x - ray luminosities range from @xmath255ergss@xmath27 for goods - s_30274 to @xmath256ergss@xmath27 for cosmos-11363 . this high agn fraction is consistent with previous studies of massive star forming galaxies at these redshifts ( e.g. , papovich et al . 2006 ; daddi et al . 2007 ; kriek et al . 2007 ; barro et al . 2013 ; frster schreiber et al . the four galaxies with the highest velocity dispersions are all classified as x - ray agn . is no longer significant when these four objects are removed . ] their kinematics are complex ( see fig . [ spectra.fig ] ) , and their ratios range from 0.8 to 2.2 . it is likely that the observed emission line properties of these galaxies are affected by the presence of the agn , either directly through emission from the broad line region or indirectly through agn - driven winds ( see frster schreiber et al . 2014 ; genzel et al . 2014a ) . however , it is not clear whether agns or winds _ dominate _ the observed , galaxy - integrated kinematics , even for these four objects and whether the presence of a central point source influenced their selection as apparently compact , apparently massive galaxies . as shown in fig . [ seds.fig ] the uv near - ir seds of all galaxies are well fit by stars - only models . most galaxies have strong balmer breaks ( including the most powerful x - ray source in the sample , cosmos-11363 ) , and as discussed in kriek et al . ( 2007 ) and later studies ( e.g. , marsan et al . 2015 ) this strongly constrains the contribution of continuum emission from an agn at @xmath257 . as we show below and in the following section , the properties of most of the galaxies can be understood in a model where agn are present but do not dominate the kinematics , line ratios , line luminosities , or morphology . the h@xmath8 luminosities can be converted to star formation rates if it is assumed that the h@xmath8 emission largely originates in hii regions . by comparing these star formation rates to those derived from the uv and the bolometric uv+ir luminosities we can assess whether this assumption is reasonable , and also constrain the amount of obscuration in the galaxies . the h@xmath8 star formation rates were determined using the kennicutt ( 1998 ) relation , converted to a chabrier ( 2003 ) imf . the uv luminosities come from the best - fitting brammer et al . ( 2008 ) models at @xmath258 , and the ir luminosities are converted spitzer / mips 24@xmath33 m fluxes ( see whitaker et al . 2012 and sect . the relation between the uv / uv+ir star formation rates and the h@xmath8 star formation rate is shown in fig . [ sfrsfr.fig ] . only the 20 galaxies from our own spectroscopy are considered here , as we do not have self - consistent measurements of @xmath146(h@xmath8 ) for the five objects from barro et al . ( 2014b ) . the h@xmath8 star formation rates range from 6/yr 58/yr . they correlate with the uv star formation rates ( 98% significance ) and with the uv+ir star formation rates , which are dominated by the ir ( 96% significance ) . the mean offset between sfr(h@xmath8 ) and sfr(uv ) is @xmath259 dex , with an rms scatter of 0.22 dex . the offset between sfr(h@xmath8 ) and sfr(uv+ir ) is @xmath260 dex , with a scatter of 0.27 dex . the implication is that the h@xmath8 emission misses @xmath261% of the star formation , and the uv misses @xmath262% . the ratios between the three indicators are broadly consistent with expectations from a calzetti et al . ( 2000 ) reddening curve , if there is @xmath263% more dust toward nebular emission line gas than toward the uv continuum . = 8.5 cm the x - ray agns are indicated by orange points in fig . [ sfrsfr.fig ] . remarkably , they are indistinguishable from the other objects : they span the same range in h@xmath8 luminosity , and they follow the same relations with the uv and uv+ir luminosities . the offsets between the agn and non - agn are consistent with zero . this suggests , but does not prove , that the h@xmath8 , uv , and ir luminosities of most galaxies are dominated by star formation . the velocity dispersions we measure come from gaussian fits to the galaxy - integrated , luminosity - weighted h@xmath8 line profile and are equivalent to the second moment of the velocity distribution of the gas . they should not be confused with the rotation - corrected gas dispersions within spatially - resolved disks , such as discussed by , e.g. , kassin et al . ( 2012 ) and frster schreiber et al . the measured dispersions are a complex function of the dynamics and gas distribution in the galaxies : @xmath264 with @xmath265 ( franx 1993 ; rix et al . 1997 ; weiner et al . 2006 ; see also appendix c ) , @xmath266 the inclination of the galaxy ( @xmath267 is face - on , and @xmath268 is edge - on ) , @xmath269 the galaxy - integrated dispersion within the gas clouds , and @xmath270 an inclination - dependent term that takes non - gravitational motions into account . a further complication is that eq . [ siggen.eq ] is the result of an integral over the area of the galaxy that falls within the slit , weighted by the spatially - varying luminosity of the h@xmath8 line . we first assume that the gas in the s `` behaves '' in a similar way as the stars in s. that is , we assume that the stars in s were formed directly out of the ( detected ) gas in s , such that they have the same distribution and kinematics . this has been assumed in previous studies of the kinematics of compact massive star forming galaxies ( nelson et al . 2014 ; barro et al . 2014b ) and it may be reasonable if many compact , massive quiescent galaxies are direct descendants of the s. as discussed in sect . [ predsig.sec ] the stellar velocity dispersions of quiescent galaxies can be predicted from their stellar masses and effective radii ( e.g. , taylor et al . 2010a ; bezanson et al . 2011 ; belli et al . figure [ sigsig.fig]a shows the relation between @xmath271 and the predicted velocity dispersion . the predicted dispersions were calculated using the sersic - dependent relation eq . [ sigpredn.eq ] . there is no significant correlation between @xmath271 and @xmath77 , for either the full sample or the sample with the x - ray agn removed . the rms scatter in @xmath272 is 0.26 dex . given that we are ignoring the effects of non - gravitational motions , it is striking that many galaxies have _ lower _ velocity dispersions than the expectations . the mean offset is @xmath273 dex for the full sample , and @xmath274 dex when the agn are excluded . these results stand in sharp contrast to the stellar velocity dispersions of quiescent galaxies . red squares are seven galaxies with @xmath74 and measured @xmath275 , @xmath87 , @xmath85 , and @xmath50 from van dokkum et al . ( 2009 ) , van de sande et al . ( 2013 ) , and belli et al . ( 2014b ) . they have a mean offset in @xmath276 of @xmath277 dex and an rms scatter of only @xmath278 dex . as dynamical mass is proportional to @xmath279 the offsets of the s are even more dramatic in fig . [ sigsig.fig]b , which shows the relation between dynamical mass and stellar mass . here dynamical mass was calculated using @xmath280 as derived by cappellari et al . ( 2006 ) and following studies of quiescent galaxies at high redshift ( e.g. , van de sande et al . 2013 ) . for s @xmath281 and for quiescent galaxies @xmath282 . note that , given our definition of @xmath77 ( see eq . [ sigpredn.eq ] ) , panels a and b of fig . [ sigsig.fig ] are two different ways of presenting the same information . the mean mass offset of the s is @xmath274 dex for the full sample , and @xmath283 dex for the sample with the agn removed . that is , the dynamical masses of the non - agn galaxies are on average a factor of two lower than the stellar masses . several of the galaxies have apparent dynamical masses that are a factor of @xmath284 lower than their stellar masses . again , the quiescent galaxies show a tight relation in fig.[sigsig.fig]b , with a mean offset of @xmath285 dex . we conclude that the gas dynamics of s are _ not _ similar to the stellar dynamics of quiescent galaxies in the same mass and redshift range . the stellar masses and sizes are not useful indicators of the observed gas velocity dispersions ; in fact , the observed ratio is a better predictor of the observed h@xmath8 linewidth of a galaxy than its compactness is . there are many ways to _ increase _ the velocity dispersion of a galaxy so it falls above the lines of equality in the two panels of fig . [ sigsig.fig ] : the broad line region of an agn , agn - induced winds , and supernova - driven winds can all lead to broad h@xmath8 lines ( e.g. , westmoquette et al . 2009 ; le tiran et al . 2011 ; frster schreiber et al . 2014 ; banerji et al . this is likely the case for several galaxies in the sample : the four galaxies with the largest dynamical masses are all x - ray agn with ratios in the range @xmath286 . however , it is difficult to _ decrease _ the observed dispersion . setting aside the possibility that the stellar masses of some galaxies could be in error by a factor of @xmath7 , this is only possible if the detected ionized gas is s is distributed very differently from the stars in quiescent galaxies . as we show below , there is strong evidence that this is indeed the case . a possible interpretation of the large range of velocity dispersions is that the dynamics are dominated by rotation , and we are seeing disks under a large range of viewing angles . in fig . [ incl.fig]a we show the distribution of projected axis ratios @xmath287 in our sample , as determined from the @xmath30 data ( see van der wel et al . . the axis ratios of the 25 galaxies are inconsistent with a uniform distribution , which would be expected for thin , randomly oriented disks . we find no galaxies with @xmath288 and the distribution peaks at @xmath289 . the distribution is consistent with that observed for s , shown by the red line in fig . [ incl.fig]a : according to the kolgomorov - smirnov test the probability that both samples were drawn from the same distribution of axis ratios is 27% . the distributions are also consistent with results for the general population of massive galaxies at @xmath18 ( chang et al . 2013 ; van der wel et al . we note that we do not detect a significant wavelength dependence of the mean axis ratio of the 25 scmgs : we find @xmath290 in @xmath28 and @xmath291 in @xmath30 . = 8.5 cm even though the stars are not in thin disks , the gas can be . if the gas is in rotationally - supported disks that are aligned with the stellar distribution , the measured velocity dispersions are expected to show an anti - correlation with the observed axis ratios of the galaxies . as shown in fig . [ sigq.fig]a we see precisely this effect : there is an anti - correlation , with a correlation coefficient of @xmath292 and a significance of 94% . this is strong evidence that the gas is in disks and that the measured dispersions are dominated by gravitational motions . and @xmath293 has slightly less scatter , and equal significance . ] this anti - correlation is _ not _ consistent with m82-style galactic winds : outflows that are perpendicular to the disk lead to the highest observed velocities ( and hence integrated velocity dispersions ) when the disk is viewed face - on . = 8.5 cm going back to eq . [ siggen.eq ] , we now assume that @xmath269 and @xmath294 can be neglected , so that @xmath295 to derive rotation velocities we need to determine the relation between inclination and axis ratio in our sample . we constructed a model with long , intermediate , and short axes @xmath296 , @xmath297 , and @xmath117 that reproduces the observed axis ratio distribution for random viewing angles . the orange line in fig . [ incl.fig]a shows the predicted distribution of @xmath293 for thick disks or oblate spheroids with @xmath298 and @xmath299 uniformly distributed between @xmath300 and @xmath301 . this model is an excellent fit , @xmath297 , and @xmath117 ( see , e.g. , franx et al . although there is some evidence that the stellar distribution of compact @xmath18 galaxies is oblate or disk - like rather than triaxial ( e.g. , van der wel et al . 2014a ; zolotov et al . 2015 ) , in our paper this is an assumption , not a result . ] to the observed distribution of @xmath293 . it should be emphasized that this is a model for the intrinsic shapes of the _ stellar _ distribution , not for the gas distribution : the gas is likely in thinner disks , than the stellar distribution , they are probably not as thin as disks in the local universe ( see , e.g. , cresci et al . 2009 ) . ] and all we assume is that the gas disks of the galaxies are aligned with their stellar distributions . for galaxies with intrinsic thickness @xmath302 the relation between the inclination and the observed axis ratio is given by @xmath303 as @xmath302 is not a constant in our model the relation between @xmath266 and @xmath293 is not single - valued . the solid line in fig . [ incl.fig]b shows the median relation , and the broken lines indicate the @xmath155 scatter . figure [ incl.fig]c shows the inclination correction @xmath304 as a function of @xmath293 . the inclination - corrected rotation velocities are shown in fig.[sigq.fig]b . they are derived from the gas velocity dispersions and the observed axis ratios of the galaxies using the average relation in [ incl.fig]c and assuming @xmath305 ( see rix et al . 1997 ; weiner et al . 2006 ) . in appendix c we show that this value is a reasonable approximation for the geometries of both the mass and the ionized gas that we derive in this paper . the large uncertainty reflects the fact that the conversion of dispersion to rotation velocity depends on the spatial distribution of the gas , and the underlying velocity field ( see appendix c ) . data of much higher spatial resolution and s / n ratio are needed to measure @xmath8 directly for these extremely compact galaxies . and but two narrow peaks in a `` double - horned '' profile that happen to have exactly the separation of h@xmath8 and @xmath3066584 . this may happen when the h@xmath8 emission originates from a narrow ring rather than a disk . in most cases that interpretation can readily be ruled out , from the spatially - resolved line profile ( see sect.[morphline.sec ] ) or from the detection of the @xmath3066548 line , but in a few cases ( e.g. , aegis_41114 ) it is difficult to exclude this possibility without observing other emission lines . ] the uncertainty in @xmath8 and 50% of the ( logarithmic ) inclination correction were added in quadrature to the error budget . the median rotation velocity for the full sample is @xmath307 . excluding the x - ray agn we find @xmath308 . if it is assumed that @xmath87 is not only the half - light radius in the @xmath30 band but also the half - light radius of the h@xmath8 emission , we can define the dynamical mass as @xmath309 this is not a true total mass but simply twice the enclosed mass within the half - light radius . in fig . [ vdyn.fig ] this dynamical mass is compared to the stellar mass . although the inclination corrections have lessened the offsets of the most extreme outliers , it is clear that orientation effects are not sufficient to explain the relatively low velocities that have been measured for a large fraction of the sample . the mean offset for the whole sample is @xmath310 dex , and the scatter is @xmath311 dex . in the next section we show that variation in the spatial extent of the ionized gas with respect to that of the stars is a likely source of both the offset and scatter in fig . [ vdyn.fig ] . in the previous section we showed that many galaxies have galaxy - integrated velocity dispersions that are much smaller than expected from their stellar masses and sizes . as demonstrated in sect . [ rot.sec ] this is partly caused by the @xmath312 reduction of the velocity of rotating disks . however , even after correcting the observed dispersions to inclination - corrected rotation velocities the dynamical masses are typically lower than the stellar masses , particularly for galaxies that do not host an x - ray agn . so far we have assumed that the spatial extent of the gas is similar to that of the stars , that is , @xmath313 , where @xmath314 is the half - light radius of the measured h@xmath8 distribution . '' requires molecular line measurements with high spatial resolution . ] there is no a priori reason why this should be the case ; e.g. , in the models of zolotov et al . ( 2015 ) compact galaxies often have rings of gas and young stars around their dense centers , which originate from ongoing accretion from the halo . furthermore , as shown earlier @xmath261% of the star formation in s is obscured , and the extinction is expected to be particularly high toward the central regions ( e.g. , gilli et al . 2014 ; nelson et al . . the distribution of detected h@xmath8 emission may therefore be less centrally concentrated than the distribution of star formation . the radius of the gas disks can be inferred from @xmath315 if we assume that the observed velocity is the circular velocity of the stellar body at the radius of the gas . the gas radius then depends on @xmath315 , the stellar mass , and the structural parameters of the galaxy : @xmath316 with @xmath315 the inclination - corrected rotation velocity and @xmath317 a function that depends on the mass distribution of the galaxies : @xmath318 here @xmath319 is the best - fitting sersic profile to the light distribution . for @xmath320 ( @xmath321 ) , @xmath322 and eq . [ rgas.eq ] is equivalent to eq . [ mvrot.eq ] with @xmath323 . these expressions ignore the fact that the 2d radii are not identical to the 3d radii , assume that the stellar mass distribution can be approximated by the @xmath30 luminosity distribution , and assume that the contributions of gas and dark matter to the total mass can be neglected on the scales that are probed by the h@xmath8 emission . solving eq . [ rgas.eq ] numerically , we find that the inferred gas disk sizes range from @xmath324kpc to @xmath284kpc . , as the gas could in principle also be located in the inner @xmath325pc where the rotation curve is still rising ( see , e.g. , fig . this is unlikely given that the galaxies have , by selection , star forming seds with a spatial extent of @xmath99kpc . furthermore , as we show later , the large radius solutions are corroborated by the measured spatial extent of the h@xmath8 emission . ] this large range is not surprising , as it is effectively an interpretation of the large variation that is seen in fig . [ vdyn.fig ] . figure [ sizepred.fig ] shows the relation between inferred @xmath314 and the stellar effective radius . the gas radii are typically larger than the stellar radii , particularly for the galaxies that do not have an x - ray agn ( black points ) . the ratio between the gas radius and the stellar radius is shown explicitly in the bottom panel of fig . [ sizepred.fig ] . the mean ratio , calculated with the biweight statistic ( beers et al . 1990 ) , is @xmath326 for the full sample . excluding galaxies with an agn , we find @xmath327 . that is , the gas disks are a factor of @xmath328 more extended than the stellar distribution . this is strictly a lower limit , as it is assumed that only stars contribute to the stellar mass , the galaxies have a relatively `` light '' chabrier ( 2003 ) imf , and there are no contributions from non - gravitational motions to the measured velocity dispersions . = 8.5 cm we can test directly whether the s are embedded in large gas disks by examining the observed spatial extent of the emission lines . even though the galaxies were selected to be extremely compact , the inferred spatial extent of the emission line gas is so large that it should ( just ) be detectable in ground - based , seeing - limited observations . the 2d spectra are shown in fig . [ 2dspec.fig ] ; they cover a rest - frame wavelength range from 6551 to 6596 and a spatial extent along the slit of @xmath329 . the five empty panels are the s from barro et al . ( 2014b ) . remarkably , about 1/3 of the galaxies show velocity gradients . they are most prominent in uds_33334 , uds_26012 , and uds_16442 , but also visible in goods - s_5981 , uds_42571 , and uds_35673 . the seeing ranged from @xmath113 to @xmath330 , and the stellar half - light radii of the galaxies are typically @xmath331 . therefore , the fact that we spatially resolve the h@xmath8 emission immediately demonstrates that the ionized gas extends to larger radii than the stars in these galaxies . we emphasize here that we do _ not _ attempt to measure rotation curves directly from these velocity gradients , as this can only be done reliably when the sizes of galaxies are similar to , or larger than , the spatial resolution of the data ( see , e.g. , vogt et al . 1996 ; miller et al . 2011 ; newman et al . 2013 ) . for the nine galaxies that were observed with mosfire we can measure the spatial extent of the h@xmath8 emission . as discussed in sect.[mosfire.sec ] a bright star was included in all mosfire masks , and the profile of this star in the spatial direction can be used to approximate the psf . we extracted spatial profiles of the combined and emission for the nine galaxies by averaging the data in the wavelength direction . each column was weighted by the inverse of the noise ( which is dominated by sky emission lines ) ; we did not weight by the signal as this would bias the profile towards the central regions . the spatial profiles are shown in fig . [ 2dspec.fig ] ( black points with errorbars ) . each panel also shows the profile of the star that was observed in that particular mask ( orange points ) ; the fwhm of this profile is also indicated . the profiles were fit by a model to determine the half - light radii of the ionized gas . the model has the form @xmath332 with @xmath333 the position along the slit , @xmath334 the model for the one - dimensional surface brightness profile of along the slit , @xmath335 a gaussian fit to the profile of the star , and @xmath118 denoting a convolution . the gaussian fits to the stellar profiles are shown by orange lines in fig . [ 2dspec.fig ] . parameterizing @xmath335 with the sum of two gaussians does not improve the fit to the stellar profile or change the results . it is not possible to constrain the functional form of the surface brightness profile with our data . instead , we assume that the h@xmath8 is in an exponential disk ( see nelson et al . 2013 ) : @xmath336 here @xmath337 is a normalization factor , @xmath338 is the center of the profile , and @xmath314 is the half - light radius of the ionized gas . we fitted this model to the data using the emcee code , as described for the fits in the wavelength direction in sect . [ fitting.sec ] . again , the priors are top hats with bounds that do not constrain the fits or the errorbars . rather than @xmath314 itself we fit @xmath339 : the error distribution of @xmath314 is highly asymmetric , which means that the peak of the distribution of samples does not coincide with its 50@xmath134 percentile . the distribution of the @xmath339 samples is symmetric . the resulting measured gas radii , converted to kpc , are listed in the panels of fig.[2dspec.fig ] . for seven out of nine galaxies the value of @xmath314 is different from zero with @xmath340 significance . a geometric correction needs to be applied to the measured values of @xmath314 to account for the fact that the slit is typically not aligned with the major axis of the gas disk . this correction depends on the orientation of the slit and on the inclination of the gas disk : @xmath341^{-0.5 } r_{\rm gas } % } \right ) % r^{\rm c}_{\rm gas } \approx \left ( \cos\,(i ) + \frac{1-\cos\,(i ) } % { \|\cos\,({\rm pa}_{\rm slit } - { \rm pa}_{\rm gal})\| } \right ) % r_{\rm gas},\ ] ] with @xmath266 the inclination ( as derived in sect . [ rot.sec ] ) , @xmath342 the position angle of the slitmask , and @xmath343 the orientation of the galaxy on the sky ( as determined with galfit ) . note that the corrected @xmath344 is measured along the major axis ( and is not a circularized radius ) , consistent with our interpretation that the gas is in thin , rotating disks . the median correction is small at 9% . for goods - s_30274 we use the median correction of the other eight galaxies , as its pa mostly reflects the orientation of its tidal tail . we use the corrected radii when comparing the measured radii to predicted radii and when deriving the rotation curve of the galaxies in sect . [ rotcurve.sec ] . for three galaxies , uds_35673 , goods - s_30274 , and goods - n_6215 , we obtained an independent measurement of the extent of the emission line gas from their wfc3/g141 grism spectra . these are the only galaxies in the sample of 25 that have grism spectra covering the redshifted @xmath345 lines and a detection of these lines with @xmath346 significance . as shown in nelson et al . ( 2012 ) emission lines in grism spectra are images of the galaxy in the light of that line , providing direct information on the distribution of ionized gas at @xmath347 resolution . the interpretation of the lines is complicated by the fact that the two lines are very close together on the detector . we fit the lines simultaneously with galfit ( peng et al . 2002 ) , keeping their relative location and flux ratio fixed and using a psf generated with tiny tim ( krist 1995 ) . two of the three galaxies ( uds_35673 and goods - s_30274 ) are also in the mosfire sample . the best - fit g141 radii of these objects are @xmath348kpc and @xmath349kpc , in excellent agreement with the mosfire values ( @xmath350kpc and @xmath351kpc , respectively ) . the third galaxy , goods - n_6215 , has a g141 radius of @xmath352kpc . in the following , we show all twelve measurements in figures ( nine from mosfire , three from hst ) , with lines connecting the two independent measurements for uds_35673 and goods - s_30274 . for the ten galaxies with gas size measurements we can directly compare the inferred sizes to the measured ones . the results are shown in fig . [ comparesize.fig ] . there is a clear correlation , with a significance of @xmath250% . furthermore , the offset between the two sets of radii is small . giving equal weight to all twelve measurements we find a difference of only @xmath353 dex . this excellent agreement between inferred and measured radii provides support to our modeling of the observed kinematics of s. = 8.5 cm = 15.5 cm this result is presented in a different way in fig.[massmass.fig ] , which shows the relation between dynamical mass and stellar mass . the left panel is identical to fig . [ vdyn.fig ] , but here we only show the ten galaxies with measured h@xmath8 effective radii . the dynamical masses in the right panel were calculated using @xmath354 with @xmath317 accounting for the ( small ) fraction of the mass that is outside @xmath314 ( see sect . [ sigcomp.sec ] ) . the dynamical masses in the right panel are consistent with the stellar masses for all galaxies , although we note that the sample is small . the mean offset is @xmath355 , and the rms scatter is 0.25dex . summarizing the results from this and the previous section , we have inferred that s have rotating gas disks whose observed spatial extent is larger by a factor of @xmath9 than their stellar distribution . this is based on four related results : 1 ) many of the galaxies have very low galaxy - integrated velocity dispersions ; this shows that the gas does not have the same spatial distribution as the stars and that galactic - scale winds do not dominate the kinematics for the majority of the sample ( fig . [ sigsig.fig]a ) . 2 ) the observed dispersions display a significant anti - correlation with the axis ratios of the galaxies ; this is consistent with disks under a range of viewing angles and difficult to reconcile with m82-style galactic winds ( fig . [ sigq.fig]a ) . 3 ) nearly all galaxies with spatially - resolved gas distributions show velocity gradients ( fig . [ 2dspec.fig ] ) . 4 ) inferring the sizes of the gas disks from the inclination - corrected rotation velocities , we find good agreement between the inferred sizes and the measured sizes ( fig . [ comparesize.fig ] ) . the measured kinematics can be used to constrain the total mass within @xmath356kpc . we can derive a spatially - resolved rotation curve by making use of the fact that the measured spatial extent of the gas varies by a factor of 10 ( see fig . [ comparesize.fig ] ) , under the assumption that the galaxies have similar inclination - corrected dynamics after scaling them to a common mass . the validity of this approach is demonstrated in appendix c , where we calculate the relation between the observed galaxy - integrated linewidths and the actual rotation velocity at @xmath357 . to bring all galaxies to the same normalization , we first define the scaled rotation velocity as @xmath358 with @xmath359the median stellar mass of the full sample of 25 s. we note that this scaling does not change the velocities by a large amount as the galaxies in our sample span a small mass range . = 8.5 cm in fig . [ kepler.fig ] the scaled velocities are plotted as a function of the measured gas half - light radius @xmath314 ( corrected for slit orientation ) for the 10 galaxies that have this measurement . the rotation curve declines : in galaxies where h@xmath8 is measured at a larger distance from the center , the inclination - corrected rotation velocity is lower . the decline has a formal significance @xmath250% . falling rotation curves have been seen previously in some individual ( large , non - compact ) @xmath18 galaxies ( e.g. , the galaxies d3a6397 and zc400690 in genzel et al . the solid line is the predicted rotation curve for an @xmath360 galaxy with the median effective radius ( @xmath361kpc ) and median sersic index ( @xmath362 ) of the s , calculated with eq . [ rgas.eq ] . this model is a good description of the data : @xmath363 with 12 degrees of freedom . the grey line is a model with two mass components : in addition to the stellar component this model has a gas component with the same mass as the stars ( i.e. , the gas fraction in this model is @xmath364 ) . for consistency with the previous sections , the spatial distribution of the gas is assumed to be exponential with @xmath365 . the grey line overpredicts the observed velocities : with @xmath366 this model can be ruled out with @xmath367% confidence . we can derive an upper limit to the gas mass within 7kpc by assuming that the uncertainty in the stellar mass is small and allowing the mass in the gas component to vary . the 95% confidence upper limit to the gas mass is @xmath368 , corresponding to a limit on the gas fraction of @xmath369 . it appears that the gas is mostly a tracer of , rather than a contributor to , the kinematics . finally , we derive the best fitting mass within @xmath370kpc by assuming that @xmath371 and allowing @xmath50 to vary : @xmath372 , where the errorbars are 95% confidence limits . although this estimate assumes that mass follows light , we verified that the results are very similar for more extended mass distributions . we conclude that the dynamical mass within @xmath373kpc is fully consistent with the stellar mass that is implied by the stellar population fit ; and that there is little room for additional stars , gas , or dark matter inside this radius . in the previous sections we have shown that a population of star forming galaxies exists at @xmath4 whose dynamical mass within @xmath356kpc is dominated by a massive , compact , stellar component . we now ask whether these galaxies can be progenitors of the population of massive , compact , quiescent galaxies , by considering their number densities , morphologies , and star formation rates . this question has been discussed before , by , e.g. , williams et al . ( 2014 , 2015 ) , bruce et al . ( 2014 ) , nelson et al . ( 2014 ) , dekel & burkert ( 2014 ) , zolotov et al . arguably the most extensive observational study is a series of papers by barro et al . ( barro et al . 2013 , 2014a , 2014b ) , using data over two ( barro et al . 2013 , 2014b ) or one ( barro et al . 2014a ) of the five fields that we study here . using our larger data set and more restrictive selection we find broadly similar results . a star forming compact massive galaxy will resemble a quiescent compact massive galaxy if star formation stops ( quenching ) . however , the opposite is also true : a quiescent compact galaxy that starts forming stars due to the accretion of new gas ( see , e.g. , zolotov et al . 2015 ; graham , dullo , & savorgnan 2015 ) could resemble a star forming compact galaxy ( rejuvenation ) . we can determine whether quenching or rejuvenation dominates by measuring the number density of s and s as a function of redshift . the selection criteria of sect . [ candidates.sec ] were applied in small redshift bins , and the number density was determined by dividing the number of galaxies in the bin by its volume . the result is shown in fig . [ evo.fig ] ( filled points and solid curves ) . = 8.5 cm at @xmath6 the number densities of the two populations are very similar , as already noted in sect . [ predsig.sec ] . however , at higher and lower redshifts the number densities are different : the s have a roughly constant number density from @xmath374 to @xmath375 , whereas the number density of s increases by an order of magnitude over that same redshift range . : straatman et al . ( 2015 ) recently reported the existence of a sizeable population of compact , massive quiescent galaxies at @xmath376 , based on medium - band near - ir photometry . ] a straightforward interpretation is that star forming galaxies continuously enter the `` compact massive '' selection box ( because of a decrease in their size and/or an increase in their mass ) , and quench shortly after . this continuous quenching then leads to a rapid build - up of the number of quiescent galaxies in the compact / massive selection region . we conclude that quenching likely dominates over rejuvenation : if rejuvenation dominated , we would have expected to see quiescent galaxies disappear as their star formation ( re-)started , unless there are other channels to create quiescent compact galaxies . we note that the evolution of the number densities of the two populations is qualitatively similar to the simulations of zolotov et al . ( 2015 ) . it is difficult to determine how long it takes before a compact star forming galaxy turns into a quiescent galaxy , as this depends on the rate with which new galaxies enter the sample . the number density of s is constant from @xmath374 to @xmath375 , which means that new s enter the sample at approximately the same rate as existing ones quench . we can obtain a very rough estimate of the `` compact life time '' of star forming galaxies @xmath377 by adding the number densities of the s in the three redshift bins that cover this period : if the average quenching time is much shorter than the time interval between redshift bins , all galaxies in each bin are new arrivals and should be added to the sample of progenitors of quiescent galaxies . the combined number density in these bins ( which are of nearly equal volume ) is @xmath378mpc@xmath53 , higher than the increase in the number density of the s over this period ( @xmath379mpc@xmath53 ) . this implies that only about half of the star forming galaxies disappear from one redshift bin to the next , and that the average quenching timescale is roughly equal to the time interval between the redshift bins : @xmath380gyr . this is the average lifetime of star forming galaxies in the `` compact massive '' selection box , assuming that they all turn into quiescent galaxies . it is slightly lower than the value of @xmath381gyr derived by barro et al . ( 2013 ) , but judging from their fig . 5 the two studies are broadly consistent . although somewhat outside of the scope of this paper , we briefly discuss the number density evolution at lower redshift . the number density of s drops precipitously after @xmath375 . this drop leads to a plateau in the number density of s : as the number of star forming progenitors decreases , no new quiescent galaxies are added to the sample . at the lowest redshifts the number density of compact quiescent galaxies decreases ( as was also found by taylor et al . 2010b , van der wel et al . 2014b , and van dokkum et al . 2014 , among others ) , while the number density of _ all _ massive quiescent galaxies rises steeply ( dashed red curve ) . the likely explanation is that the compact galaxies accrete extended envelopes through merging from @xmath3 to the present day ( e.g. , bezanson et al . 2009 ; naab et al . 2009 ; van dokkum et al . 2010 , 2014 ; newman et al . 2012 ; hilz et al . 2013 ) . finally , we note that fig . [ evo.fig ] is not new : the peak in the number density of compact , massive quiescent galaxies was also shown in cassata et al . ( 2011 , 2013 ) , barro et al . 2013 , and van der wel et al . ( 2014b ) . barro et al . ( 2013 ) derive a similar lifetime for star forming galaxies in the compact selection region . although uncertainties remain ( particularly at low redshift ; see , e.g. , carollo et al . 2013 ) , it is encouraging that these largely independent samples give similar results . the large spatial extent of the ionized gas raises the question whether the stellar half - light radii and masses of the compact star forming galaxies have been underestimated : although it is difficult to bias galfit measurements in this direction ( see , e.g. , davari et al . 2014 ) , it is possible that the galaxies have extended low surface brightness envelopes ( see , e.g. , hopkins et al . if such envelopes exists this would call into question whether s can be direct progenitors of compact quiescent galaxies with the same apparent mass and half - light radius . images of the galaxies are shown in fig . [ hstim.fig ] and in fig . a1 ( see sect.[parent.sec ] ) . visually , most of the objects have a compact luminosity distribution and no spiral arms , clumps , star forming complexes , or other features outside of the dense center . several of the reddest galaxies do not appear very compact : for example , uds_42571 and , in particular , uds_16442 are faint and fuzzy rather than bright and compact . the reason for their relatively low surface brightness is that dust obscuration has dramatically lowered their luminosity : as galaxies can have high @xmath382 ratios , compact in mass does not necessarily imply compact in light . two objects show unambiguous evidence for ongoing or past mergers : goods - s_30274 has an asymmetric feature resembling a tidal tail , and cosmos_11363 is one component of a spectacular merger between two compact galaxies with a projected separation of @xmath113 ( 5kpc ) . the companion of cosmos_11363 is cosmos_11337 in the skelton et al . ( 2014 ) catalog . our keck / nirspec and hst / wfc3 spectroscopy confirms that they are at the same redshift . with @xmath383kpc and @xmath384 cosmos_11337 is actually significantly more compact than cosmos_11363 . its rest - frame @xmath47 colors ( just ) give it a quiescent classification . magnitudes of these objects are somewhat uncertain as they rely on accurate deblending of the irac fluxes ; it may well be that both galaxies are s. ] this merging pair seems to suggest that cmgs can form in mergers ( hopkins et al . 2009b ) , but that is not the right interpretation : as both galaxies already fall in the `` compact massive '' selection region , this particular type of merger actually _ decreases _ their number . even if the result of the merger falls in the selection region , there will be one less cmg . interestingly , several other galaxies show evidence for distorted outer isophotes in fig . a1 . this could indicate interactions are common for these galaxies , but the evidence is not conclusive at the depth of the candels imaging . = 8.5 cm to quantify the stellar emission on scales @xmath385kpc we stacked the @xmath30 images of the 25 s and measured their averaged radial surface brightness profile to faint levels . each galaxy was normalized by its total @xmath30 flux prior to stacking , so that the stack is not dominated by a few bright objects . neighboring objects , identified from the sextractor segmentation map ( see bertin & arnouts 1996 ; skelton et al . 2014 ) , were masked . the resulting surface brightness profile is shown in the top panel of fig . [ sbprof.fig ] ( blue points ) . we fit the stack with a psf - convolved sersic profile to determine whether there is evidence for an additional component at large radii . this fit , done with galfit ( peng et al . 2002 ) , is shown by the blue line . it is an excellent description of the data out to 15kpc ( @xmath386 ) : there is no excess light beyond a single sersic profile . furthermore , the best - fitting effective radius ( @xmath361kpc ) and sersic index ( @xmath387 ) are similar to the median values of the 25 galaxies that went into the stack : @xmath388kpc and @xmath389 . the stacked profile is compared to a stacked profile , shown in red in fig . [ sbprof.fig ] . the s in this figure are a subset of the full population : they were selected in narrow bins of mass and effective radius , centered on the median values of the 25 s. this ensures that any differences between the stacks are not caused by a difference in the mean size or mass of the samples . the quiescent profile is virtually indistinguishable from that of the star forming galaxies . finally , @xmath390 color profiles of both stacks are shown in the bottom panel of fig . [ sbprof.fig ] . both stacks are bluer at larger radii and the gradients are small , qualitatively consistent with previous work ( szomoru , franx , & van dokkum 2012 ) . the negative color gradients imply that the galaxies are even more compact in mass than in light , and that any stellar emission at @xmath391 is not missed because it is enshrouded in dust . we conclude that the morphologies of the s are consistent with being direct progenitors of s. when selected to have the same mass and effective radius , their surface brightness profiles are indistinguishable out to at least 15kpc . we find a relatively high sersic index for both populations . such high values ( and the relatively round 3d morphologies ; see sect . [ rot.sec ] ) are consistent with violent relaxation following a merger , but also with composite structures , such as envelopes of material around extremely compact exponential disks . accepting that the s are direct progenitors of s , an important question is whether they are forming a large fraction of the stars that are present in their quiescent descendants . if the life times of the s are short , or the star formation rates are low , they may account for only a small fraction of the total stellar mass in compact massive galaxies at @xmath18 . we address this question in fig . [ ssfr.fig ] , which shows the relation between the specific star formation rate and compactness within the sample of compact , massive galaxies at @xmath74 . = 8.5 cm the right axis of this figures shows the fraction of the total stellar mass that is formed in the compact phase : @xmath392 with ssfr the specific star formation rate , @xmath393 a correction for mass loss due to stellar winds , and @xmath394 the average life time of star forming galaxies in the compact , massive selection region . the median specific star formation rate of the s is ssfr@xmath395yr@xmath27 , and for @xmath396 ( chabrier 2003 ) and @xmath397gyr ( sect . [ evoorder.sec ] ) we find @xmath398 . as they are , on average , observed halfway through their lifetime in the compact selection region , their final mass before quenching will be @xmath399 , and the fraction of @xmath400 that is formed in the compact phase is then @xmath401 . we conclude that s are responsible for forming a significant fraction of the stars that are present in compact quiescent galaxies . an implication of this result is that the spatial distribution of the emission in s is probably more extended than the spatial distribution of star formation in these galaxies . this is qualitatively similar to results for galaxies at @xmath402 ( nelson et al . 2012 , 2015 ) , and may indicate that star formation has ceased in the inner regions of the galaxies ( e.g. , genzel et al . 2014b ; tacchella et al . however , as discussed in sect . [ sfr.sec ] most of the star formation in s is obscured , and the observed emission accounts for only @xmath7% of the total star formation . as the column density is a very strong function of radius in these compact galaxies ( see , e.g. , gilli et al . 2014 ; nelson et al . 2014 ) , the obscuration - corrected distribution of star formation is almost certainly much more compact than the observed distribution of emission at least for the galaxies with low observed velocity dispersions . a somewhat puzzling aspect of the s is that they have very high specific star formation rates even though their observed kinematics leave little room for a large gas reservoir ( see sect . [ rotcurve.sec ] ) . many studies have found that the molecular gas and dust content of galaxies increases with redshift , and reaches @xmath403% of the total baryonic mass for @xmath18 galaxies with the highest star formation rates ( e.g. , tacconi et al . 2010 ; daddi et al . 2010 ; genzel et al . 2015 ; scoville et al . 2015 ) . using the scaling relations derived in genzel et al . ( 2015 ) , the expected gas fraction for the galaxies in our sample is @xmath404% . one possible explanation for their relatively low gas fraction is that the galaxies have nearly exhausted their reservoir and are about to quench . if the galaxies typically build @xmath405% of their mass inside the compact , massive selection region , the average should have @xmath406% of their mass in gas ( for @xmath396 ) ; this is just consistent with the 95% confidence upper limit on the gas fraction of 40% that we derived in sect . [ rotcurve.sec ] . another explanation is that newly accreted gas is continuously and efficiently funneled into the central regions , and the star formation rates are `` accretion throttled '' ( dekel et al . 2009 ) ; in that case the gas depletion time can be shorter than the actual duration of star formation ( see , e.g. , genzel et al . direct observations of the dust and molecular gas in s , at @xmath99kpc resolution , are needed to address these questions . finally , we note that star forming galaxies tend to be less compact than quiescent galaxies even _ within _ the population of compact massive galaxies at @xmath74 ( see fig . [ ssfr.fig ] ) . as discussed earlier in the context of the sample selection ( sect . [ predsig.sec ] ) , star forming galaxies are always less compact than quiescent galaxies , irrespective of the precise criteria for their selection . in the next section we interpret the distribution of galaxies in the size - mass plane in the context of a simple model , in which star forming galaxies become gradually more compact and the probability of quenching rises smoothly as their compactness increases . in this section we turn to the _ formation _ of compact , massive star forming galaxies . several distinct mechanisms have been discussed in the literature , including mergers of gas - rich galaxies ( tacconi et al . 2008 ; hopkins et al . 2009b ; hammer et al . 2009 ; wellons et al . 2015 ) , in - situ , inside - out growth of even more compact progenitors ( oser et al . 2010 ; johansson , naab , & ostriker 2012 ; williams et al . 2014 ; nelson et al . 2014 ; wellons et al . 2015 ) , `` compaction '' of the gas in star forming galaxies due to disk instabilities ( dekel & burkert 2014 ) , and hybrid models that include several of these effects ( zolotov et al . 2015 ) . although individual massive galaxies likely have complex formation histories , including periods of compaction , mergers , and star bursts , the _ population _ of massive galaxies should follow a particular track in the size - mass plane that is determined by the dominant mode of growth when the evolution of many galaxies is averaged . tracks derived from observations and simulations are shown in fig . [ meangrowth.fig ] . the blue and red tracks show the evolution of galaxies matched by their cumulative number density , for ( relatively ) low mass galaxies ( van dokkum et al . 2013 , blue ) and high mass galaxies ( patel et al . 2013 , red ) . the solid parts of the curves are for @xmath407 and the dotted parts for @xmath408 . low mass galaxies evolve along a single track with a slope of @xmath409 . high mass galaxies evolve along a similar track from @xmath2 to @xmath3 but then turn `` upward '' , around the time when star formation ceases and the growth becomes dominated by dry mergers ( see sect.[lowz.sec ] ) . = 8.5 cm magenta , orange , and black curves are from simulations . the magenta tracks are the wind models shown in fig . 10 of hirschmann et al . ( 2013 ) , for two different mass ranges . these models are the same as those in genel et al . ( 2012 ) , and are updated versions of the momentum - driven wind models of oppenheimer & dav ( 2006 ) in cosmological simulations . they include both winds and metal enrichment ; as shown in hirschmann et al . ( 2013 ) models without winds predict somewhat steeper relations between size growth and mass growth . the orange curve is the track of galaxies in the illustris project ( vogelsberger et al . 2014 ) , as shown in fig . 5 of wellons et al . this is the average track of all galaxies with a stellar mass in the range @xmath410 at @xmath15 . the thin black curves show the evolution from @xmath411 to @xmath412 of individual galaxies in the simulations of zolotov et al . we include all 34 simulations , irrespective of whether they have a `` compaction '' phase . the thick dashed curve was created by averaging the evolution in these simulations . the number density - matched observational samples and the simulations all suggest that the ensemble - averaged evolution of star forming galaxies in the size - mass plane is well approximated by @xmath413 that is , galaxies increase their size by a factor of 2 for every factor of 10 evolution in their mass . this simple inside - out growth model is qualitatively consistent with a host of other data and theory , including the expected growth of disks in @xmath414cdm ( e.g. , mo , mao , & white 1998 ) and the distributions of star formation and existing stars in galaxies ( e.g. , nelson et al . interestingly , this track corresponds to an approximately constant 3d density within the effective radius ( as @xmath415 , it follows that @xmath416 if the density is constant ) . although the 3d density within the effective radius stays constant , a direct consequence of eq . [ growth.eq ] is that the stellar density within a physical radius , the stellar surface density , and the stellar velocity dispersion all gradually increase as galaxies form stars . we assume that galaxies have an increasing likelihood of quenching as their velocity dispersion reaches a particular threshold . this is motivated by numerous studies showing that the specific star formation rates of galaxies correlate much better with compactness than with mass ( e.g. , kauffmann et al . 2003 ; franx et al . we parameterize this process by a dispersion - dependent quenching probability @xmath417 : @xmath418 with @xmath419 ( see fig . [ q.fig ] ) . galaxies begin to quench at @xmath420 , or @xmath421 ( eq . [ sigpred.eq ] ) . as we show below this particular choice of @xmath422 provides a reasonably good fit to the data over the redshift range @xmath423 . we use a single value in this paper , but we note that the threshold is a function of redshift : low redshift galaxies quench at a lower density or dispersion than high redshift galaxies ( franx et al . 2008 ) . = 8.5 cm the average mass growth of the population is assumed to be a simple function of the star formation rate , modified by the quenching function : @xmath424 the parameter @xmath425 encompasses mass loss due to stellar winds , possible effects of mergers , and the well - documented offset between the evolution of the star forming sequence and the evolution of the stellar mass function ( see leja et al . 2015 ; papovich et al . 2015 , and references therein ) . we adopt @xmath426 ; values of @xmath427 produce very similar results . a pure mass loss model would have @xmath428 for a chabrier ( 2003 ) imf . the star formation rate is given by the star forming `` main sequence '' . we adopt the mass - dependent parameterization of whitaker et al . ( 2014 ) : @xmath429 with @xmath430 , @xmath431 , and @xmath432 for the redshift range of interest . as shown in fig . [ select.fig]c the actual star formation rates of s are broadly consistent with this relation . the model is illustrated in fig . [ moddemo.fig ] , which shows galaxies in the size - mass plane at @xmath433 . the color indicates the fraction of galaxies that are quiescent according to the @xmath47 criteria . galaxies move along the green curves until they cross the yellow line , when their quenching probability rises steeply . in this model galaxies follow parallel tracks in the size - mass plane , which means that large galaxies and small galaxies at fixed mass have different formation histories . however , we emphasize that individual galaxies likely have complex histories , involving excursions above and below these mean tracks ( see , e.g. , zolotov et al . our description is qualitatively similar to the work of williams et al . ( 2014 , 2015 ) , who identified low mass lyman break galaxies with small sizes as possible progenitors of quiescent compact massive galaxies . = 8.5 cm = 15.5 cm = 15.5 cm we test the model in the following way . we first quantify the distribution of galaxies in the size - mass plane at @xmath434 , by measuring the number of galaxies in bins of [email protected] ( see fig . [ modtest1.fig]a ) . next , we evolve this distribution forward in time , using timesteps of @xmath436myr . for each combination of @xmath437 we can calculate the sfr from eq . [ ms.eq ] , @xmath417 from eq . [ quench.eq ] , the change in mass from eq . [ massgrowth.eq ] , and the corresponding change in size from eq . [ growth.eq ] . the evolved distribution after 10 timesteps ( i.e. , 1 gyr ) is shown in fig . [ modtest1.fig]b , with a small ( 4% ) correction to account for the volume difference between @xmath438 and @xmath439 . as expected , the galaxies have shifted to larger masses and to slightly larger radii in the size - mass plane . the distribution artificially falls off at low masses due to the @xmath440 limit in fig.[modtest1.fig]a . this limit was chosen to ensure that the galaxies with the lowest masses and highest redshifts have robust size measurements : the median brightness of the 28 galaxies with @xmath441 and @xmath442 is @xmath443 , well within the regime where size measurements are reliable ( see van der wel et al . 2014b ) . the _ observed _ distribution of galaxies at @xmath439 is shown in fig . [ modtest1.fig]c . in panel ( d ) this observed distribution is multiplied by a weight mask , to account for the artificial fall - off at low masses in panel ( b ) . the weight mask was constructed by evolving a galaxy population with a uniform density distribution in the size - mass plane and a cutoff at @xmath444 forward in time ( in the same way as described above ) . the distribution in fig.[modtest1.fig]d is remarkably similar to that in fig.[modtest1.fig]b . furthermore , the total number density of galaxies in the two panels is almost identical ; panel ( d ) has 7% less galaxies than panel ( b ) . in fig . [ modtest2.fig ] the color - coding reflects the specific star formation rates of the galaxies , with redder squares indicating a lower ssfr . the figure looks very similar when the fraction of quiescent galaxies is used for the color coding instead of the ssfr . the sizes of the squares are proportional to the number of galaxies . the model naturally produces a population of quiescent galaxies with @xmath445and @xmath1kpc . in our model , the progenitors of these galaxies have masses of @xmath446and sizes of @xmath447kpc at @xmath2 . the model does not produce the right fraction of quiescent galaxies at the highest masses and largest sizes : many of these galaxies are forming stars at @xmath448 even though they have high galaxy - averaged velocity dispersions . this suggests that our quenching prescription is too simplistic in this regime ( see sect . [ modelsum.sec ] ) . we compare the predicted to the observed number densities explicitly in fig . [ compmod.fig ] . this figure highlights the excellent match of our model to the size distribution of all galaxies over the entire mass range @xmath449 : it not only reproduces the peak in the distribution at @xmath450kpc but also the `` shoulder '' of compact quiescent galaxies . it also demonstrates that the modeling of quenching is too simplistic for large galaxies , as was already clear from the comparison of panels ( b ) and ( d ) of fig.[modtest2.fig ] . in particular , nearly 100% of galaxies with @xmath451kpc are forming stars in the model , whereas the observed star forming fraction is only @xmath452% . = 8.5 cm in summary , we have shown that the population of compact , massive galaxies at @xmath18 can be explained by a model in which galaxies form stars at a rate that is dictated by the star forming sequence , experience a modest increase in size for a given increase in mass , and quench after passing a velocity dispersion threshold . this was demonstrated by evolving the observed galaxy population at @xmath453 forward by 1gyr to @xmath448 . this is a critical period as the number density of s increases by an order of magnitude over that redshift range . although it is beyond the scope of this ( already somewhat unwieldy ) paper , we note that the modeling can easily be extended . in particular , it would be straightforward to fit for the two tunable parameters ( the quenching dispersion @xmath454 and the parameter @xmath425 , which relates the mass growth to the star formation rate ) . furthermore , our quenching description is inadequate in the high mass / large size regime ; the yellow line in fig . [ moddemo.fig ] is somewhat too steep . a possible explanation is that quenching depends on the galaxy properties in the central @xmath99kpc , and the simple @xmath455 criterion no longer `` works '' in a regime where @xmath456kpc . some evidence for this comes from a study of the mass in the central @xmath457kpc of galaxies ( van dokkum et al . 2014 ) : as we showed in fig . 9 of that paper the mass inside of 1kpc is an excellent predictor of quiescence at all redshifts . finally , the modeling can be extended to lower redshifts , taking evolution in @xmath454 into account ( see sect . [ lowz.sec ] ) . in the preceding sections we discussed a simple model for the evolution of massive galaxies at @xmath458 : they grow inside - out with @xmath459 ( eq . [ growth.eq ] ) while they are forming stars , and quench when they reach a density or velocity dispersion threshold . this model provides an explanation for the fact that large galaxies have younger stellar populations than small galaxies at fixed mass ( e.g. , franx et al . 2008 ) , as only the smallest galaxies have reached the quenching threshold . galaxies enter the massive , compact selection region in the size - mass plane `` from the left '' , that is , by increasing their masses . this seems different from models in which large , massive galaxies enter this region `` from above '' , that is , by decreasing their sizes through mergers ( e.g. , hopkins et al . 2009b ) or by gas `` compaction '' followed by star formation ( dekel & burkert 2014 ) . this apparent difference may reflect a difference in approach : in this paper we are concerned with the average evolution of the population of massive galaxies , whereas simulations such as those of zolotov et al . ( 2015 ) are able to follow the tracks of individual galaxies in the size - mass plane . judging from the zolotov et al . ( 2015 ) tracks , eq . [ growth.eq ] may simply be the time- and population average of periods of proportional size and mass growth ( @xmath460 ) , periods of compaction , and the effects of mergers . at lower redshifts massive galaxies evolve along a markedly different track in the size - mass plane : van dokkum et al . ( 2010 ) , patel et al . ( 2013 ) , and others find that the size and mass evolution of massive galaxies are related through @xmath461 at @xmath462 ( as indicated by the dotted section of the red curve in fig . [ meangrowth.fig ] ) . this evolution can be explained by minor , gas - poor mergers building up the outer envelopes of galaxies ( bezanson et al . 2009 ; naab et al . 2009 ; hopkins et al . 2010 ; hilz et al . 2013 ) . in van dokkum ( 2010 ) we showed that _ any _ physical process that deposits mass at @xmath463 leads to a steep track in the size - mass plane , due to the definition of the effective radius . a schematic of the growth of massive galaxies from @xmath2 to @xmath464 is shown in fig . [ cartoon.fig ] . after galaxies quench , their mass growth per unit time is reduced , but their effective radii continue to increase . this figure suggests that there are multiple paths leading to large , massive , quiescent galaxies in the local universe , as was also noted in , e.g. , cappellari et al . ( 2013 ) and barro et al . ( 2014a ) . their @xmath18 progenitors can be large star forming ( disk ) galaxies , such as those studied extensively by , e.g. , genzel et al . ( 2008 ) and frster schreiber et al . ( 2011 ) , or compact , massive , quiescent galaxies that have grown through mergers ( e.g. , trujillo et al . 2011 ; patel et al . 2013 ; ownsworth et al . 2014 ) . as shown in fig . 2 of van dokkum et al . ( 2014 ) massive @xmath16 galaxies have a large range of central densities at fixed total mass , as expected in such scenarios . it is possible that massive s0 galaxies formed from large star forming galaxies and massive elliptical galaxies formed from compact star forming galaxies , although it remains to be seen whether the stellar populations of massive early - type galaxies are sufficiently diverse to accommodate a large range in formation histories ( gallazzi et al . 2005 ; van dokkum & van der marel 2007 ) . in this paper we mostly ignored the effects of agn , despite the fact that nearly half of the 25 galaxies with keck spectra have x - ray luminosities above the canonical agn limit of @xmath465ergss@xmath27 . the reason is that these effects are difficult to constrain and quantify . barro et al . ( 2013 ) discuss the high occurrence rate of agn in compact star forming galaxies extensively , and argue that they are the agent of quenching . this may be true : in many galaxy formation models agns play a crucial role in quenching star formation precisely in this mass and redshift range ( e.g. , croton et al . 2006 ; hopkins et al . 2008 ) . however , the star formation rates of the s are ( still ) high and consistent with the @xmath66 star forming sequence ( whitaker et al . 2014 ) , and there is no evidence for a direct effect of the agns on star formation . turning this around , it is obviously the case that the black holes are growing in these galaxies , and that they are growing at a time when the dense stellar centers are also growing . this is not surprising , as it is difficult to see how to _ avoid _ a high accretion rate onto the central object in these extremely dense , highly star forming galaxies . an obvious point of concern is that the presence of agns causes errors in the derived physical parameters of the galaxies . in principle , an agn in a relatively low mass , relatively large , and relatively quiescent galaxy could push the galaxy in the category : the extra light of the agn could be mistaken for star light , increasing the mass ; the combination of a point source with a normal galaxy could be mistaken for a compact bulge - dominated object ; and the hot ir flux from the agn could be mistaken for pah features from star formation . this can only be addressed properly with data of much higher spatial resolution than is available today , but we note here that the galaxies with agns do not stand out in any of the figures . the only exception is that the four galaxies with the highest measured velocity dispersions all have x - ray agn , and also ratios of @xmath99 . we have treated these four galaxies in the same way as the others . a related issue is the almost - certain presence of galactic - scale winds and outflows . such winds can be driven by star formation ( e.g. , heckman , armus , & miley 1987 ) and/or agns ( e.g. , proga , stone , & kallman 2000 ) and are ubiquitous in star forming galaxies at high redshifts ( franx et al . 1997 ; pettini et al . 1998 ; frster schreiber et al . 2014 ; genzel et al . galactic superwinds can create bubbles and shock fronts whose kinematics , spatial extent , and emission line ratios are very similar to what we observe . in at least one of the galaxies in our sample , cosmos_1014 , there is evidence for a broad h@xmath8 line in addition to a narrow component , similar to iras110950238 ( soto & martin 2012 ) and galaxies in frster schreiber et al . ( 2014 ) . furthermore , four of the galaxies in our sample are part of the sample of massive galaxies of genzel et al . ( 2014a ) ( cosmos_11363 , goods - s_30274 , goods - s_37745 , and goods - s_45068 ) , and they find broad nuclear velocity components in two of them ( cosmos_11363 and goods - s_30274 ) . a detailed study of the kinematics and line ratios of goods - s_30274 was also done by van dokkum et al . ( 2005 ) . although winds are almost certainly _ present _ , two results suggest that they are not dominating the galaxy - integrated emission line widths . first , winds tend to escape in a direction perpendicular to the plane of the galaxy ( heckman , armus , & miley 1990 ) , which is difficult to reconcile with the observed anti - correlation between velocity dispersion and axis ratio ( fig . [ sigq.fig ] ) . second , the observed kinematics are fully explained by the stellar mass , leaving little room for additional broadening due to winds . in fact , broad components in the velocity profiles are expected just from rotating gas at small radii : as shown in fig . [ kepler.fig ] gas at @xmath99kpc should have fwhm@xmath466 even in the absense of winds . judging from other @xmath18 galaxies the disks are also likely to be highly turbulent , with a relatively high internal dispersion ( see , e.g. , cresci et al . 2009 ; frster schreiber et al . the gaseous environments of s may be similar to those of ulirgs , which are highly complex : as shown in soto & martin ( 2012 ) they can have rotating , large - scale disks in addition to outflows and shocks . finally , we note that the presence of spatially - extended gas disks in these galaxies had been predicted by zolotov et al . they also predicted that the gas dispersions are , on average , lower than the stellar dispersions ( fig . [ sigsig.fig]a ) , as the gas is in disks which are sometimes seen face - on . interestingly , zolotov et al . ( 2015 ) also find that the gas constitutes only a small fraction of the total baryonic mass of the simulated compact massive star forming galaxies , although they note that this result is sensitive to the feedback prescription . similarly , johansson et al . ( 2012 ) predicted that compact , massive galaxies are stellar mass - dominated and have keplerian rotation curves ; the model rotation curves in their fig . 7 are remarkably similar to the inferred rotation curve shown in our fig . [ kepler.fig ] . this study begins with an hst / wfc3-selected sample in a total area of @xmath467 square degrees . many other studies have found extreme star forming galaxies by selecting them on the basis of their far - infrared , submm , or radio emission instead ( e.g. , kormendy & sanders 1992 ; sanders & mirabel 1996 ; barger et al . 1998 ; smail et al . 2000 ; barger et al . 2001 ; casey et al . these extreme galaxies are plausible ancestors of early - type galaxies ; as an example , tacconi et al . ( 2008 ) , toft et al . ( 2014 ) , and simpson et al . ( 2015 ) have suggested that many submm galaxies could be direct progenitors of compact quiescent galaxies at @xmath18 . we do not select _ against _ such objects , and our sample should include the proper number of submm galaxies , radio galaxies , and other extreme objects . however , there are ( at least ) two possible reasons why galaxies selected at other wavelenghts could be underrepresented in our sample : some fraction may be too faint in the near - ir to be included ( or to be properly characterized ) in the skelton et al . ( 2014 ) catalogs , and some may be too rare to be represented in the 3d - hst / candels area . s have such high column densities in the central regions that some may be entirely obscured at rest - frame optical wavelengths ( gilli et al . 2014 ; nelson et al . wang , barger , & cowie ( 2012 ) and caputi et al . ( 2014 ) show that objects exist that are relatively bright in the irac bands but that are undetected in deep near - ir data . it is obviously difficult to measure the redshifts and masses of these objects with traditional means , but it may be possible using molecular lines ( see walter et al . 2012 ; riechers et al . 2013 ) . in the context of the study presented here the question is not whether _ any _ massive , compact , `` optically - dark '' galaxies were missed , but what fraction of mass and star formation is in such objects . the second class of potentially missed objects are extremely rare , extremely luminous galaxies . the median star formation rates of s in our study is @xmath468yr@xmath27 , and we have 112 such objects at @xmath74 . therefore , objects that are so rare that there are only a few ( or zero ) in our survey volume must have star formation rates @xmath469yr@xmath27 to have a significant impact on our results . this seems extreme , but such objects probably exist : the most extreme herschel - selected galaxies at @xmath470 have estimated star formation rates up to @xmath471yr@xmath27 ( casey et al . furthermore , recently identified highly obscured quasars have bolometric luminosities of @xmath472ergss@xmath27 ( banerji et al . 2012 , 2015 ) , and it seems likely that the growth of the black holes in these objects is accompanied by prodigious star formation . it remains to be seen whether such objects are sufficiently common ( or rather , long - lived ) to impact results derived from candels - sized areas . finally , we note that we do not find a correlation between size and ir luminosity at fixed stellar mass , that is , an ir selection does not preferentially select compact galaxies but objects with a wide range of rest - frame optical sizes ( see also wiklind et al . 2014 ; simpson et al . 2015 ) . as an ir selection is effectively a star formation selection at high masses ( see , e.g. , whitaker et al . 2012 ; rodighiero et al . 2014 ) , this is perhaps not surprising . in this paper we have identified a population of star forming , compact , massive galaxies in the five fields of the candels and 3d - hst surveys . such objects have been studied previously by barro et al . ( 2013 , 2014b , 2014a ) and nelson et al . ( 2014 ) , and we build on their results . compared to the barro et al . studies , our selection is more restrictive , focusing only on the most massive and most compact galaxies ; we study an area that is @xmath473 times larger ; and our redshift catalogs make use of the 3d - hst grism spectra for all objects brighter than @xmath32 . we first confirm the redshifts and masses of the galaxies using keck mosfire and nirspec spectroscopy of 25 compact massive star forming galaxies at @xmath74 . the gas dynamics suggest that the galaxies are embedded in spatially - extended rotating disks ; this explains the low measured dispersions of a large fraction of the sample and the observed anti - correlation between the disperion and the axis ratio of the galaxies . support for this interpretation comes from direct measurements of the sizes of the h@xmath8 disks for 10 galaxies ; the fact that this is possible at all from ground - based , seeing - limited data already shows that the gas extends to scales @xmath4741kpc . the derived sizes of the gas disks , and the fall - off of the rotation curve that we construct for the galaxies , are in very good agreement with recent models for the formation of massive galaxies ( johansson et al . 2012 ; zolotov et al . 2015 ) . it is important to note that , in our interpretation , the measured gas velocity dispersions of the galaxies generally do not reflect the true @xmath315 in the stellar body . we predict that the ( inclination - corrected ) velocities at @xmath475kpc are @xmath476for all galaxies . this can be tested with adaptive optics - assisted observations of the h@xmath8 line . there is evidence for broad components in several of the velocity profiles ( see sect . [ complex.sec ] ) , and these complex profiles may reflect the combined effect of high rotation velocities at small radii and lower velocities at larger radii . a more direct measurement could come from co line widths , as these likely probe much smaller radii than the h@xmath8 emission ( see , e.g. , downes & solomon 1998 ) . next , we interpret the existence of star forming , compact galaxies at @xmath74 in the context of a simple model for the evolution of galaxies in the size - mass plane . we describe the average evolution of star - forming galaxies by the simple relation @xmath477 , with the mass evolution proportional to the main sequence star formation rate . we show that this evolution is a consistent feature in galaxy formation models of hirschmann et al . ( 2013 ) , wellons et al . ( 2015 ) , and zolotov et al . ( 2015 ) , and is also seen in observations of number density - matched samples of galaxies ( van dokkum et al . 2013 ; patel et al . 2013 ; ownsworth et al . 2014 ) . as galaxies move along this track their average 3d density within @xmath63 remains approximately constant ( as @xmath415 , it follows that @xmath416 if the density is constant ) . however , their density within a fixed physical radius increases , as does their projected ( 2d ) density and their velocity dispersion . following many other studies ( e.g. , franx et al . 2008 ; bell et al . 2012 ) , we assume that quenching occurs when galaxies reach a threshold in either velocity dispersion or physical density . we show that this model explains the evolution of the distribution of galaxies in the size - mass plane from @xmath453 to @xmath448 , the redshift range when the number density of massive compact quiescent galaxies increases by nearly an order of magnitude . in the context of this straightforward model , the progenitors of compact massive star forming galaxies at @xmath478 were simply somewhat less massive and slightly smaller galaxies at @xmath479 . our study has several important systematic uncertainties . first , the stellar masses of the galaxies are derived from fitting stellar population synthesis models to the photometry , and these models have not been tested for the extreme galaxies that are under discussion in this paper . such tests are urgently needed but they are difficult , even for quiescent galaxies and for `` normal '' star forming galaxies in the local universe ( muzzin et al . 2009b ; conroy 2013 ) . one interpretation of fig . [ sigsig.fig]b is that the stellar masses are off by factors up to @xmath7 ; however , as we show in the remainder of sect . 5 the dynamical masses and stellar masses are consistent with each other once orientation effects and the spatial extent of the gas are taken into account . our final dynamical result ( @xmath480 ; sect . [ rotcurve.sec ] ) suggests that the contributions of dark matter and gas to the mass within @xmath356kpc are small . we have assumed a relatively bottom - light chabrier ( 2003 ) imf when deriving stellar masses ; if we assume a salpeter ( 1955 ) imf instead ( see , e.g. , van dokkum & conroy 2010 ; conroy & van dokkum 2012 ; cappellari et al . 2012 ) we find @xmath481 , and even tighter constraints on the amount of gas and dark matter . we emphasize , however , that the conversion of light to stellar mass for these dusty , compact star forming galaxies is highly uncertain . we also note here that the stellar masses are not corrected for the contribution of emission lines to the seds . these corrections are generally small ( @xmath7% ) . second , the role of winds and active nuclei in these galaxies is not well understood ( sect . [ complex.sec ] ) . they almost certainly influence the measured dynamics and line ratios , but without spatially - resolved data it is very difficult to disentangle the effects of winds , a falling rotation curve , and the spatial distribution of the ionized gas . third , the fact that the galaxies are all very dusty may imply that we are missing part of the population due to selection effects ( sect . [ extreme.sec ] ) . we could be missing galaxies outright ( see fig . 3 in nelson et al . 2014 ) , or they could be misclassified as less compact , lower mass galaxies if only their outer edges are detected in the currently available data . another potential effect of the dust is that the stellar population modeling may produce incorrect stellar masses : the modeling uses a screen approximation for dust , whereas in reality the dust and stars are almost certainly mixed . fortunately , the prospects for addressing these uncertainties are excellent . adaptive optics - assisted spectroscopic observations with integral field units on 8 m 10 m telescopes can be used to measure kinematics and line ratios on @xmath482kpc scales ( e.g. , newman et al . the morphology of the dust and molecular gas emission can be studied with interferometers such as the very large array , the plateau de bure interferometer , and the atacama large millimeter array ( see , e.g. , simpson et al . 2015 , for impressive early alma results on submm - selected galaxies ) . these instruments can also measure the kinematics of the molecular gas ( e.g. , tacconi et al . 2008 ) . on a longer timescale , the james webb space telescope can measure the stellar kinematics of the galaxies , as well as identify and characterize compact galaxies that are entirely obscured in the @xmath5 band ( wang et al . finally , the upcoming generation of extremely large ground - based optical / near - ir telescopes is needed to spatially resolve these compact , massive galaxies within their effective radius . we thank adi zolotov for providing model tracks of galaxies in digital form , and thorsten naab and peter johansson for pointing us to the key figures that show model tracks in their simulations . the comments from the anonymous referee improved the manuscript , and prompted us to add the three appendices . support from stsci grant go-12177 is gratefully acknowledged . the data presented herein were obtained at the w. m. keck observatory , which is operated as a scientific partnership among the california institute of technology , the university of california and the national aeronautics and space administration . the observatory was made possible by the generous financial support of the w. m. keck foundation . the authors recognize and acknowledge the very significant cultural role and reverence that the summit of mauna kea has always had within the indigenous hawaiian community . we are most fortunate to have the opportunity to conduct observations from this mountain . , a. , garca - marn , m. , rodrguez zaurn , j. , monreal - ibero , a. , colina , l. , & arribas , s. 2010 , , 522 , a7 , j. a. , phillips , m. m. , & terlevich , r. 1981 , , 93 , 5 , m. , alaghband - zadeh , s. , hewett , p. c. , & mcmahon , r. g. 2015 , , 447 , 3368 , m. , mcmahon , r. g. , hewett , p. c. , alaghband - zadeh , s. , gonzalez - solares , e. , venemans , b. p. , & hawthorn , m. j. 2012 , , 427 , 2275 , m. , huler , b. , peng , c. y. , mcintosh , d. h. , & guo , y. 2012 , , 422 , 449 , a. j. , cowie , l. l. , mushotzky , r. f. , & richards , e. a. 2001 , , 121 , 662 , a. j. , cowie , l. l. , sanders , d. b. , fulton , e. , taniguchi , y. , sato , y. , kawara , k. , & okuda , h. 1998 , , 394 , 248 , g. , faber , s. m. , prez - gonzlez , p. g. , koo , d. c. , williams , c. c. , kocevski , d. d. , trump , j. r. , mozena , m. , et al . 2013 , , 765 , 104 , g. , faber , s. m. , prez - gonzlez , p. g. , pacifici , c. , trump , j. r. , koo , d. c. , wuyts , s. , guo , y. , et al . 2014a , , 791 , 52 , g. , trump , j. r. , koo , d. c. , dekel , a. , kassin , s. a. , kocevski , d. d. , faber , s. m. , van der wel , a. , et al . 2014b , , 795 , 145 , t. c. , flynn , k. , & gebhardt , k. 1990 , , 100 , 32 , p. s. , marchesini , d. , wechsler , r. h. , muzzin , a. , papovich , c. , & stefanon , m. 2013 , , 777 , l10 , e. f. , van der wel , a. , papovich , c. , kocevski , d. , lotz , j. , mcintosh , d. h. , kartaltepe , j. , faber , s. m. , et al . 2012 , , 753 , 167 , s. , newman , a. b. , & ellis , r. s. 2014a , , 783 , 117 , s. , newman , a. b. , ellis , r. s. , & konidaris , n. p. 2014b , , 788 , l29 , e. & arnouts , s. 1996 , , 117 , 393 , r. , franx , m. , & van dokkum , p. g. 2015 , , 799 , 148 , r. , van dokkum , p. g. , franx , m. , brammer , g. b. , brinchmann , j. , kriek , m. , labb , i. , quadri , r. f. , et al . 2011 , , 737 , l31 + , r. , van dokkum , p. g. , tal , t. , marchesini , d. , kriek , m. , franx , m. , & coppi , p. 2009 , , 697 , 1290 , r. , van dokkum , p. g. , van de sande , j. , franx , m. , leja , j. , & kriek , m. 2013 , , 779 , l21 , g. b. , snchez - janssen , r. , labb , i. , da cunha , e. , erb , d. k. , franx , m. , fumagalli , m. , lundgren , b. , et al . 2012a , , 758 , l17 , g. b. , van dokkum , p. g. , & coppi , p. 2008 , , 686 , 1503 , g. b. , van dokkum , p. g. , franx , m. , fumagalli , m. , patel , s. , rix , h .- w . , skelton , r. e. , kriek , m. , et al . 2012b , , 200 , 13 , g. b. , whitaker , k. e. , van dokkum , p. g. , marchesini , d. , franx , m. , kriek , m. , labbe , i. , lee , k .- s . , 2011 , , 739 , 24 , j. , pettini , m. , & charlot , s. 2008 , , 385 , 769 , v. a. , dunlop , j. s. , mclure , r. j. , cirasuolo , m. , buitrago , f. , bowler , r. a. a. , targett , t. a. , bell , e. f. , et al . 2014 , , 444 , 1660 , d. , armus , l. , bohlin , r. c. , kinney , a. l. , koornneef , j. , & storchi - bergmann , t. 2000 , , 533 , 682 , m. , bacon , r. , bureau , m. , damen , m. c. , davies , r. l. , de zeeuw , p. t. , emsellem , e. , falcn - barroso , j. , et al . 2006 , , 366 , 1126 , m. , mcdermid , r. m. , alatalo , k. , blitz , l. , bois , m. , bournaud , f. , bureau , m. , crocker , a. f. , et al . 2012 , nature , 484 , 485 , m. , mcdermid , r. m. , alatalo , k. , blitz , l. , bois , m. , bournaud , f. , bureau , m. , crocker , a. f. , et al . 2013 , , 432 , 1862 , k. i. , michaowski , m. j. , krips , m. , geach , j. e. , ashby , m. l. n. , huang , j .- s . , fazio , g. g. , koekemoer , a. m. , et al . 2014 , , 788 , 126 , c. m. , bschorr , t. j. , renzini , a. , lilly , s. j. , capak , p. , cibinel , a. , ilbert , o. , onodera , m. , et al . 2013 , , 773 , 112 , c. m. , berta , s. , bthermin , m. , bock , j. , bridge , c. , burgarella , d. , chapin , e. , chapman , s. c. , et al . 2012 , , 761 , 139 , p. , giavalisco , m. , guo , y. , renzini , a. , ferguson , h. , koekemoer , a. m. , salimbeni , s. , scarlata , c. , et al . 2011 , , 743 , 96 , p. , giavalisco , m. , williams , c. c. , guo , y. , lee , b. , renzini , a. , ferguson , h. , faber , s. f. , et al . 2013 , , 775 , 106 , g. 2003 , , 115 , 763 , y .- y . , van der wel , a. , rix , h .- w . , holden , b. , bell , e. f. , mcgrath , e. j. , wuyts , s. , hussler , b. , et al . 2013 , , 773 , 149 , c. 2013 , , 51 , 393 , c. , van dokkum , p. , & kravtsov , a. 2015 , , 803 , 77 , c. & van dokkum , p. g. 2012 , , 760 , 71 , c. 2014 , ara&a , 52 , 291 , g. , hicks , e. k. s. , genzel , r. , frster schreiber , n. m. , davies , r. , bouch , n. , buschkamp , p. , et al . 2009 , , 697 , 115 , d. j. , springel , v. , white , s. d. m. , de lucia , g. , frenk , c. s. , gao , l. , jenkins , a. , kauffmann , g. , et al . 2006 , , 365 , 11 , e. , alexander , d. m. , dickinson , m. , gilli , r. , renzini , a. , elbaz , d. , cimatti , a. , chary , r. , et al . 2007 , , 670 , 173 , e. , bournaud , f. , walter , f. , dannerbauer , h. , carilli , c. l. , dickinson , m. , elbaz , d. , morrison , g. e. , et al . 2010 , , 713 , 686 , e. , renzini , a. , pirzkal , n. , cimatti , a. , malhotra , s. , stiavelli , m. , xu , c. , pasquali , a. , et al . 2005 , , 626 , 680 , i. , abraham , r. g. , glazebrook , k. , mccarthy , p. j. , caris , e. , carlberg , r. g. , chen , h .- w . , crampton , d. , et al . 2011 , , 739 , l44 , r. , ho , l. c. , peng , c. y. , & huang , s. 2014 , , 787 , 69 , a. , birnboim , y. , engel , g. , freundlich , j. , goerdt , t. , mumcuoglu , m. , neistein , e. , pichon , c. , et al . 2009 , , 457 , 451 , a. & burkert , a. 2014 , , 438 , 1870 , g. , terlevich , r. , & terlevich , e. 2002 , , 330 , 69 , m. a. & sutherland , r. s. 1995 , , 455 , 468 , d. & solomon , p. m. 1998 , , 507 , 615 , d. , dickinson , m. , hwang , h. s. , daz - santos , t. , magdis , g. , magnelli , b. , le borgne , d. , galliano , f. , et al . 2011 , , 533 , a119 , d. k. , shapley , a. e. , steidel , c. c. , pettini , m. , adelberger , k. l. , hunt , m. p. , moorwood , a. f. m. , & cuby , j. 2003 , , 591 , 101 , f. , grazian , a. , santini , p. , puccetti , s. , brusa , m. , feruglio , c. , fontana , a. , giallongo , e. , et al . 2008 , , 672 , 94 , d. , hogg , d. w. , lang , d. , & goodman , j. 2013 , , 125 , 306 , n. m. , genzel , r. , bouch , n. , cresci , g. , davies , r. , buschkamp , p. , shapiro , k. , tacconi , l. j. , et al . 2009 , , 706 , 1364 , n. m. , genzel , r. , newman , s. f. , kurk , j. d. , lutz , d. , tacconi , l. j. , wuyts , s. , bandara , k. , et al . 2014 , , 787 , 38 , n. m. , shapley , a. e. , erb , d. k. , genzel , r. , steidel , c. c. , bouch , n. , cresci , g. , & davies , r. 2011 , , 731 , 65 , m. 1993 , in iau symposium , vol . 153 , galactic bulges , ed . h. dejonghe & h. j. habing , 243 , m. , illingworth , g. , & de zeeuw , t. 1991 , , 383 , 112 , m. , illingworth , g. d. , kelson , d. d. , van dokkum , p. g. , & tran , k. 1997 , , 486 , l75 , m. , van dokkum , p. g. , schreiber , n. m. f. , wuyts , s. , labb , i. , & toft , s. 2008 , , 688 , 770 , m. , patel , s. g. , franx , m. , brammer , g. , van dokkum , p. , da cunha , e. , kriek , m. , lundgren , b. , et al . 2012 , , 757 , l22 , m. , labb , i. , patel , s. g. , franx , m. , van dokkum , p. , brammer , g. , da cunha , e. , frster schreiber , n. m. , et al . 2014 , , 796 , 35 , a. , charlot , s. , brinchmann , j. , white , s. d. m. , & tremonti , c. a. 2005 , , 362 , 41 , s. , naab , t. , genzel , r. , frster schreiber , n. m. , sternberg , a. , oser , l. , johansson , p. h. , dav , r. , et al . 2012 , , 745 , 11 , r. , burkert , a. , bouch , n. , cresci , g. , frster schreiber , n. m. , shapley , a. , shapiro , k. , tacconi , l. j. , et al . 2008 , , 687 , 59 , r. , frster schreiber , n. m. , rosario , d. , lang , p. , lutz , d. , wisnioski , e. , wuyts , e. , wuyts , s. , et al . 2014a , , 796 , 7 , r. , frster schreiber , n. m. , lang , p. , tacchella , s. , tacconi , l. j. , wuyts , s. , bandara , k. , burkert , a. , et al . 2014b , , 785 , 75 , r. , tacconi , l. j. , gracia - carpio , j. , sternberg , a. , cooper , m. c. , shapiro , k. , bolatto , a. , bouch , n. , et al . 2010 , , 407 , 2091 , r. , tacconi , l. j. , lutz , d. , saintonge , a. , berta , s. , magnelli , b. , combes , f. , garca - burillo , s. , et al . 2015 , , 800 , 20 , r. , norman , c. , vignali , c. , vanzella , e. , calura , f. , pozzi , f. , massardi , m. , mignano , a. , et al . 2014 , , 562 , a67 , a. w. , dullo , b. t. , & savorgnan , g. a. d. 2015 , , 804 , 32 , n. a. , kocevski , d. d. , faber , s. m. , ferguson , h. c. , koekemoer , a. m. , riess , a. g. , acquaviva , v. , alexander , d. m. , et al . 2011 , , 197 , 35 , f. , flores , h. , puech , m. , yang , y. b. , athanassoula , e. , rodrigues , m. , & delgado , r. 2009 , a&a , 507 , 1313 , t. m. , armus , l. , & miley , g. k. 1987 , , 93 , 276 . 1990 , , 74 , 833 , m. , naab , t. , & ostriker , j. p. 2013 , , 429 , 2924 , m. , naab , t. , dav , r. , oppenheimer , b. d. , ostriker , j. p. , somerville , r. s. , oser , l. , genzel , r. , et al . 2013 , , 436 , 2929 , p. f. , bundy , k. , hernquist , l. , wuyts , s. , & cox , t. j. 2010 , , 401 , 1099 , p. f. , bundy , k. , murray , n. , quataert , e. , lauer , t. r. , & ma , c .- p . 2009a , , 398 , 898 , p. f. , cox , t. j. , kere , d. , & hernquist , l. 2008 , , 175 , 390 , p. f. , hernquist , l. , cox , t. j. , keres , d. , & wuyts , s. 2009b , , 691 , 1424 , p. h. , naab , t. , & ostriker , j. p. 2009 , , 697 , l38 . 2012 , , 754 , 115 , s. a. , weiner , b. j. , faber , s. m. , gardner , j. p. , willmer , c. n. a. , coil , a. l. , cooper , m. c. , devriendt , j. , et al . 2012 , , 758 , 106 , g. , heckman , t. m. , white , s. d. m. , charlot , s. , tremonti , c. , peng , e. w. , seibert , m. , brinkmann , j. , et al . 2003 , , 341 , 54 , r. c. 1998 , , 36 , 189 , l. j. , dopita , m. a. , leitherer , c. , dav , r. , yuan , t. , allen , m. , groves , b. , & sutherland , r. 2013 , , 774 , 100 , a. m. , faber , s. m. , ferguson , h. c. , grogin , n. a. , kocevski , d. d. , koo , d. c. , lai , k. , lotz , j. m. , et al . 2011 , , 197 , 36 , j. & sanders , d. b. 1992 , , 390 , l53 , m. , shapley , a. e. , reddy , n. a. , siana , b. , coil , a. l. , mobasher , b. , freeman , w. r. , de groot , l. , et al . 2015 , , 218 , 15 , m. , van dokkum , p. g. , franx , m. , illingworth , g. d. , coppi , p. , frster schreiber , n. m. , gawiser , e. , labb , i. , et al . 2007 , , 669 , 776 , m. , van dokkum , p. g. , franx , m. , illingworth , g. d. , & magee , d. k. 2009a , , 705 , l71 , m. , van dokkum , p. g. , labb , i. , franx , m. , illingworth , g. d. , marchesini , d. , & quadri , r. f. 2009b , , 700 , 221 , j. 1995 , in asp conf . 77 : astronomical data analysis software and systems iv , vol . 4 , 349 , i. , huang , j. , franx , m. , rudnick , g. , barmby , p. , daddi , e. , van dokkum , p. g. , fazio , g. g. , et al . 2005 , , 624 , l81 , l. , lehnert , m. d. , van driel , w. , nesvadba , n. p. h. , & di matteo , p. 2011 , , 534 , l4 , j. , van dokkum , p. , & franx , m. 2013a , , 766 , 33 , j. , van dokkum , p. g. , franx , m. , & whitaker , k. e. 2015 , , 798 , 115 , j. , van dokkum , p. g. , momcheva , i. , brammer , g. , skelton , r. e. , whitaker , k. e. , andrews , b. h. , franx , m. , et al . 2013b , , 778 , l24 , j. , tremaine , s. , richstone , d. , bender , r. , bower , g. , dressler , a. , faber , s. m. , gebhardt , k. , et al . 1998 , , 115 , 2285 , d. , van dokkum , p. g. , frster schreiber , n. m. , franx , m. , labb , i. , & wuyts , s. 2009 , , 701 , 1765 , z. c. , marchesini , d. , brammer , g. b. , stefanon , m. , muzzin , a. , fernndez - soto , a. , geier , s. , hainline , k. n. , et al . 2015 , , 801 , 133 , i. s. , becklin , e. e. , bendiksen , o. , brims , g. , canfield , j. , figer , d. f. , graham , j. r. , hare , j. , et al . 1998 , in society of photo - optical instrumentation engineers ( spie ) conference series , vol . 3354 , infrared astronomical instrumentation , ed . a. m. fowler , 566578 , i. s. , steidel , c. c. , epps , h. w. , konidaris , n. , matthews , k. y. , adkins , s. , aliado , t. , brims , g. , et al . 2012 , in society of photo - optical instrumentation engineers ( spie ) conference series , vol . 8446 , society of photo - optical instrumentation engineers ( spie ) conference series , 0 , e. , abraham , r. g. , glazebrook , k. , mccarthy , p. j. , yan , h. , odonnell , d. v. , le borgne , d. , savaglio , s. , et al . 2009 , , 706 , 1020 , s. h. , bundy , k. , sullivan , m. , ellis , r. s. , & treu , t. 2011 , , 741 , 115 , h. j. , mao , s. , & white , s. d. m. 1998 , , 295 , 319 , j. & kennicutt , jr . , r. c. 2006 , , 164 , 81 , a. , marchesini , d. , stefanon , m. , franx , m. , mccracken , h. j. , milvang - jensen , b. , dunlop , j. s. , fynbo , j. p. u. , et al . 2013 , , 777 , 18 , a. , marchesini , d. , van dokkum , p. g. , labb , i. , kriek , m. , & franx , m. 2009a , , 701 , 1839 , a. , van dokkum , p. , franx , m. , marchesini , d. , kriek , m. , & labb , i. 2009b , , 706 , l188 , a. , van dokkum , p. , kriek , m. , labb , i. , cury , i. , marchesini , d. , & franx , m. 2010 , , 725 , 742 , t. , johansson , p. h. , & ostriker , j. p. 2009 , , 699 , l178 , e. , van dokkum , p. , franx , m. , brammer , g. , momcheva , i. , schreiber , n. f. , da cunha , e. , tacconi , l. , et al . 2014 , , 513 , 394 , e. j. , van dokkum , p. g. , brammer , g. , frster schreiber , n. , franx , m. , fumagalli , m. , patel , s. , rix , h .- w . , 2012 , , 747 , l28 , e. j. , van dokkum , p. g. , momcheva , i. , brammer , g. , lundgren , b. , skelton , r. e. , whitaker , k. e. , da cunha , e. , et al . 2013 , , 763 , l16 , a. b. , ellis , r. s. , bundy , k. , & treu , t. 2012 , , 746 , 162 , s. f. , genzel , r. , frster schreiber , n. m. , shapiro griffin , k. , mancini , c. , lilly , s. j. , renzini , a. , bouch , n. , et al . 2013 , , 767 , 104 , p. a. , van dokkum , p. g. , illingworth , g. d. , bouwens , r. j. , momcheva , i. , holden , b. , roberts - borsani , g. w. , smit , r. , et al . 2015 , , 804 , l30 , b. d. & dav , r. 2006 , , 373 , 1265 , l. , ostriker , j. p. , naab , t. , johansson , p. h. , & burkert , a. 2010 , , 725 , 2312 , j. r. , conselice , c. j. , mortlock , a. , hartley , w. g. , almaini , o. , duncan , k. , & mundy , c. j. 2014 , , 445 , 2198 , c. , labb , i. , quadri , r. , tilvi , v. , behroozi , p. , bell , e. f. , glazebrook , k. , spitler , l. , et al . 2015 , , 803 , 26 , c. , moustakas , l. a. , dickinson , m. , le floch , e. , rieke , g. h. , daddi , e. , alexander , d. m. , bauer , f. , et al . 2006 , , 640 , 92 , s. g. , van dokkum , p. g. , franx , m. , quadri , r. f. , muzzin , a. , marchesini , d. , williams , r. j. , holden , b. p. , et al . 2013 , , 766 , 15 , c. y. , ho , l. c. , impey , c. d. , & rix , h .- w . 2002 , , 124 , 266 , m. , kellogg , m. , steidel , c. c. , dickinson , m. , adelberger , k. l. , & giavalisco , m. 1998 , , 508 , 539 , m. & pagel , b. e. j. 2004 , , 348 , l59 , b. m. , calvi , r. , bindoni , d. , donofrio , m. , moretti , a. , valentinuzzi , t. , fasano , g. , fritz , j. , et al . 2013 , , 762 , 77 , s. h. , kriek , m. , brammer , g. b. , conroy , c. , frster schreiber , n. m. , franx , m. , fumagalli , m. , lundgren , b. , et al . 2014 , , 788 , 86 , d. , stone , j. m. , & kallman , t. r. 2000 , , 543 , 686 , n. a. , kriek , m. , shapley , a. e. , freeman , w. r. , siana , b. , coil , a. l. , mobasher , b. , price , s. h. , et al . 2015 , , 806 , 259 , d. a. , bradford , c. m. , clements , d. l. , dowell , c. d. , prez - fournon , i. , ivison , r. j. , bridge , c. , conley , a. , et al . 2013 , , 496 , 329 , h. , guhathakurta , p. , colless , m. , & ing , k. 1997 , , 285 , 779 , g. , renzini , a. , daddi , e. , baronchelli , i. , berta , s. , cresci , g. , franceschini , a. , gruppioni , c. , et al . 2014 , , 443 , 19 , e. e. 1955 , , 121 , 161 , d. b. & mirabel , i. f. 1996 , , 34 , 749 , n. , sheth , k. , aussel , h. , vanden bout , p. , capak , p. , bongiorno , a. , casey , c. m. , murchikova , l. , et al . 2015 , apjl , submitted ( arxiv:1505.02159 ) , j. l. 1968 , atlas de galaxias australes ( cordoba , argentina : observatorio astronomico , 1968 ) , a. e. , reddy , n. a. , kriek , m. , freeman , w. r. , sanders , r. l. , siana , b. , coil , a. l. , mobasher , b. , et al . 2015 , , 801 , 88 , j. m. , smail , i. , swinbank , a. m. , almaini , o. , blain , a. w. , bremer , m. n. , chapman , s. c. , chen , c .- c . , 2015 , , 799 , 81 , r. e. , whitaker , k. e. , momcheva , i. g. , brammer , g. b. , van dokkum , p. g. , labb , i. , franx , m. , van der wel , a. , et al . 2014 , , 214 , 24 , i. , ivison , r. j. , owen , f. n. , blain , a. w. , & kneib , j .- p . 2000 , , 528 , 612 , k. t. & martin , c. l. 2012 , , 203 , 3 , c. c. , rudie , g. c. , strom , a. l. , pettini , m. , reddy , n. a. , shapley , a. e. , trainor , r. f. , erb , d. k. , et al . 2014 , , 795 , 165 , c. m. s. , labbe , i. , spitler , l. r. , glazebrook , k. , tomczak , a. , allen , r. , brammer , g. b. , cowley , m. , et al . 2015 , , 808 , l29 , d. , franx , m. , & van dokkum , p. g. 2012 , , 749 , 121 , s. , carollo , c. m. , renzini , a. , frster schreiber , n. m. , lang , p. , wuyts , s. , cresci , g. , dekel , a. , et al . 2015 , science , 348 , 314 , l. j. , genzel , r. , neri , r. , cox , p. , cooper , m. c. , shapiro , k. , bolatto , a. , bouch , n. , et al . 2010 , , 463 , 781 , l. j. , genzel , r. , smail , i. , neri , r. , chapman , s. c. , ivison , r. j. , blain , a. , cox , p. , et al . 2008 , , 680 , 246 , e. n. , franx , m. , brinchmann , j. , van der wel , a. , & van dokkum , p. g. 2010a , , 722 , 1 , e. n. , franx , m. , glazebrook , k. , brinchmann , j. , van der wel , a. , & van dokkum , p. g. 2010b , , 720 , 723 , s. , smoli , v. , magnelli , b. , karim , a. , zirm , a. , michalowski , m. , capak , p. , sheth , k. , et al . 2014 , , 782 , 68 , i. , frster schreiber , n. m. , rudnick , g. , barden , m. , franx , m. , rix , h .- w . , caldwell , j. a. r. , mcintosh , d. h. , et al . 2006 , , 650 , 18 , i. , ferreras , i. , & de la rosa , i. g. 2011 , , 415 , 3903 , d. , kriek , m. , labb , i. , conroy , c. , & fumagalli , m. 2014 , , 783 , l30 , j. , kriek , m. , franx , m. , van dokkum , p. g. , bezanson , r. , bouwens , r. j. , quadri , r. f. , rix , h .- w . , 2013 , , 771 , 85 , a. , bell , e. f. , hussler , b. , mcgrath , e. j. , chang , y .- y . , guo , y. , mcintosh , d. h. , rix , h .- w . , et al . 2012 , , 203 , 24 , a. , chang , y .- y . , bell , e. f. , holden , b. p. , ferguson , h. c. , giavalisco , m. , rix , h .- w . , skelton , r. , et al . 2014a , , 792 , l6 , a. , franx , m. , van dokkum , p. g. , skelton , r. e. , momcheva , i. g. , whitaker , k. e. , brammer , g. b. , bell , e. f. , et al . 2014b , , 788 , 28 , p. g. , bezanson , r. , van der wel , a. , nelson , e. j. , momcheva , i. , skelton , r. e. , whitaker , k. e. , brammer , g. , et al . 2014 , , 791 , 45 , p. g. , brammer , g. , fumagalli , m. , nelson , e. , franx , m. , rix , h .- w . , kriek , m. , skelton , r. e. , et al . 2011 , , 743 , l15 , p. g. & conroy , c. 2010 , , 468 , 940 , p. g. , franx , m. , f " orster schreiber , n. m. , illingworth , g. d. , daddi , e. , knudsen , k. k. , labb ' e , i. , moorwood , a. , et al . 2004 , , 611 , 703 , p. g. , franx , m. , kriek , m. , holden , b. , illingworth , g. d. , magee , d. , bouwens , r. , marchesini , d. , et al . 2008 , , 677 , l5 , p. g. , kriek , m. , & franx , m. 2009 , , 460 , 717 , p. g. , kriek , m. , rodgers , b. , franx , m. , & puxley , p. 2005 , , 622 , l13 , p. g. , leja , j. , nelson , e. j. , patel , s. , skelton , r. e. , momcheva , i. , brammer , g. , whitaker , k. e. , et al . 2013 , , 771 , l35 , p. g. & van der marel , r. p. 2007 , , 655 , 30 , p. g. , whitaker , k. e. , brammer , g. , franx , m. , kriek , m. , labb , i. , marchesini , d. , quadri , r. , et al . 2010 , , 709 , 1018 , m. , genel , s. , springel , v. , torrey , p. , sijacki , d. , xu , d. , snyder , g. , nelson , d. , et al . 2014 , , 444 , 1518 , n. p. , forbes , d. a. , phillips , a. c. , gronwall , c. , faber , s. m. , illingworth , g. d. , & koo , d. c. 1996 , , 465 , l15 , f. , decarli , r. , carilli , c. , bertoldi , f. , cox , p. , da cunha , e. , daddi , e. , dickinson , m. , et al . 2012 , , 486 , 233 , w .- h . , barger , a. j. , & cowie , l. l. 2012 , , 744 , 155 , b. j. , willmer , c. n. a. , faber , s. m. , melbourne , j. , kassin , s. a. , phillips , a. c. , harker , j. , metevier , a. j. , et al . 2006 , , 653 , 1027 , s. , torrey , p. , ma , c .- p . , rodriguez - gomez , v. , vogelsberger , m. , kriek , m. , van dokkum , p. , nelson , e. , et al . 2015 , , 449 , 361 , m. s. , smith , l. j. , gallagher , iii , j. s. , trancho , g. , bastian , n. , & konstantopoulos , i. s. 2009 , , 696 , 192 , k. e. , franx , m. , leja , j. , van dokkum , p. g. , henry , a. , skelton , r. e. , fumagalli , m. , momcheva , i. g. , et al . 2014 , , 795 , 104 , k. e. , labb , i. , van dokkum , p. g. , brammer , g. , kriek , m. , marchesini , d. , quadri , r. f. , franx , m. , et al . 2011 , , 735 , 86 , k. e. , van dokkum , p. g. , brammer , g. , & franx , m. 2012 , , 754 , l29 , t. conselice , c. j. , dahlen , t. , dickinson , m. e. , ferguson , h. c. , grogin , n. a. , guo , y. , koekemoer , a. m. , et al . 2014 , , 785 , 111 , c. c. , giavalisco , m. , cassata , p. , tundo , e. , wiklind , t. , guo , y. , lee , b. , barro , g. , et al . 2014 , , 780 , 1 , c. c. , giavalisco , m. , lee , b. , tundo , e. , mobasher , b. , nayyeri , h. , ferguson , h. c. , koekemoer , a. , et al . 2015 , , 800 , 21 , r. j. , quadri , r. f. , franx , m. , van dokkum , p. , toft , s. , kriek , m. , & labb , i. 2010 , , 713 , 738 , g. , faber , s. m. , & gonzalez , j. j. 1992 , , 398 , 69 , s. , cox , t. j. , hayward , c. c. , franx , m. , hernquist , l. , hopkins , p. f. , jonsson , p. , & van dokkum , p. g. 2010 , , 722 , 1666 , s. , forster schreiber , n. m. , van der wel , a. , magnelli , b. , guo , y. , genzel , r. , lutz , d. , aussel , h. , et al . 2011 , , 742 , 96 , s. , labb , i. , franx , m. , rudnick , g. , van dokkum , p. g. , fazio , g. g. , frster schreiber , n. m. , huang , j. , et al . 2007 , , 655 , 51 , a. , dekel , a. , mandelker , n. , tweed , d. , inoue , s. , degraf , c. , ceverino , d. , primack , j. r. , et al . 2015 , , 450 , 2327 in the main text we show color images of the 25 star forming compact massive galaxies , created from the @xmath28 and @xmath30 candels data ( fig . 6 ) . in fig . [ hband_color.fig ] we show the @xmath30 images separately , with a higher dynamic range than in fig . 6 . the tidal features around goods - s_30274 and cosmos_11363 are very clear , and several other galaxies also show structure at faint surface brightness . we fit all galaxies with a single sersic profile , which is an excellent approximation of the average surface brightness profile of the full sample ( see sect . [ sb.sec ] ) ; however , it is clear that these fits do not capture the full information in the hst images . as described in sect . 3.4.1 we fit gaussian models to the emission lines . the fits are done with the emcee code , with the observed 1d spectrum and a noise model as inputs for each galaxy . here we briefly analyze the residuals from these fits to determine the accuracy of the noise models . in fig . [ spectra_noise.fig ] we show the spectra of the 20 galaxies that were observed by us . for convenience , the figure has the same format as figs . 5 , 6 , 7 , and 15 in the main text , except that the five galaxies from barro et al . ( 2014 ) are left blank . for each galaxy three subpanels are shown . the top subpanel is identical to the main panel of fig . 5 , and shows the observed spectrum in black along with the best - fitting model in red . the middle subpanel shows the noise model ( empirical in the case of mosfire and theoretical in the case of nirspec ; see sect . 3.1 and sect . the bottom subpanel is the residual from the fit divided by the noise model . the residuals are well - behaved , and generally exhibit no indications of poorly subtracted sky lines or other irregularities . we quantified this by calculating the biweight scatter @xmath483 ( see beers et al . 1990 ) in the distribution of residuals . the value of @xmath483 deviates by more than @xmath406% from unity in only two cases , uds_35673 and cosmos_11363 . both galaxies have very high s / n ratio spectra , and the higher than expected residuals are not caused by errors in the noise spectra but by the fact that the velocity distributions are not exactly gaussian . the average scatter of the remaining 18 galaxies is @xmath484 , which means that the noise models that we use are accurate to @xmath7% . in sect . [ rotcurve.sec ] we construct the average rotation curve for star forming compact massive galaxies . this is done by combining information for 10 different galaxies : all galaxies have approximately the same stellar masses and @xmath30 half - light radii , but they have a wide range of h@xmath8 effective radii . for each galaxy we measure the galaxy - integrated velocity dispersion and the inclination , and convert these to an inclination - corrected rotation velocity at @xmath357 , where @xmath314 is the half - light radius of the h@xmath8 emission . the rotation velocities of the galaxies are then plotted versus @xmath314 in fig . [ kepler.fig ] , and the resulting relation is interpreted as a rotation curve . here we test whether this method is viable , that is , whether the actual rotation curve of a model galaxy can be reconstructed in this way . we also test whether we are using the correct conversion constant to go from a galaxy - integrated velocity dispersion to a rotation velocity at the half - light radius of h@xmath8 . this constant , together with an inclination correction , relates the velocity dispersion @xmath127 to the rotation velocity @xmath315 : @xmath485 ( see eq . [ siggen.eq ] and eq . [ correcti.eq ] ) . in the main text we use @xmath486 , based on previous studies ( see sect.[gasstars.sec ] ) . however , these studies did not consider the specific model of a compact , @xmath487-law mass distribution combined with an extended , exponential gas distribution . we simulated the observations in the following way . we constructed a model mass distribution that follows a sersic surface density profile . this mass distribution is characterized by three parameters : the sersic index @xmath85 , the effective radius @xmath488 ( this parameter is equivalent to both @xmath489 and @xmath87 in the main text ) , and the total mass @xmath490 . we fixed @xmath491kpc and @xmath492 , and for the initial model we set @xmath362 . apart from a slight rescaling of the effective radius , this model closely matches the actual average stellar mass distribution of the scmgs , if mass traces the @xmath30 light . the model is shown in fig . [ alpha_1.fig]a by the grey line . next , we constructed 10 model galaxies , each with the same mass distribution but with different distributions of the h@xmath8 emission . the ionized gas is in thin exponential disks , with effective radii ranging from @xmath493kpc ( and hence @xmath494 ) to @xmath495kpc . four of these model gas distributions are shown by the black lines in fig . [ alpha_1.fig]a . the gas disks mimic the derived extended ionized gas of scmgs , with @xmath496 equivalent to the parameter @xmath314 in the main text . galaxy - integrated velocity profiles were created by integrating the projected velocities along the line of sight and over the full spatial extent of the model galaxies . the velocities were calculated from the mass profile shown in fig . [ alpha_1.fig]a and weighted by the h@xmath8 flux . in order to model the observed profiles as closely as possible , we used an inclination of 60@xmath497 ( where @xmath498 is edge - on ) and an instrumental resolution of 60 ( in between the mosfire and nirspec resolution ) . the velocity profiles of the four model galaxies are shown in panels c - f of fig . [ alpha_1.fig ] . as expected they have the classic `` double - horned '' form that is characteristic of rotating disks . the profile is not the same for all four models even though the mass distribution , and hence the underlying velocity field , is identical in all cases . the more extended the h@xmath8 distribution is with respect to the mass , the narrower the profile becomes , and the more closely it resembles a gaussian . the reason for this behavior is that the h@xmath8 emission is more weighted toward larger radii , where the rotation velocity is lower . velocities in excess of @xmath499 are still sampled , but they have relatively low weight and are responsible for the high velocity tails of the profile . [ [ relation - between - global - dispersion - and - rotation - velocity - at - rr_rm - halpha ] ] relation between global dispersion and rotation velocity at @xmath500 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ we fitted gaussian models to the line profiles , just as we do in the data analysis described in the main text . these gaussian fits are shown by the red curves in panels c - f of fig . [ alpha_1.fig ] . the width of these gaussians decreases with increasing @xmath501 , as discussed above . we note here that the actual profile shape is not very well approximated by a gaussian , particular in panels c and d. interestingly , we see hints of double - horned profiles in the data for some of the galaxies ( e.g. , uds_16442 and , particularly , goods - n_774 , which was published in nelson et al . 2014 ) , although the s / n ratio is not high enough to quantify this . in fig . [ alpha_2.fig]a these measured galaxy - integrated velocity dispersions are plotted versus the half - light radii of the h@xmath8 disks , after correcting for inclination and instrumental broadening ( open squares ) . all ten galaxy models are shown , with h@xmath8 effective radii ranging from @xmath502 to @xmath503 . for comparison , the black curve shows the actual rotation curve of the galaxies . the squares show the same fall - off as the actual rotation curve , with a roughly constant multiplicitative offset . the solid squares are obtained by dividing the measured dispersions by 0.8 , which is the value of @xmath504 that we used in the analysis of sect . [ rotcurve.sec ] . they are in almost perfect agreement with the black curve , demonstrating that it is possible to reconstruct the average rotation curve of scmgs with our method . the analysis is generalized in fig . [ alpha_2.fig]b , where we show the value of @xmath8 as a function of the ratio of the effective radius of h@xmath8 and the effective radius of the mass . we repeated the analysis for different assumed mass profiles , ranging from exponential ( @xmath505 ; dotted ) to an @xmath487 law ( @xmath362 ; solid ) . the ratio between dispersion and rotation velocity at @xmath506 is remarkably constant : it does not vary appreciably either with @xmath333 or with @xmath85 . we conclude that the assumed value of @xmath507 is reasonable for the mass and h@xmath8 profiles discussed in this paper .	in this paper we study a key phase in the formation of massive galaxies : the transition of star forming galaxies into massive ( @xmath0 ) , compact ( @xmath1kpc ) quiescent galaxies , which takes place from @xmath2 to @xmath3 . we use hst grism redshifts and extensive photometry in all five 3d - hst / candels fields , more than doubling the area used previously for such studies , and combine these data with keck mosfire and nirspec spectroscopy . we first confirm that a population of massive , compact , _ star forming _ galaxies exists at @xmath4 , using @xmath5-band spectroscopy of 25 of these objects at @xmath6 . they have a median ratio of 0.6 , are highly obscured with sfr(tot)/sfr()@xmath7 , and have a large range of observed line widths . we infer from the kinematics and spatial distribution of h@xmath8 that the galaxies have rotating disks of ionized gas that are a factor of @xmath9 more extended than the stellar distribution . by combining measurements of individual galaxies , we find that the kinematics are consistent with a nearly keplerian fall - off from @xmath10 at 1kpc to @xmath11 at 7kpc , and that the total mass out to this radius is dominated by the dense stellar component . next , we study the size and mass evolution of the progenitors of compact massive galaxies . even though individual galaxies may have had complex histories with periods of compaction and mergers , we show that the _ population _ of progenitors likely followed a simple inside - out growth track in the size - mass plane of @xmath12 . this mode of growth gradually increases the stellar mass within a fixed physical radius , and galaxies quench when they reach a stellar density or velocity dispersion threshold . as shown in other studies , the mode of growth changes after quenching , as dry mergers take the galaxies on a relatively steep track in the size - mass plane .
the starting point of this investigation was the discussion of the carnot - bekenstein - process in the environment of a schwarzschild black hole @xcite . there it was assumed that the black hole is surrounded by a cloud of radiation with a local temperature according to tolman s equation @xcite @xmath0 where @xmath1 is the usual hawking temperature at @xmath2 . this equation also follows for the equilibrium distribution of photons in relativistic kinetic theory @xcite or from elementary thermodynamical gedanken experiments @xcite . these yield a modified formula for the efficiency of the carnot - bekenstein - process @xmath3 between two heat reservoirs at different height . in the equilibrium we have @xmath4 which implies ( [ 1 ] ) . all this holds as long as the back reaction of the radiation to the metric can be neglected . but , since @xmath5 , at the schwarzschild radius @xmath6 the temperature and the energy density diverges and can not longer be regarded as a mere perturbation . rather one would have to solve the field equation anew , this time allowing for the energy - stress tensor of the photon gas to act as a source of gravitation . it is not at all clear whether the resulting metric will be a modified black hole in some sense . so we have the following problem which may be considered independently of the original motivation : calculate the static , spherically symmetric metric of einstein s field equations with the energy - stress tensor of a perfect fluid @xmath7 consisting of photons , i.e. @xmath8 in section 2 the corresponding field equations are transformed into an autonomous two dimensional system of differential equations which is discussed in section 3 . in section 4 we investigate the metric of a photon star for @xmath9 and study its global properties by means of numerical solutions . physical characteristics like radius , mass and temperature are defined in section 5 . section 6 contains concluding remarks . with except of eq . ( [ 4 ] ) the problem stated above is just the well - known problem of constructing the interior solution of a star . we may choose coordinates @xmath10 such that the metric is given by @xmath11 where @xmath12 and @xmath13 are unknown functions and @xmath14 is the surface element of a unit sphere . with respect to these coordinates the field equations boil down to a system of three coupled differential equations ( cf . @xcite 6.2.36.2.5 ) @xmath15 we recall that tolman s equation @xcite was derived under the same assumptions we made , except spherical symmetry . hence we may adopt ( [ 1 ] ) or , equivalently , @xmath16 since @xmath17 from local statistical mechanics , where @xmath18 is the stefan - boltzmann constant . we write @xmath19 and obtain from ( [ 6 ] ) and ( [ 7 ] ) @xmath20 eq . ( [ 8 ] ) is identically satisfied , which confirms tolman s result . otherwise the system of differential equations would be overdetermined . the set of solutions @xmath21 of the system ( [ 12 ] ) , ( [ 13 ] ) can be parameterized by 3 parameters , e. g. @xmath22 and two initial values for @xmath12 and @xmath13 . but @xmath23 is invariant under the 2-parameter group of scale transformations @xmath24 thus there exists only a 1-parameter family of solutions looking qualitatively different . an equivalent second order equation is obtained by eliminating @xmath13 using @xmath25 and inserting the derivative of ( [ 15 ] ) into ( [ 12 ] ) : @xmath26 for @xmath27 we have the well - known equations which lead to a 2-parameter family of schwarzschild metrics . for @xmath28 we may scale every solution such that it becomes a solution with @xmath29 the remaining subgroup of ( [ 14 ] ) with @xmath30 may be used to simplify the differential equation ( [ 16 ] ) . we perform the transformation @xmath31 where @xmath32 is arbitrary . the resulting reduced equations read @xmath33 note that this system is autonomous . the resulting symmetry @xmath34 reflects the scale invariance of ( [ 16 ] ) w.r . to the subgroup @xmath30 . even if we can not solve it exactly , it seems to be better accessible for intuition and developing approximation schemes . any two different solutions of ( [ 19 ] ) , ( [ 20 ] ) correspond to different similarity classes of @xmath35 . we calculate some typical solutions of ( [ 19 ] ) , ( [ 20 ] ) numerically and display them as curves in the @xmath36-plane . it is obvious that @xmath37 is a stable stationary point of ( [ 19 ] ) , ( [ 20 ] ) which is an attractor of the whole open quadrant @xmath38 . it follows by inverting the transformation ( [ 18 ] ) that @xmath39 is an exact solution of ( [ 19 ] ) , ( [ 20 ] ) which is asymptotically approached by any other solution for @xmath40 . it follows that the space - time of photon stars is not asymptotically flat . the other exact solution of ( [ 19 ] ) , ( [ 20 ] ) @xmath41 yields @xmath42 and is hence unphysical . a typical solution curve of ( [ 19 ] ) , ( [ 20 ] ) starts from @xmath43 , @xmath44 , @xmath45 and runs close to the solution ( [ 23 ] ) until it reaches small values of @xmath46 . then according to the `` @xmath46 '' in the denominator of ( [ 20 ] ) , @xmath47 increases rapidly and the curve is turned up towards the @xmath48-axis . then it describes a parabolic - like bow and approaches the stationary point by a clockwise vortex . it is instructive to draw the general schwarzschild solutions ( @xmath49 ) @xmath50 into the @xmath36-diagram . they are given by the family of parabolas @xmath51 which approximate the solution curves of ( [ 19 ] ) , ( [ 20 ] ) having the same vertex at @xmath52 in this way , for each solution @xmath53 of ( [ 12 ] ) , ( [ 13 ] ) we can define a unique schwarzschild approximation @xmath54 . we now turn to the discussion of the solutions of ( [ 16 ] ) for @xmath55 which yield @xmath56 by ( [ 15 ] ) . to study the behaviour for smaller @xmath57 we expand @xmath12 into a laurent series and insert this series into ( [ 16 ] ) . it turns out that the series starts with @xmath58 in accordance with a schwarzschild solution for @xmath49 . the coefficients @xmath59 are left undetermined since they represent the two initial values for ( [ 16 ] ) . the next terms are uniquely determined by ( [ 16 ] ) . it is straight forward to calculate the first , say , 20 terms by using a computer algebra software like mathematica . here we only note down the first nonvanishing extra terms for @xmath12 and @xmath60 : @xmath61 for small @xmath57 , @xmath12 and @xmath13 are also approximated by schwarzschild solutions , but unlike the approximation discussed above , we have to choose @xmath62 in order to obtain a positive solution @xmath63 . @xmath12 can not change its sign in a continuous way , since @xmath64 implies @xmath65 by ( [ 16 ] ) . ( according to the scale invariance ( [ 14 ] ) , @xmath55 may be multiplied by @xmath66 , but this gives no physically different solution . ) thus we may state that for @xmath67 the metric looks like that of a schwarzschild black hole with negative mass , independent of @xmath22 . for the geodesic motion close to @xmath68 we may thus adopt the effective potential of schwarzschild theory ( @xcite 6.3.15 ) ( with @xmath69 ) : @xmath70 where @xmath71 it follows that @xmath68 can never be reached by particles or photons due to the infinite high potential barrier . although curvature blows up for @xmath67 , as in the schwarzschild case , the nature of the singularity is less harmful . we conjecture that it can not be regarded as a `` naked singularity '' in whatever technical sense ( see @xcite for details of the various definitions ) and that cauchy surfaces still exist in the spacetime given by ( [ 12 ] ) , ( [ 13 ] ) , if the line @xmath68 is excluded from the spacetime manifold . from a computational point of view , the singularity of @xmath55 at @xmath68 suggests to transform ( [ 16 ] ) into a differential equation for @xmath72 and to re - transform to @xmath55 after a numerical solution for @xmath73 has been obtained . we used the ndsolve - command of mathematica to produce the following numerical solutions : a typical solution is shown in fig . 2 . recall that for the schwarzschild metric @xmath56 diverges at @xmath74 and @xmath75 . for the solution of fig . 2 @xmath56 has a relatively sharp maximum at @xmath76 and @xmath77 is becoming small . for @xmath78 , @xmath12 and @xmath13 are comparable with their schwarzschild approximations @xmath79 and @xmath80 , if @xmath57 is not too large . for @xmath81 , @xmath12 remains small within some shell @xmath82 and diverges for @xmath83 according to ( [ 29 ] ) . by ( [ 9 ] ) this means that the energy density is concentrated within that shell and @xmath76 may be viewed as the `` radius of the photon star '' . other solutions with larger values of @xmath22 show a more diffuse cloud of photons and a less sharp maximum of @xmath13 , see fig . 3 and 4 . these solutions are all scaled to the same value of @xmath76 . we have seen that the set of solutions @xmath21 may be characterized by 3 parameters , e. g. @xmath59 and @xmath22 in ( [ 29 ] ) . from the analogy with the schwarzschild case ( @xmath49 ) we expect that only a 2-parameter family represents physically different spacetimes . in the schwarzschild case , one parameter is set to 1 by the choice of the units , and the remaining parameter @xmath84 distinguishes between black holes of different mass . more specifically , one postulates that the velocity of light , expressed by @xmath85 approaches 1 for @xmath86 . the metric then obtains the form @xmath87 in the case of the photon star , we can not proceed in the same way , since the metric will not be asymptotically flat . but instead we may postulate the `` gauge condition '' that the schwarzschild approximation of @xmath88 , defined in section 3 , should obey condition ( [ 34 ] ) . if this is not the case , one has to perform a suitable scale transformation ( [ 14 ] ) . in this way we obtain a two - fold of physically different solutions . we now consider physical parameters characterizing this two - fold of solutions . one could be the radius @xmath76 of the photon star defined above . in analogy to the schwarzschild theory ( @xcite 6.2.7 ) we introduce the ( gravitational ) mass function @xmath89 in the domain where @xmath90 this is the `` would - be - mass '' of an equivalent black hole . in the domain @xmath91 the interpretation of ( [ 35 ] ) is not so obvious . as to be expected from the above discussion of the metric for @xmath67 , it turns out that @xmath92 . a typical mass function is shown in fig . 5 , where also the `` proper mass '' @xmath93 and the difference @xmath94 is displayed . it may be , as in this case , that the majority of the photons are `` hidden '' by the apparent negative mass in the center with respect to gravitation . nevertheless , we could use @xmath95 as a further physical parameter characterizing a photon star . generally , by ( [ 35 ] ) and @xmath96 @xmath97 where @xmath98 if @xmath99 we have @xmath100 , as in the schwarzschild case . + as another physical parameter we consider the `` surface temperature '' @xmath101 if we have only a two - fold of physically different solutions , as we claimed above , @xmath102 should be a function of @xmath76 and @xmath103 . indeed , ( [ 12 ] ) together with @xmath104 shows that @xmath77 is a function of @xmath76 and @xmath105 , hence of @xmath76 and @xmath103 . then , by ( [ 39 ] ) and ( [ 9 ] ) , also @xmath102 depends only on @xmath76 and @xmath103 . since @xmath106 depends on @xmath107 , the result can be conveniently expressed by using planck units , indicated by a subscript p : @xmath108^{1/4}. \label{40}\ ] ] to give a numerical example , if we take @xmath103 as the mass of the sun , @xmath76 as the corresponding schwarzschild radius , we obtain @xmath109k . this would correspond to a very hard gamma radiation with a wavelength @xmath110 m. the temperature of this example is @xmath111k , since @xmath112 whereas @xmath113 for @xmath114 . it is difficult to assess the physical relevance of our findings . but one point seems to be clear : the global character of the solutions @xmath115 is completely different from the schwarzschild approximations @xmath116 , no matter how small @xmath22 is . so the class @xmath21 of solutions of ( [ 12 ] ) , ( [ 13 ] ) does not depend continuously on @xmath22 in any reasonable sense . any small amount of radiation will destroy the event horizon , at least if an equilibrium is approached . if this would also be the case in simulations of the birth of `` black holes '' by collapsing matter , then they would never be born , and perhaps wo nt exist at all . , _ bangs , crunches , whimpers , and shrieks _ , oxford university press , new york 1995 . , _ zur thermodynamik schwarzer lcher _ , diploma thesis , university of osnabrck , 1998 , + http:@xmath117www.physik.uni-osnabrueck.de@[email protected] , _ relativistische thermodynamik _ , vieweg , braunschweig 1980 . , _ relativity , thermodynamics and cosmology _ , clarendon press , oxford 1934 . , _ general relativity _ , the university of chicago press , chicago 1984 .	we discuss numerical solutions of einstein s field equation describing static , spherically symmetric conglomerations of a photon gas . these equations imply a back reaction of the metric on the energy density of the photon gas according to tolman s equation . the 3-fold of solutions corresponds to a class of physically different solutions which is parameterized by only two quantities , e. g. mass and surface temperature . the energy density is typically concentrated on a shell because the center contains a repelling singularity , which can , however , not be reached by timelike or null geodesics . the physical relevance of these solutions is completely open , although their existence may raise some doubts w.r . to the stability of black holes .
the dynamics of charge carriers in crystalline solids normally resembles that exhibited by free electrons , only with numerical values of parameters such as mass and gyromagnetic ratio renormalized due to the influence of the crystal structure @xcite . in certain instances , changes in the properties of band electrons turn out to be dramatic . few - layer samples of graphite @xcite are one example , and a new class of so - called _ topological _ materials @xcite another . here we focus on bilayer graphene @xcite ( blg ) , i.e. , two single - layer sheets of graphite stacked as shown in fig . [ fig : lattice](a ) . electrons in this material turn out to be pseudospin - carrying chiral fermions having a finite band mass but zero rest energy @xcite . our study shows that the interaction of these exotic charge carriers with electromagnetic fields is very unusual . normally , electric fields couple to electric charges at rest and moving magnetic moments , whereas magnetic fields couple to moving electric charges and magnetic moments at rest . in contrast , every coupling of a bilayer electron s degree of freedom to an electric field is matched by an analogous coupling to a magnetic field . we explore physical consequences of this magneto - electric equivalence , including anomalous polarizations of real spin and the layer - index - related pseudospin , as well as an anisotropic version of axion electrodynamics @xcite . the blg band structure near the @xmath0 point in the brillouin zone [ see fig . [ fig : lattice](b ) ] is described by the effective hamiltonian @xmath1 where @xmath2 is planck s constant , @xmath3 is the electrons wave vector measured from @xmath0 , and the pauli matrices @xmath4 are associated with the sublattice ( or , equivalently , the layer - index ) pseudospin degree of freedom @xcite . in our notation , @xmath5 is the @xmath6 unit matrix , @xmath7 , and @xmath8 . numerical values for the effective masses @xmath9 and the speed @xmath10 are well known @xcite . very close to the @xmath0 point , the energy dispersion resulting from ( [ eq : effm_blg ] ) mimics that of massless dirac electrons , as is the case in single - layer graphene ( slg ) . however , as @xmath11 , the dominant behavior of electrons in blg is captured by the quadratic dispersion shown in fig . [ fig : lattice](c ) . basic structural and electronic properties of bilayer graphene . ( a ) honeycomb structure of a bilayer stack of graphene . atoms in sublattice @xmath12 ( @xmath13 ) are marked in grey ( red ) . a @xmath14 plane is marked in light grey . ( b ) brillouin zone and its two inequivalent corner points @xmath0 and @xmath15 . the remaining corners are related with @xmath0 or @xmath15 by reciprocal lattice vectors . ( c ) dispersion @xmath16 near the @xmath0 point . we have @xmath17 . ] rrs1emcs1emll & & + orbital zeeman splitting ( @xmath18 field ) & b_z_z_z & ( 1 ) & _ z_z_0_z_z_0@xmath19 & -layer ( pseudo - spin ) gap - layer ( pseudo - spin ) gap@xmath19 + magnetic spin splitting ( @xmath18 field ) & b_z s_z_0 & ( 2 ) & _ z s_z _ z_z s_z _ z@xmath19 & spin splitting ( @xmath18 field)spin splitting ( @xmath18 field)@xmath19 + magnetic spin splitting ( @xmath20 field ) & ( b_x s_x + b_y s_y)_0 & ( 3 ) & ( _ x s_x + _ y s_y)_z(_x s_x + _ y s_y)_z@xmath19 & spin splitting ( @xmath20 field)spin splitting ( @xmath20 field)@xmath19 + -orbital zeeman ( @xmath18 field)-orbital zeeman ( @xmath18 field)@xmath19 & _ z ( s_y _ x_0 - s_x _ y_z)_z ( s_y _ x_0 - s_x _ y_z)@xmath19 & ( 4 ) & _ z ( s_y _ x_z - s_x _ y_0 ) & rashba spin splitting ( @xmath18 field ) + -orbital zeeman ( @xmath20 field)-orbital zeeman ( @xmath20 field)@xmath19 & _ z ( b_y _ x_0 - b_x _ y_z)_z ( b_y _ x_0 - b_x _ y_z)@xmath19 & ( 5 ) & s_z ( _ y _ x_z - _ x _ y_0 ) & rashba spin splitting ( @xmath20 field ) + zeeman splitting ( @xmath20 field)zeeman splitting ( @xmath20 field)@xmath19 & ( k_x b_y - k_y b_x ) _ z _ 0(k_x b_y - k_y b_x ) _ z _ 0@xmath19 & ( 6 ) & ( k_x_y - k_y _ x ) _ z_z & orbital rashba splitting + trigonal zeeman splitting & @r@ ( s_y b_y - s_x b_x ) _ x_0 + + ( s_x b_y + s_y b_x ) _ y _ z & ( 7 ) & @l@ ( s_y_y - s_x _ x ) _ x_z + + ( s_x _ y + s_y _ x)_y_0 @l@ ( s_y_y - s_x _ x ) _ x_z + + ( s_x _ y + s_y _ x)_y_0 @xmath19 & rashba splittingrashba splitting@xmath19 + external fields can have striking effects on the electronic properties of charge carriers in solids . previously , only the effects of electric fields @xmath21 and magnetic fields @xmath22 directed perpendicular to the blg sheet have been considered , giving the extended effective hamiltonian @xmath23{l } \mathcal{h}_{{\ensuremath{\bm{\mathrm{k } } } } } ( { { { \ensuremath{\bm{\mathrm{k } } } } } } , { \mathcal{e}}_z , b_z ) = \mathcal{h}_{{\ensuremath{\bm{\mathrm{k } } } } } ( { { { \ensuremath{\bm{\mathrm{k}}}}}}+e{{\ensuremath{\bm{\mathrm{a } } } } } ) \\[0.5ex ] \hspace{3em } { \displaystyle } { } + \frac{\lambda_z}{2}\ , { \mathcal{e}}_z\ , \sigma_z - \frac{g}{2}\,\mu_\mathrm{b}\,b_z\,\sigma_z + c_z \,{\mathcal{e}}_z b_z\ , \sigma_0 \ , . \hspace{1em } \end{array}\ ] ] accordingly , ( i ) a potential difference between the two layers ( equivalent to finite @xmath21 ) opens up a pseudospin gap @xcite @xmath24 , ( ii ) @xmath22 induces a pseudospin zeeman splitting @xcite @xmath25 , and ( iii ) the simultaneous presence of fields @xmath21 and @xmath22 leads to an ( actually , valley - contrasting see below ) overall energy shift @xcite @xmath26 . in eq . ( [ eq : fieldham ] ) , @xmath27 is the gyromagnetic ratio , @xmath28 is the bohr magneton , and @xmath29 is the electromagnetic vector potential satisfying @xmath30 . the matter - field interactions ( i)(iii ) generate sizeable effects for typical values of @xmath21 and @xmath22 ( see appendix [ app : c ] ) . in addition , the chiral nature of electrons in blg leads to unconventional quantum - hall physics @xcite . inspection of eq . ( [ eq : fieldham ] ) reveals a surprising feature : disregarding constant prefactors , the electron s interaction with fields @xmath21 and @xmath22 is symmetric with respect to the interchange of @xmath21 and @xmath22 . indeed , this observation is not an accident . it reflects the unusual property of blg that its crystal symmetry does not distinguish between polar vectors such as the electric field @xmath31 and axial vectors such as the magnetic field @xmath32 . moreover , the familiar constraints due to time - reversal invariance are modified in blg such that symmetry under time reversal likewise permits that @xmath31 and @xmath32 become interchangeable . using symmetry@xcite we have obtained the complete set of interactions of electrons near the @xmath0 point in blg with magnetic and electric fields @xmath32 and @xmath31 , strain @xmath33 , and also spin @xmath34 . see appendix [ app : a ] for details on our method and full results . all interactions that exist in slg are also present in blg . in addition , there are new interactions in blg that can be obtained from field - dependent slg interactions by replacing the magnetic ( electric ) field components with their electric ( magnetic ) counterparts . table [ tab : newterms ] summarizes this magneto - electric equivalence for couplings that are linear in electric or magnetic field components . the interaction @xmath35 coupling the pseudospin to a perpendicular electric field thus arises by magneto - electric equivalence from the pseudospin zeeman term @xmath36 that exists in slg . moreover , we obtain a rather counter - intuitive purely electric - field - dependent spin splitting @xmath37 , which can be directly measured using electron spin resonance . also interesting is the spin - orbital zeeman splitting @xmath38 , which is the magneto - electric equivalent of rashba spin splitting in slg @xcite . we note that the electrons interact not only with the fields @xmath32 and @xmath31 , but their quantum dynamics depends also via the usual minimal coupling @xcite on the electromagnetic potentials @xmath29 and @xmath39 . the latter contributions are not affected by the magneto - electric equivalence . the interactions shown in table [ tab : newterms ] give rise to a diverse set of textures for the induced spin and pseudospin orientations in @xmath40 space , which are illustrated in fig . [ fig : visual ] . the grey arrow in the lower left of each panel indicates the direction of the external field while a large arrow in the lower right indicates ( if present ) the spin polarization induced by the particular interaction when averaging over all occupied states in the @xmath41 plane . cc & + [ 0pt ] [ cols="^ " , ] & in addition to the layer - index pseudospin @xmath42 , electrons in blg carry a valley - isospin @xmath43 that distinguishes states near the two inequivalent @xmath0 and @xmath44 points in the brillouin zone @xcite [ fig . [ fig : lattice](b ) ] . the effective hamiltonian for electrons in the @xmath15 valley can be obtained from that for the @xmath0 valley by a reflection of the vectors @xmath40 , @xmath34 , @xmath31 , and @xmath32 at the @xmath14 plane , see fig . [ fig : lattice](a ) @xcite . choosing the convention that the @xmath32 and @xmath31-dependent terms have the same sign in the @xmath0 valley ( first column of fig . [ fig : visual ] ) , the corresponding term in the @xmath15 valley involving the axial vector @xmath32 ( second column ) differs by an overall minus sign from the term involving the polar vector @xmath31 ( third column of fig . [ fig : visual ] , see also table [ tab : newterms ] ) . for example , the orbital zeeman term @xmath45 couples @xmath22 to the pseudospin associated with the sublattice degree of freedom @xmath46 . however , the field @xmath22 does not induce a global pseudo - spin ( sublattice ) polarization , because this term has opposite signs in the two valleys so that the pseudospin polarization in the two valleys is antiparallel . as a result , the two sublattices remain indistinguishable even for finite @xmath22 . only an electric field @xmath21 polarizes the sublattices in blg via the term @xmath47 , consistent with the fact that the pseudospin @xmath46 is even under time reversal @xcite . this behaviour is opposite to that of real spin , where @xmath48 induces a real - spin polarization @xmath49 ( when averaging over the occupied states in both valleys ) while the term @xmath50 does not , consistent with time reversal symmetry . similarly , an in - plane magnetic field @xmath51 gives rise to a macroscopic in - plane polarization of real spins , whereas the real - spin polarization induced by an electric field @xmath52 is anti - parallel in the two valleys . the fields @xmath51 and @xmath52 also couple to the in - plane pseudospin @xmath53 . term ( 6 ) of table [ tab : newterms ] induces an out - of - plane tilt of the spin orientation of individual states which , in the @xmath15 valley , has opposite signs for @xmath51 and @xmath52 . remarkably , on average this yields an in - plane polarization @xmath54 which is nonetheless the same in each valley for fields @xmath51 and @xmath52 ( fig . [ fig : visual ] ) . this result reflects the fact that the macroscopic pseudospin polarization @xmath54 is neither even nor odd under time reversal @xcite . more precisely , in each valley the direction of @xmath54 is well - defined only up to a gauge - dependent angular offset . yet the _ change _ of @xmath54 induced by a change in the in - plane orientation of the applied field _ is _ well - defined and it points clockwise in one valley and counterclockwise in the other valley ( for both @xmath51 and @xmath52 and all terms in tab . [ tab : newterms ] giving rise to an in - plane pseudospin orientation of individual states ) . specifically for the term ( 6 ) , if we change the in - plane orientation of the fields @xmath51 or @xmath52 by an angle @xmath55 , this changes the resulting average polarization @xmath54 by @xmath56 . this implies , in particular , that reverting the direction of the external field yields the same orientation of the induced pseudospin polarization . we see here that the pseudospin polarization induced by external fields @xmath31 and @xmath32 behaves qualitatively different from the polarization of real spins . numeric values for the prefactors of the field - dependent interactions in fig . [ fig : visual ] can not be determined with the methods used here , but the important qualitative trends do not depend on these values . therefore , in the calculations yielding fig . [ fig : visual ] , the prefactors were chosen large enough to provide a visually clear picture . up to now , we have considered implications of the magneto - electric equivalence for interactions that are linear in either @xmath32 or @xmath31 . in every material , we also have interactions proportional to the squared field components . in slg , these are interactions proportional to @xmath57 , @xmath58 , @xmath59 and @xmath60 . the magneto - electric equivalence implies that we may replace one factor in the square of one field by the corresponding components of the other field . we thus get scalar magneto - electric couplings @xmath61 as well as their trigonally anisotropic counterpart @xmath62 \ , .\ ] ] such scalars enter not only the hamiltonian describing the dynamics of electrons in graphene , but they also enter the lagrangian @xcite from which we obtain maxwell s equations in graphene . the bi - linear coupling of electric and magnetic fields displayed in eq . ( [ eq : me_scalar ] ) is reminiscent of the @xmath63 contribution to the lagrangian of axions @xcite , hypothetical particles that were introduced to solve the strong cp problem in particle physics . here the bi - linear couplings also have a valley dependence . assuming independent dynamics for the electrons in the two valleys , the new term in the lagrangian becomes @xmath64 where @xmath65 and @xmath66 . the presence of @xmath67 modifies gauss law in blg by inducing valley - dependent charge densities @xmath68\ ] ] proportional to @xmath51 , and it modifies ampre s law by inducing current densities @xmath69\ ] ] proportional to @xmath21 ( see appendix [ app : b ] for details ) , thus realizing a valley - contrasting axion electrodynamics in blg , where the usual coupling @xcite of charges to electric fields ( gauss law ) and currents to magnetic fields ( ampre s law ) is completely reversed . the finiteness of any real sample leads to a spatial variation of the coefficients @xmath70 at its edge . as a result , equilibrium valley - isospin _ densities _ are induced at the blg sheet s boundaries in the @xmath71 plane when the magnetic field has an in - plane component directed perpendicularly to such a boundary ; and a perpendicular electric field generates equilibrium valley - isospin _ currents _ flowing parallel to the system s boundaries in the @xmath71 plane . figure [ fig : axion ] illustrates these effects in panels ( a ) and ( b ) . ramifications of the valley - contrasting axion electrodynamics . ( a ) an in - plane magnetic field @xmath51 induces equilibrium valley - isospin densities @xmath72 at sample boundaries that are not aligned with @xmath51 . ( b ) a perpendicular electric field @xmath21 gives rise to equilibrium valley - isospin edge currents @xmath73 . ( c ) valley - isospin densities can be created inside the sample by engineering inhomogeneity in the in - plane magnetic field . ( d ) valley - isospin currents can be created inside the sample by engineering inhomogeneity in the perpendicular electric field . ] the valley - asymmetric character of the edge densities and currents in blg distinguishes them from similar edge effects in topological materials @xcite that generally require the presence of strong spin - orbit coupling . also in contrast to the usually considered @xcite cases where @xmath74 , valley - isospin densities ( currents ) can be induced in the _ bulk _ of blg by spatially inhomogeneous fields @xmath51 ( @xmath21 ) . scenarios for this are illustrated in panels ( c ) and ( d ) of fig . [ fig : axion ] , which can be realized , e.g. , by placing nanomagnets on the surface of a blg sample [ for ( c ) ] or arranging pairs of front and back gates with opposite voltage polarity around adjacent parts of the sample @xcite [ for ( d ) ] . more generally , the valley - contrasting axion electrodynamics discovered here enlarges the scope of valley - dependent electronic effects in graphene materials @xcite by providing a comprehensive framework to design magneto - electric effects at boundaries and interfaces in blg , including associated valley - helical edge states @xcite . one of the most exciting aspects of magneto - electric equivalence in blg is the wide range of novel matter - field interactions resulting in this material . to achieve any type of effect within a valley , either an electric or a magnetic field component can be utilized . this includes , in particular , the different real - spin and pseudo - spin couplings of both zeeman and spin - orbit character . ingenious combinations of nano - magnetic and nano - electronic fabrication capabilities can thus be used to realize new device architectures that are impossible to achieve in other materials . additional variability is provided by the valley - isospin degree of freedom , allowing the selective creation of global or valley - contrasting real - spin or pseudo - spin polarizations and their exploitation for both fundamental research and new nano - spintronic applications . finally , it is fascinating that the phenomena discussed here are present not only in bilayer graphene , but they exist in any material with similar symmetries . in particular , analogous behaviour can be realized in appropriately designed metamaterials @xcite or cold atoms in optical lattices @xcite . the intrinsic tuneability of system parameters in these latter realizations opens intriguing avenues to tailor the relative magnitude of these matter - field interactions beyond the regime accessible in blg . the authors thank m. meyer and h. schulthei for help with generating fig . [ fig : visual ] . useful discussions with j. j. heremans , a. h. macdonald , j. l. maes , and m. morgenstern are also gratefully acknowledged . this work was supported by marsden fund contract no . vuw0719 , administered by the royal society of new zealand , and at argonne national laboratory by the doe bes via contract no . de - ac02 - 06-ch11357 . the hamiltonian ( [ eq : effm_blg ] ) describes the band structure of graphene as a function of the wave vector @xmath40 measured from the @xmath0 point . the theory of invariants @xcite makes it possible to also include perturbations @xmath75 that are combinations of various quantities in addition to the wave vector @xmath40 , e.g. , electric and magnetic fields @xmath31 and @xmath32 , strain @xmath33 and the intrinsic spin @xmath34 . the electron states at the @xmath0 point transform according to the two - dimensional irreducible representation ( ir ) @xmath76 of the group @xmath77 ( ref . ) . the effective @xmath78 hamiltonian for the @xmath76 subspace can be expressed as @xmath79 here @xmath80 are prefactors , @xmath81 are @xmath6 matrices that transform according to the irs @xmath82 ( of dimension @xmath83 ) contained in the product representation @xmath84 of @xmath77 . likewise , @xmath75 can be decomposed into irreducible tensor operators @xmath85 that transform according to the irs @xmath86 of @xmath77 . using the coordinate system in fig . [ fig : lattice ] we obtain the basis matrices and tensor operators listed in tables [ tab : basemat ] and [ tab : tensorop ] . @xmath87 0.2ex cs1.0emp0.9 _ 1 & ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; + _ 2 & ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; + _ 3 & ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; + additional constraints for the hamiltonian ( [ eq : invar ] ) are due to time reversal invariance . the point group @xmath88 of blg contains symmetry elements @xmath89 mapping the basis functions @xmath90 at @xmath0 on @xmath91 at @xmath15 . these basis functions are also mapped onto each other by the time - reversal operation @xmath92 , i.e. , we have @xmath93 with a unitary matrix @xmath94 . combining these operations , we obtain @xcite @xmath95 here @xmath96 denotes transposition and @xmath97 depends on the behavior of @xmath75 under time reversal . @xmath40 , @xmath32 , and @xmath34 are odd under time reversal so that then @xmath98 , while @xmath31 and @xmath33 have @xmath99 . equation ( [ eq : timea2 ] ) provides a general criterion for determining which terms in the expansion ( [ eq : invar ] ) are allowed by time - reversal invariance and which terms are forbidden . the matrix @xmath94 depends on the choice for the operation @xmath89 . if @xmath89 is the reflection @xmath100 at the @xmath14 plane [ thus mapping the atoms in each sublattice in each layer onto each other , see fig . [ fig : lattice](a ) ] , we obtain @xmath101 and the matrix @xmath94 is simply the identity matrix . we note that under @xmath100 polar ( @xmath102 ) and axial ( @xmath103 ) vectors transform as [ eq : transval ] @xmath104 the transformational properties for the components of the second - rank strain tensor @xmath105 can be expressed similarly @xcite . if we denote by @xmath106 ( @xmath107 ) the unity ( diagonal pauli ) matrix acting in valley - isospin space , eq . ( [ eq : transval ] ) can be summarized by writing general vector operators as @xmath108 and @xmath109 . the group @xmath77 characterizing the @xmath0 point in blg is a subgroup of the group @xmath110 for the @xmath0 point in slg @xcite so that any term allowed by spatial symmetries in @xmath111 for slg is likewise allowed in blg . moreover , the constraint ( [ eq : timea2 ] ) due to time reversal invariance is exactly equivalent to the constraint in slg @xcite . thus it follows immediately that the invariant expansion for blg contains all terms that exist already for slg . a more detailed analysis shows that the point group @xmath110 for slg distinguishes , as is usual , between polar vectors ( such as the electric field @xmath31 ) and axial vectors ( such as the magnetic field @xmath32 ) . thus each term in the slg hamiltonian with a certain functional form and linear in the field @xmath31 or @xmath32 is forbidden for the other field . however , the point group @xmath112 relevant for blg contains only rotations as symmetry elements so that it can not distinguish between polar and axial vectors . therefore , the @xmath113 and @xmath114 components of any vector transform according to the ir @xmath76 , whereas the @xmath115 component transforms according to @xmath116 . this implies that spatial symmetries can not distinguish electric and magnetic fields in blg . moreover , eq . ( [ eq : timea2 ] ) treats electric and magnetic fields symmetrically , too . thus it follows that every @xmath31-dependent term in the blg hamiltonian ( [ eq : invar ] ) is accompanied by another term where @xmath31 is simply replaced by @xmath32 ( and vice versa for @xmath32-dependent terms ) . however , the prefactors of these terms are , in general , unrelated @xcite . table [ tab : newterms ] summarizes some of the new terms arising from this magneto - electric equivalence . similar to slg , pseudospin up and down in blg corresponds to atoms in sublattice @xmath13 and @xmath117 , see fig . [ fig : lattice](a ) @xcite . with the convention that for both valleys @xmath118 the up and down eigenstates of pseudospin @xmath46 correspond to the same sublattice , the above procedure determines the hamiltonians @xmath119 up to phases @xmath120 corresponding to pseudospin rotations @xmath121 of the basis matrices @xmath122 ( @xmath123 ) entering the invariant expansion ( [ eq : invar ] ) ( ref . ) . due to this @xmath124 gauge freedom the direction of the average in - plane spin polarization @xmath125 induced within a valley by fields @xmath32 and @xmath31 is well - defined only up to a uniform field - independent angular offset @xmath120 . the presence of @xmath67 leads to modifications of the inhomogeneous maxwell s equations ( i.e. , gauss and ampre s laws ) amounting to the replacements @xcite [ eq : macroscopic ] @xmath126 where @xmath127 and @xmath128 denote the macroscopic electric and magnetic fields in matter @xcite . equations ( [ eq : macroscopic ] ) imply the existence of extra valley - dependent charge densities and currents that are most generally given by @xmath129 the coefficients @xmath130 are finite constants within blg but must vanish outside the sample . hence , there is a boundary contribution to @xmath131 , and we have @xmath132 .\ ] ] using the homogeneous maxwell equation @xmath133 , the magnetic - field - induced charge densities are then given by eq . ( [ eq : axion_rho ] ) . furthermore , the two - dimensionality of blg implies that currents flow only in - plane . hence , both @xmath134 and @xmath135 do not contribute to the current , which is just given by @xmath136 \ , .\ ] ] the other homogeneous maxwell equation can be written as @xmath137 whose in - plane projection inserted into eq . ( [ eq : j_intermed ] ) yields eq . ( [ eq : axion_j ] ) . for homogeneous , time - independent fields @xmath31 and @xmath32 the ordinary couplings of @xmath31 and @xmath32 to @xmath138 and @xmath139 vanish so that we only get the valley - contrasting axion electrodynamics of eqs . ( [ eq : axion_rho ] ) and ( [ eq : axion_j ] ) . the valley - contrasting axion electrodynamics discussed here can be expected to apply in situations when the valley - isospin degree of freedom is conserved . this will generally be the case for inhomogeneities induced by smoothly varying external fields ( such as the one proposed in ref . ) . in contrast , the physical termination of the blg sheet is abrupt on an atomic scale , and details will usually matter @xcite . nevertheless , the recently observed @xcite emergence of a universal behavior at disordered blg edges consistent with that obtained from our valley - contrasting axion - electrodynamics suggests a wider applicability of this formalism . recent experiments @xcite demonstrated that a displacement field of @xmath140 v / nm generates a bandgap of @xmath141 ev in blg , implying that @xmath142 @xmath143 nm ( with electron charge @xmath144 ) consistent with first - principles calculations @xcite . for the remaining prefactors in eq . ( [ eq : fieldham ] ) , the slonczewski - weiss - mcclure ( swm ) model @xcite applied to blg yields ( see also ref . ) @xmath145 the swm parameters @xmath146 , @xmath147 , and @xmath148 are defined in ref . , and we have @xmath149 , where @xmath150 is the electron mass in vacuum and @xmath151 is the lattice constant of blg s planar honeycomb structure .	it is a fundamental paradigm that the physical effects induced by electric fields are qualitatively different from those induced by magnetic fields . here we show that electrons at a dirac point in bilayer graphene experience an unusual type of electromagnetism where magnetic and electric fields are virtually equivalent : every coupling of an electron s degrees of freedom to a _ magnetic _ field is matched by an analogous coupling of the _ same _ degrees of freedom to an _ electric _ field . this counter - intuitive duality of matter - field interactions enables novel ways to create and manipulate spin and pseudo - spin polarizations in bilayer graphene via external fields and leads to the emergence of a valley - contrasting axion electrodynamics , where the traditional association of charges at rest with electric fields and charge currents with magnetic fields is reversed .
hochschild cohomology is a useful tool for studying deformation theory and it was studied extensively over the years ( for example see @xcite , @xcite , @xcite , @xcite , @xcite and @xcite ) . higher order hochschild ( co)homology was introduced by pirashvili in @xcite ( see also @xcite and @xcite ) . it associated to a simplicial set @xmath0 , a commutative @xmath1-algebra @xmath2 and an @xmath2-bimodule @xmath3 , the higher hochschild cohomology groups @xmath4 . when @xmath0 is the standard simplicial set associated to the sphere @xmath5 , one recovers the usual hochschild cohomology . one important feature of these cohomology groups is that they depend only on the homotopy type of the geometric realization of the simplicial set @xmath0 . for more recent results about higher hochschild cohomology see @xcite and @xcite . secondary hochschild cohomology was introduced in @xcite where it was used to study @xmath6-algebra structures on the algebra @xmath7 $ ] . it associates to a triple @xmath8 ( where @xmath9 gives the @xmath6-algebra structure on @xmath2 ) , and an @xmath2-bimodule @xmath3 that is @xmath6-symmetric , the secondary hochschild cohomology groups @xmath10 . the main difference from the usual hochschild cohomology is that the secondary hochschild cohomology gives information about both , the product and the @xmath6-algebra structure on @xmath7 $ ] . many of the results that are known for hochschild cohomology can be extended to secondary hochschild cohomology . one can study @xmath6-extensions of the algebra @xmath2 , show that @xmath11 is a multiplicative operad and admits a hodge type decomposition ( @xcite ) . there is also a cyclic version of the secondary cohomology that can be described using a generalization of the bar resolution ( @xcite ) . our main goal in this paper is to show that secondary hochschild cohomology is a certain version of higher order hochschild cohomology . more precisely , we consider a simplicial pair @xmath12 ( where @xmath13 is a simplicial set and @xmath0 is simplicial subset of @xmath13 ) , a triple @xmath8 ( where @xmath2 and @xmath6 are commutative @xmath1-algebras , and @xmath14 is a morphism of @xmath1-algebras ) and @xmath3 a symmetric @xmath2-bimodule . to this setting we associate the groups @xmath15 . when @xmath16 we recover the higher order hochschild cohomology @xmath17 . when @xmath18 with the natural simplicial structure , we recover the secondary hochschild cohomology @xmath19 . in this paper we fix a field @xmath1 and denote @xmath20 by @xmath21 . we assume that the reader is familiar with hochschild cohomology , but provide some details for the discussion of higher order and secondary hochschild cohomology . we also assume familiarity with simplicial sets . we follow the description in @xcite ( see also @xcite ) . assume that @xmath2 is a commutative @xmath1-algebra and @xmath3 is a symmetric @xmath2-bimodule . let @xmath22 be a finite pointed set such that @xmath23 ( we identify it with @xmath24 with @xmath25 the fixed element ) and define @xmath26 . for @xmath27 we define @xmath28 determined as follows : if @xmath29 then @xmath30 where @xmath31 take @xmath0 to be a pointed simplicial set . suppose that @xmath32 , we identify the set @xmath33 with @xmath34 then define @xmath35 for each @xmath36 we define @xmath37 and take @xmath38 defined as @xmath39 . the homology of this complex is denoted by @xmath4 and is called the higher order hochschild cohomology group . one interesting fact is that these groups depend only on the homotopy type of the geometric realization of the simplicial set @xmath0 . we recall the construction from @xcite . let @xmath2 be a @xmath1-algebra , @xmath6 a commutative @xmath1-algebra , @xmath14 a morphism of @xmath1-algebras such that @xmath40 and @xmath3 an @xmath2-bimodule that is @xmath6-symmetric . we define @xmath41 and @xmath42 such that for @xmath43 , we have : @xmath44 @xmath45 where @xmath46 and @xmath47 . it was proved in @xcite that @xmath48 is a complex . the homology of this complex is called the secondary hochschild cohomology and is denoted by @xmath49 . the homology and the cyclic version of this theory were discussed in @xcite . in this section we introduce a new cohomology associated to a simplicial pair @xmath12 and a triple @xmath8 ( where @xmath2 and @xmath6 are commutative @xmath1-algebra and @xmath50 is a morphism of @xmath1-algebras ) . we start with a few notations . we consider @xmath51 to be the category whose objects are pairs @xmath52 , where @xmath22 is a finite pointed set with base point @xmath53 , and @xmath54 is a pointed subset of @xmath22 . a morphism @xmath55 is a map of pointed sets @xmath56 such that @xmath57 . the category of finite pointed sets @xmath58 can be see as a full subcategory of @xmath51 in two different ways . first we can take the inclusion given by @xmath59 , second we can take the inclusion @xmath60 . a @xmath51-module is a functor from @xmath61 to @xmath1-modules . let @xmath2 and @xmath6 be two commutative @xmath1-algebras , @xmath14 a morphism of @xmath1-algebras and @xmath3 a symmetric @xmath2-bimodule . we construct @xmath62 to be the @xmath51-module determined as follows . for @xmath63 such that @xmath64 and @xmath65 define @xmath66 if @xmath67 is a morphism in @xmath68 , we define @xmath69 for @xmath70 then @xmath71 where for @xmath72 we have @xmath73 and for @xmath74 we have @xmath75 with the convention that if the product is taken over the empty set then we put @xmath76 and @xmath77 . we say that a pair @xmath78 is a simplicial pair if @xmath79 is a simplicial set and @xmath80 a simplicial subset of @xmath79 . in other words we have a functor @xmath81 for a simplicial pair @xmath78 we define the higher order hochschild cohomology associated to the triple @xmath8 and a symmetric @xmath2-bimodule @xmath3 , to be the homology of the complex defined as follows . for every @xmath82 we consider @xmath83 and take @xmath84 . we construct a complex by taking the differential induced by the simplicial structure on @xmath85 . more precisely if @xmath86 then we define @xmath87 and take @xmath88 , @xmath89 the homology of the above complex is called the higher order hochschild cohomology associated to the simplicial pair @xmath78 , of the triple @xmath8 with coefficients in @xmath3 and is denoted by @xmath90 . in the event that @xmath91 this definition agrees with the definition of higher order hochschild cohomology @xmath92 . in this section we show that when @xmath2 is commutative and @xmath3 is a symmetric @xmath2-bimodule , then the secondary hochschild cohomology @xmath93 is a particular case of the construction from the previous section . consider the simplicial pair @xmath94 , where the sphere @xmath5 is a obtained from the interval @xmath95 $ ] by identifying the ends of the interval , and the disk @xmath96 is obtained from the @xmath97-simplex @xmath98 $ ] by collapsing the edges @xmath99 $ ] and @xmath100 $ ] ( i.e. the boundary of @xmath96 is the edge @xmath101 $ ] ) . more precisely , we take @xmath80 to be the simplicial set where the only nondegenerate @xmath102-simplex is @xmath103 $ ] . we denote by @xmath104 the base point in dimension @xmath105 , and by @xmath106 the simplex in dimension @xmath107 , where we iterate the @xmath108 $ ] vertex @xmath109 times , and the @xmath110 $ ] vertex @xmath111 times . for example , @xmath112 is the interval @xmath113 with @xmath114 , and @xmath115 is a @xmath97-simplex @xmath116 $ ] such that @xmath117 and @xmath118 . for @xmath79 , besides the above simplices , we also have a nondegenerate @xmath97-simplex @xmath98 $ ] . denote it by @xmath119 and take @xmath120 and @xmath121 . more generally , take @xmath122 the @xmath123-dimensional simplex obtained by iterating the @xmath108 $ ] vertex @xmath109 times , the @xmath124 $ ] vertex @xmath111 times , and the @xmath110 $ ] vertex @xmath125 times . for example @xmath126 is a @xmath127-simplex @xmath128 $ ] with @xmath129 , @xmath130 , and @xmath131 . in general we have @xmath132 and @xmath133 . the @xmath134 are defined as follows : @xmath135 @xmath136 @xmath137 notice that @xmath106 is degenerate if @xmath138 , and @xmath122 is degenerate if @xmath139 . also we have that @xmath140 and @xmath141 . in particular we get that @xmath142 next we need to make the identification with the notation from @xcite . first recall that an element in @xmath143 was represented by a tensor matrix @xmath144 where @xmath145 and @xmath47 . for @xmath146 with @xmath147 the element @xmath148 corresponds to the position @xmath149 in the tensor matrix . for @xmath150 with @xmath151 the element @xmath152 corresponds to the position @xmath153 in the tensor matrix . we also add the symbol @xmath154 to correspond to @xmath155 . with the above identifications the formulas ( [ f2 ] ) and ( [ f3 ] ) become : @xmath156 @xmath157 if we use the above identification the formula for @xmath158 from equation ( [ eq2 ] ) is the same as the formula for differential @xmath159 from equation ( [ delta ] ) . to summarize we have the following result . let @xmath2 and @xmath6 be commutative @xmath1-algebras , @xmath14 a morphism of @xmath1-algebras and @xmath3 a symmetric @xmath2-bimodule , then we have @xmath160 one can see that @xmath90 is functorial with respect to all of its entries . more precisely , let @xmath161 and @xmath162 be two triples , @xmath3 a symmetric @xmath163-bimodule , and @xmath164 a morphism of @xmath1-algebras such that @xmath165 and @xmath166 , then we have the natural morphism @xmath167 where the @xmath168-bimodule structure on @xmath3 is induced by @xmath169 . also , if @xmath170 is a morphism of symmetric @xmath2-bimodules then @xmath171 moreover if @xmath172 is a morphism of simplicial sets then @xmath173 if we take the natural inclusion of simplicial pairs @xmath174 with the simplicial structure discussed in the previous section , then @xmath175 is nothing else but the morphism @xmath176 discussed in @xcite . one natural question is whether the construction in this paper depends only of the homotopy type of the geometric realization of the simplicial pair @xmath177 ( or maybe invariant under a certain equivalence relation among simplicial pairs ) . we explored this problem but we were not able to prove any interesting result . the main issue is finding an equivalence relation among simplicial pairs that is manageable at the algebraic level . the authors would like to thank andrew salch for conversations and suggestions about this research . bruce would also like to thank his wife kendall for her continued support .	in this note we give a generalization for the higher order hochschild cohomology and show that the secondary hochschild cohomology is a particular case of this new construction .
the agegraphic de model is initiated from quantum mechanics through the uncertainty relation along with the gravitational consequences in general relativity . this model takes the spacetime and fluctuations of matter content for the observed de in the universe . the original agegraphic de model was proposed by cai @xcite to discuss the expansion of the universe with acceleration . it contains the age @xmath5 of the universe in the expression of energy density as @xmath6 . the numerical factor @xmath7 is used to overcome some uncertaninties and @xmath8 represents the reduced plank mass . however , this model bears some drawbacks , for example it does not explain matter dominated era of the universe . cai @xcite proposed a new model by replacing age of the universe with the conformal time @xmath10 and called it new agegraphic de model . this model may naturally solve the coincidence problem @xcite . myung @xmath9 seo @xcite argued that nade and holographic de ( hde ) models have same energy densities , the only difference is that the former has conformal time while latter has future event horizon . however , there exists some essential differences between them @xcite . another approach to explore the accelerated expansion of the universe is modified theories of gravities . between the most studied models of modfied gravity , we mention @xmath11 gravity @xcite where @xmath12 is the ricci scalar curvature , @xmath13 gravity @xcite inherits torsion scalar @xmath14 , @xmath0 gravity takes @xmath1 as gauss - bonnet invariant @xcite , @xmath15 gravity and many more . these theories of modified gravity play a vital role , not in explaining the expanding universe but also many other issues @xcite in this respect . the @xmath0 gravity have a de - sitter point which is used for the cosmic expansion of the accelerated universe @xcite . davis @xcite has found that the model @xmath16 with @xmath17 , by keeping @xmath1 as responsible for de , is consistent with solar system tests . the reconstruction scenario or making correspondences between different de models became very appealing in cosmology recently . setare @xmath9 saridakis @xcite made correspondence between @xmath0 gravity models with hde model to discuss the cosmic acceleration of the universe . the nade model is also discussed in the framework of brans - dicke theory by liu @xmath9 zhang @xcite . setare @xcite studied the nade model in @xmath11 gravity and resulted that there may exists a phantom - like universe . jamil @xmath9 saridakis @xcite investigated the nade model in horava - lifshitz gravity , which resulted to be consistent with the observations for the accelerated expanding universe . the correspondence between @xmath0 gravity and entropy - corrected hde @xcite leads to the accelerating universe . jawad et al . @xcite investigated the hde model in the framework of @xmath0 gravity and discussed different phenomenon for the acceleration of the universe . in this paper , we reconstruct the nade model in @xmath0 gravity by assuming a power - law solution of the scale factor . we discuss the evolution of the universe with the help of equation of state ( eos ) parameter @xmath18 and use squared speed of sound @xmath3 to check the stability of the reconstructed model . the paper is arranged as follows . in section * 2 * , we review the main features of the @xmath0 gravity and nade model . we also construct the correspondence scenario . in section * 3 * , we provide the analysis and comparison of reconstructed model . section * 4 * is devoted to study the statefinder trajectories for the reconstructed model . the last section summarizes the obtained results . the action @xmath19 for @xmath0 gravity is given by : @xmath20.\ ] ] here @xmath12 is the ricci scalar curvature , @xmath21 is the coupling constant with @xmath22 representing the gravitational constant , @xmath23 is the determinant of metric tensor @xmath24 , @xmath25 is the matter lagrangian and @xmath0 is an arbitrary differentiable function of @xmath1 which is generally defined as : @xmath26 , where @xmath27 and @xmath28 are ricci curvature tensor and riemann curvature tensor respectively . by varying the action @xmath19 given in eq . ( [ 4 ] ) with respect to the metric tensor @xmath24 , the corresponding field equations are obtained as follow : @xmath29 where @xmath30 is the first derivative with respect to @xmath1 of the function @xmath31 and @xmath32 represents the energy - momentum tensor of the perfect fluid . we here consider a spatially flat ( i.e. with curvature parameter @xmath33 equal to zero ) and homogenous frw spacetime which line element is given by : @xmath34 with @xmath35 representing the scale factor which represents the expansion of the universe and @xmath36 the three spatial coordinates . taking @xmath37 and using eq . ( [ 6 ] ) in eq . ( [ 5 ] ) , the field equations become : @xmath38 the subscripts @xmath39 indicates the matter contribution of energy density and pressure and @xmath40 is the hubble parameter with overdot represening time derivative . the gauss - bonnet invariant @xmath1 for frw metric becomes @xmath41 . here we will incorporate with the pressureless matter , i.e. , @xmath42 while the energy density of matter@xmath43 satisfies the following energy conservation equation : @xmath44 this equation has a solution which is given by the following expression : @xmath45 where @xmath46 is an arbitrary constant . the field equations given in eqs . ( [ 7 ] ) and ( [ 8 ] ) can be rewritten in effective manner as : @xmath47 where : @xmath48,\\\label{12 } p_{eff}&=&\frac{1}{2}\left\{f-24h^2\left ( h^2+\dot{h}\right)f_g+(24)8 h^{2}\left [ 6\dot{h}^{3 } + 6h\dot{h}\ddot{h}+24h^{2}\dot{h}^{2}\right.\right.\\\nonumber & + & \left.\left.6h^{3}\ddot{h}+8h^{4}\dot{h}+h^{2}\dddot{h } \right]f_{gg}+(24)^28h^4\left(2h^2 + h\ddot{h}+4h^2\dot{h}\right)^2f_{ggg}\right\}.\end{aligned}\ ] ] by replacing the age of the universe @xmath14 in the energy density of agegraphic de model with the conformal time @xmath49 yields the energy density of nade model which is given by : @xmath50 we take @xmath51 just for simplification of the following calculations . here we make correspondence between nade and @xmath0 model by equating their energy densities , i.e. @xmath52 . it yields : @xmath53 to find out an analytic solution of eq . ( [ 13 ] ) , we assume a power - law form of the scale factor as follow : @xmath54 where the constant @xmath55 represents the present day value of the scale factor , @xmath56 is a constant called the finite future singularity time and @xmath57 is a real constant . the hubble parameter @xmath58 , its first and second derivatives , gauss - bonnet invariant and conformal time take the following forms @xmath59 using eq.([14 ] ) in the reconstructed equation , we obtain @xmath60.\ ] ] eq . ( [ 1 ] ) is a second order linear differential equation whose solution is given by @xmath61,\end{aligned}\ ] ] where @xmath62 and @xmath63 are integration constants . this is the reconstructed na @xmath0 de model which contains a constant term , a linear term of its argument and two terms depending upon @xmath57 as well . we restrict here @xmath64 to find out well defined results . in the following section , we discuss the behavior of model given in eq . ( [ 2 ] ) along its argument , eos parameter analysis and stability of the model with respect to time . the first thing we want to do in this section is to plot the na @xmath0 model ( [ 2 ] ) with respect to @xmath1 to check its behavior . for this purpose we assume four different values of @xmath57 , i.e. @xmath65 and fix the other constants involved by taking the values as @xmath66 . figure * 1 * shows the behavior of the model versus @xmath1 . it represents the continuously decreasing manner of the function for @xmath67 . the function approaches to the negative region and becomes approximately constant for @xmath68 incorporating all chosen values of @xmath57 . the eos parameter @xmath69 ( using eqs . ( [ 11 ] ) and ( [ 12 ] ) with corresponding quantities ) is plotted against the cosmic time @xmath70 for same values of @xmath57 and constants as for figure * 1. initially , the behavior of @xmath2 indicates a decelerating phase of the universe as shown in figure 2 * as @xmath71 guarantees for the accelerated universe . approximately for @xmath72 , the graph enters into accelerated region referred as quintessence era , @xmath73 , for all values of @xmath57 considered . the universe remains in this region for @xmath74 and @xmath75 as shown by the eos parameter . the higher values of @xmath57 , i.e. @xmath76 and @xmath77 , indicate the crossing of phantom divide line @xmath78 at @xmath79 and @xmath80 respectively and after this range , the universe faces the phantom era @xmath81 . as time elapses , at @xmath82 and @xmath83 , the universe again crosses the phantom divide line and becomes convergent in quintessence region . thus , incorporating the first two values of @xmath57 , the universe behaves as accelerated expanding in quintessence region and for last two values , the model depicts as quintom model ( crossing of phantom divide line occurs ) . plot of @xmath84 versus @xmath70 . the curves are characterize as red is for @xmath74 , green is for @xmath85 , blue is for @xmath76 , black is for @xmath86 . ] plot of @xmath84 versus @xmath70 . the curves are characterize as red is for @xmath74 , green is for @xmath85 , blue is for @xmath76 , black is for @xmath86 . ] figure * 3 * is a test to analyze the stability of the model by squared speed sound which defined as the ratio of the time derivative of total pressure to the time derivative of total energy density , i.e. : @xmath87 the stability or instability of the model depends upon the sign of this quantity . its positive sign corresponds to a stable model while instability fits for negative values . the stability of many de models is discussed using squared speed of sound . setare @xcite studied the interacting hde model with the generalized chaplygin gas model and discussed that both models are instable . the agegraphic de model is always negative which shows the instability of the model explored by kim et al . jawad et al . @xcite checked that the reconstructed hde @xmath0 model always remain instable . chattopadhyay and pasqua @xcite reconstructed the hde model in @xmath13 gravity . they investigated that the squared speed of sound for reconstructed models bear negative values led to the classical instable models . using eqs.([11 ] ) and ( [ 12 ] ) for the na @xmath0 model , the @xmath84 versus time is plotted taking same values of @xmath57 and constants . the graph indicates the instability of the model for all values of @xmath57 as squared speed of sound remain negative during the its evolution . however , we may argue that for decreasing @xmath57 it gradually shifts towards zero @xmath84 and may cross it . various candidates for the de model have been proposed till date and we often face with the problem of discrimination between them . to get rid of this problem , sahni et al . @xcite introduced the statefinder @xmath88 diagnostic pair . this pair has the following form @xcite @xmath89 here @xmath90 is the deceleration parameter and @xmath40 . thus the diagnostic pair consists of the first , second and third derivatives of the scale factor and indicates geometrically the properties of de . it has been discussed for a number of models . sharif and jawad @xcite discussed the modified hde in kaluza - klein universe and justified the limit of @xmath4cdm model by the statefinder pair . panotopoulos @xcite discussed this pair for two de models . chakraborty et al . @xcite discriminated different de models in kaluza - klein universe . as demonstrated , the statefinder can successfully differentiate between a wide variety of de models including the chaplygin gas , the cosmological constant , braneworld models , quintessence and interacting de models @xcite . the @xmath91 trajectories . the red , green , blue and black lines correspond to @xmath92 and @xmath93 respectively . other parameters are same as the previous figures . ] in the present work , based on the reconstructed energy density described in the previous sections , we have created the @xmath91 trajectories . it is apparent from the @xmath91 trajectories created for different choices of parameters stated inside the caption of figure * 4 * that the trajectories are confined inside the first quadrant of the @xmath91 plane . moreover , it is observed that although the trajectories are tending towards the fixed point @xmath94 , they can not reach it . thus , it is inferred that the na @xmath0 model is not capable of reaching the @xmath4cdm phase of the universe . the search for best fit de model is very challenging issue in cosmology recently . in this respect , the reconstruction phenomenon of different de models gain a very eminent attention to discuss accelerated expansion of the universe . the correspondence of hde , nade , their entropy - corrected versions , ghost , scalar field models within modified gravities give consistent results . in this paper , we have taken the @xmath0 gravity framework assuming exact power - law solution with dust matter . we have considered nade model in this framework to explore the cosmic expansion of the universe . the nade @xmath0 model is reconstructed and we have used it for further prospective . the results are as summarized : * the model contains highly nonlinear terms dependent upon @xmath57 , thus we have drawn it for different @xmath57 . the graphical behavior of nade @xmath0 model has shown the decreasing behavior against @xmath1 . * to characterize the different eras of the universe , we have plot eos parameter versus time @xmath70 for different values of @xmath57 . for smaller values , this parameter shows that the universe stays in the quintessence region . the higher values of @xmath57 represents the crossing of phantom divide line twice and the universe lying in quintessence state finally . thus it is interesting to mention here that for @xmath76 and @xmath77 , the reconstructed na @xmath0 model behaves like quintom model . * the stability of the model is investigated by squared speed of sound . the nade @xmath0 model has preserved the instability during the evolution of the universe . * the statefinder trajectories are plotted for the reconstructed model . it is observed that the na @xmath0 model is not subject to reach the @xmath4cdm model in the underlying scenario . jawad et al . @xcite investigated the hde model in the framework of @xmath0 gravity to discuss the acceleration of the universe and checked the stability of the model with squared speed of sound . also , they worked out on energy condition for hde @xmath0 model . in the present paper , we found the nade @xmath0 model by making correspondence between nade model in the underlying gravity . the reconstructed hde @xmath0 model contains a linear term of @xmath1 and two nonlinear terms and shows the increasing behavior with respect to @xmath1 . it shows the increasing behavior graphically . the reconstructed nade @xmath0 model inherits the same type of solution but highly nonlinear terms dependent upon @xmath57 with a constant term . its graphical behavior represents the decreasing behavior . both of the models remain instable during the evolution of the universe . however the violation of strong energy condition indicates the accelerated expansion for hde @xmath0 model whereas the quintessence region is observed for acceleration in nade @xmath0 gravity with the help of eos parameter . the first author ( aj ) wishes to thank the higher education commission , islamabad , pakistan for its financial support through the _ indigenous ph.d . 5000 fellowship program batch - vii_. the second author ( sc ) wishes to acknowledge the financial support from department of science and technology , govt . of india under project grant no . sr / ftp / ps-167/2011 .	in this work , we consider the reconstruction scenario of new agegraphic dark energy ( nade ) model and @xmath0 theory of gravity with @xmath1 representing the gauss - bonnet invariant in the flat frw spacetime . in this context , we assume a solution of the scale factor in power - law form and study the correspondence scenario . a new agegraphic @xmath0 model is constructed and discussed graphically for the evolution of the universe . using this model , we investigate the different eras of the expanding universe and stability with the help of the equation of state ( eos ) parameter @xmath2 and squared speed of sound @xmath3 , respectively . it is mentioned here that the reconstructed model represents the quintessence era of the accelerated expansion of the universe with instability . moreover , the statefinder trajectories are studied and we find out that the model is not capable of reaching the @xmath4cdm phase of the universe .
the chamaeleon i dark cloud ( @xmath10 @xmath0 11@xmath11 , @xmath12 @xmath0 @xmath1377@xmath14 ) is an active stellar nursery with 150 or more known young stars @xcite . located at a distance of @xmath0 160 pc @xcite , it has a relatively small angular size ( * ? ? ? * @xmath0 3 deg@xmath15 ) . because of its proximity and moderate extension on the sky the chamaeleon i star - forming region is an attractive place to attempt to measure the initial mass function ( imf ) , particularly at low masses ( i.e. , for masses here we report results of a @xmath16 imaging survey of the cha i dark cloud , which covers an area of @xmath0 1 deg@xmath17 and is complete for @xmath2 @xmath3 15.0 , @xmath4 @xmath3 16.0 , and @xmath5 @xmath3 16.5 . our survey complements the @xmath18 observations of denis ( deep near infrared southern sky survey , eso , 1 m tel , @xcite ) described by @xcite . the denis survey covers a larger area , 1.5@xmath19 @xmath20 3@xmath19 , of the cloud ; but , with a sensitivity of @xmath21 @xmath3 13.5 , it is less sensitive to the lowest mass cloud members . @xcite proposed 54 new candidate young stars , based on their locations in the @xmath22/@xmath23 diagram . in addition our survey area comprises two areas , in the the northern region , recently observed by @xcite and @xcite . the data reported by these authors are @xmath0 2 magnitudes deeper than ours but cover modest areas on the cloud ( @xmath0 3@xmath24 @xmath20 3@xmath24 , @xcite ; @xmath0 6@xmath24 @xmath20 6@xmath24 , @xcite ) centered close to the high - velocity c@xmath25o bipolar outflow previously detected by @xcite . @xcite identified a new class i source of the cloud . @xcite proposed 9 new very low luminosity members of the cloud . our survey recovers 40 of these denis candidates ( lying on common areas to both surveys ) and the stars with @xmath2 @xmath3 15.5 detected by @xcite and @xcite . in addition the present observation yields @xmath0 100 new candidates with @xmath2 @xmath6 12 and near - infrared excess emission in the @xmath7,@xmath8 color - color diagram . the low luminosities of these candidates suggests they may be very low mass cloud members , with masses close to or below the hydrogen burning limit . we describe our observations , data reduction , and analysis in 2 . the astrometry and photometry were obtained for @xmath0 11,090 sources at @xmath16 detected in our survey region . in 3 we derive a reliable extinction curve for background stars and use this curve to propose @xmath0 100 new potential young stellar objects based on their location in the @xmath7/@xmath8 diagram in 4 . we conclude with a brief summary in 5 . we obtained @xmath16 imaging data for the chamaeleon i dark cloud and three relatively unreddened control fields on 1316 february 1995 , 811 march 1996 , 1619 april 1997 , and 12 november 1998 with cirim ( the cerro tololo infrared imager ) at the ctio 1.5 m telescope . the cirim uses a 256 @xmath20 256 hgcdte nicmos 3 array , which provides a field of @xmath0 4.9@xmath24 @xmath20 4.9@xmath24 with a plate scale of 1.16@xmath26 per pixel . we covered an area of on the cloud on a rectangular grid , with 1@xmath24 overlap between adjacent frames . we acquired two 6 @xmath20 5 sec exposures for each field , shifted by 20@xmath26 . figure 1 shows the extent of the survey relative to the h@xmath27co contour map measured by @xcite and the distribution of previously known young members of the cloud @xcite . cha i has a remarkable low - mass star formation activity localized mainly in three groups . these small clusters are situated near three reflection nebulae : ced 112 ( hd 97300 ) , ced 110 , and ced 111 ( hd 97048 ) , from north to south . our three control fields , covering an area of @xmath0 0.05 deg@xmath17 and selected from a visual inspection of the eso red sky survey prints , are close to the survey region and relatively free from significant optical extinction . these off - cloud regions lie outside the limits of figure 1 , @xmath0 1.52 deg away from the h@xmath27co boundaries , one to the north and the other two to the east and west , respectively . table 1 gives positions and total areas covered by each of the off - cloud regions . we processed the data using standard techniques with the software package iraf . from each program frame , we subtracted an average dark frame , divided by a normalized flat - field frame appropriate for each filter , and then subtracted a flattened sky frame . the dark frame is the average of @xmath0 40 individual dark images taken at the beginning and end of each night . we constructed flat - field frames for each night by median - filtering all 6 @xmath20 5 sec frames taken in each filter . we also obtained a set of `` dome flats '' in each filter before observing each night using an illuminated white spot in the ctio 1.5-m dome . flat - fields constructed from `` dome flats '' and `` sky flats '' were identical . sky frames were generated from a median - filtered set of 2030 flattened frames in each filter . in this case , we combined individual object frames obtained close in time and in position to the program frame . to produce a combined image from each dithered pair of images , we aligned the frames using the iraf subroutines geomap and geotran , added the co - aligned frames together , and trimmed the resulting image to remove bad pixels at the edges of the frames . we selected the @xmath2 images as the reference frames and transformed the @xmath5 and @xmath4 images to the same pixel scale as the @xmath2 images . we used daofind to locate stars 4@xmath28 above the local background and added to the daofind list all stellar objects missed by this routine , found by visual inspection of each image . we then derived photometry for each image using the apphot phot task , using a circular aperture with 5@xmath26 radius . this aperture size includes the total flux for the large majority of sources . however , in some cases , we used a smaller aperture ( i.e. 2.5@xmath26 ) to avoid contamination from very close stars . several iterations of this process produced an homogeneous set of data for the complete survey . we detected @xmath0 11,090 sources at @xmath16 in our survey region . to calibrate our photometry , we observed on each night a set of 10 - 15 standards from @xcite and from the ukirt faint @xmath16 standard stars list @xcite . we estimate an uncertainty of @xmath29 0.03 mag in our calibration . the standard stars were observed at a similar airmass range as our target fields . airmass corrections for our data are smaller than our photometric uncertainties . we used common stars on dithered pairs and measurements of duplicate stars in the overlapping regions of adjacent frames to estimate photometric uncertainties for the stars in our survey region . table [ tbl-2 ] lists typical photometric uncertainties in each magnitude bin . these differences reach 0.3 mag for @xmath5 @xmath0 17 , @xmath4 @xmath0 16 , and @xmath2 @xmath0 15 . the 5@xmath28 limiting magnitudes are @xmath2 @xmath30 14.5 , @xmath4 @xmath30 15.5 , and @xmath5 @xmath30 16.0 . to make a check on the completeness of the survey , we constructed a list of sources from the three off - cloud regions ( see table 1 ) . figure 2 shows the magnitude histograms of background stars as function of the @xmath16 magnitudes , the number of stars per bin increases monotonically up to the completeness limit and then turns over . the corresponding limits in each filter are : 16.5 at @xmath5 , 16.0 at @xmath4 , and 15.0 at @xmath2 . to compare our photometry with results for previous surveys , we observed @xmath0 50 stars projected on bok globule 2 in the coalsack @xcite . @xcite and @xcite previously reported photometry of these stars and derived a comparison between the mt . stromlo / aao photometry of @xcite and the cit photometric system of @xcite . the common photometric data between @xcite and our survey , together with the transformations listed in @xcite , allow us to derive a transformation between the cirim and cit photometric systems . using the @xcite routine fitexy , straight line fits to the common data yield : @xmath31 @xmath32 @xmath33 the quoted uncertainties in these relations are the 1@xmath28 errors from our fits to the mt . stromlo / aao data and those from the @xcite fits , added in quadrature . the differences between the cirim and cit system are negligible for @xmath2 and @xmath8 , in agreement with results quoted in the cirim manual written by r. elston and j. elias . we thus assume that the natural cirim system is identical to the cit system for @xmath2 and @xmath8 . our derived color term for @xmath7 agrees with elston & elias ; the cirim @xmath7 color is bluer than the cit @xmath7 due to its different lens design . we apply the color term to our @xmath7 data and thus quote colors in the cit system . to derive coordinates for our survey stars , we used wcstool @xcite , a suite of programs to calculate a direct transformation between the coordinates of the image ( x , y ) and the sky coordinates ( @xmath10,@xmath12 ) . we measured transformation coefficients adopting matches between program stars and stars in the u.s . naval observatory sa1.o catalogue . these matches provided a good transformation for @xmath0 90% of the frames . for the rest of the frames , usually corresponding to the most obscured regions , we found few or no matches between our frames and stars in the catalog . in these cases , we used stars in common with adjacent frames with good coordinates to obtain coordinates for stars on unmatched frames . when an individual source was detected in more than one filter we adopted the average coordinate . to verify the coordinates , we measured coordinate differences between stars on adjacent frames . many stars have two good coordinates ; others have as many as four . after visually inspecting the duplicate stars and rejecting stars too close to the edges of frames , we averaged coordinates of the remaining duplicates and derived 1@xmath28 errors . finally , we compared coordinates derived from the digitized sky survey ( dss ) plates . we estimate an average uncertainty in our positions of 1@xmath26 . these errors are as large as 2@xmath26 for sources in the most obscured regions of the cloud . having verified the quality of the coordinates and photometry for our survey , we search for pre - main sequence stars with near - infrared excess emission in the cloud using the near - infrared color - color diagram . to identify these stars , we derive a reliable extinction curve for background stars in 3 and then consider the near - infrared color - color diagram in 4 . to derive the near - infrared reddening law for the cha i dark cloud , we follow @xcite , who developed a generalized photometric technique to compare the colors of reddened stars with the colors of nearby , `` unreddened '' comparison stars . @xcite assume that the stellar population behind the cloud is identical to the stellar population in off - fields several degrees away . they derive the @xmath7 and @xmath8 color excesses for each on - cloud source relative to every off - cloud source and then compute the average and median color excesses for each on - cloud source . the probable error of the average color excess is the sum in quadrature of the errors of the on - cloud and off - cloud colors . for the median color excess , the probable error is the inter - quartile range . we divide the cha i sources into a complete sample containing all stars with @xmath34 and a restricted sample , where we remove known young stars in the cloud from the complete sample . the sample of known young stars includes 94 sources ( lying on our survey region ) from the literature @xcite . this sample probably is not complete ; however , any incompleteness does not affect our analysis significantly . figure 3 shows average color excesses of cha i stars for two k @xmath35 14 samples , the complete sample and the restricted sample . the colors have been transformed to the cit system using equations 13 . the color excesses are highly correlated , with a spearman rank - order correlation coefficient of @xmath36 @xmath30 0.65 for both samples . the probability for no correlation between the two color excesses is formally zero according to the spearman rank - order test . straight line fits to color excess measurements with @xmath37 yield @xmath38 @xmath30 1.63 @xmath29 0.03 for the complete sample and @xmath38 @xmath30 1.76 @xmath29 0.02 for the restricted sample . we derive identical slopes using median color excesses for each source instead of the average color excesses . our extinction results for the restricted sample are identical to the @xcite near - infrared extinction law , @xmath38 @xmath30 1.75 transformed to the cit photometric system . the cha i results differ from the near - infrared extinction law for @xmath39 oph , @xmath38 @xmath30 1.57 @xmath29 0.03 ( * ? ? ? * ; * ? ? ? * for example ) at more than the 3@xmath28 level , and are inconsistent with results from the @xcite survey of luminous southern stars , @xmath38 @xmath30 1.47 @xmath29 0.06 , at roughly the 3@xmath28 level . results derived for the complete sample demonstrate that pre - main sequence stars with near - infrared excesses skew the extinction law to smaller values ( see also @xcite and references therein ) . figure 3 shows many previously identified sources with apparent near - infrared excesses . these sources probably skew our results for the reddening law derived from the restricted sample , although this effect should be small . @xcite described a method , using the `` reddening probability distribution '' , to correct the reddening law for previously unidentified pre - main sequence stars and to derive a sample of candidate pre - main sequence stars with near - infrared excess emission . we now apply this method to the restricted sample of cha i sources . @xcite defined the reddening probability distribution , @xmath40 , as the probability of measuring a pair of color excesses , @xmath41 and @xmath42 , where @xmath43 and @xmath44 are color indices and @xmath45 is the number of reddening measurements . the number , @xmath46 , is also the number of on - field stars in the sample . the density function , @xmath40 , depends on the distributions of colors in the on - field and the off - field . if @xmath47 is the off - field color distribution and @xmath48 as the on - field color distribution , the reddening probability distribution is : @xmath49 where @xmath50 and @xmath51 . the integers @xmath52 and @xmath53 measure the zero point offsets of the color grids ; indices @xmath54 and @xmath55 span the full range of color indices @xmath43 and @xmath44 . the color distributions use a the kernel density estimator : @xmath56 where @xmath57 is the smoothing parameter and k is the kernel @xcite . we follow @xcite and adopt @xmath58 for fast computation ( * x * = ( @xmath59 ) ) . we adopt @xmath60 , which is roughly twice the 1@xmath28 error of our photometry at the @xmath61 mag survey limit . smaller values for @xmath57 produce noisy grids ; larger values are inconsistent with the photometric errors . figure 4 shows the color distribution functions for the off - field and the complete sample ( cs ) . the off - field density in the left panel has a weak maximum at the color expected for g - type giants and a sharp maximum at the color expected for k - type dwarfs . there is a plateau consistent with colors for m - type dwarfs and an extension of the dwarf locus towards the near - infrared colors of k- and m - type giant stars . the cha i color distribution in the right panel peaks close to the off - field maximum but is much more extended . the long axis of the contours roughly follows the reddening line derived above , and there is a weak extension of sources with near - infrared excess emission at @xmath8 @xmath30 0.75 and @xmath7 @xmath30 1.75 . figure 5 shows the reddening probability functions constructed from the color distributions for the complete sample ( cs ) and the restricted sample ( rs ) . we derive the slope of the reddening law which follows the major axis of each contour in these diagrams using the technique outlined in ( * ? ? ? * equations 1215 ) . this analysis yields @xmath38 @xmath30 1.77 @xmath29 0.03 for the complete sample and @xmath38 @xmath30 1.80 @xmath29 0.03 for the restricted sample . these results are consistent with the reddening law derived in figure 3 given the 1@xmath28 errors of each fit . figure 6 shows the @xmath8/@xmath7 diagram for our complete survey region . as in figures 35 , we display stars with , where the photometric errors are @xmath62 0.1 mag . the solid line corresponds to the main sequence locus @xcite ; the dashed lines , parallel to the reddening vector derived in 3 , define the reddening band extending from the main sequence . we distinguish two groups of sources in figure 6 : a ) stars that lie along or follow the reddening band direction , and b ) sources that show near - infrared excess , located to the right of the reddening band . the first group , comprising the majority of the sources , show negligible near - infrared color excess , and can be dereddened to lie close to the standard dwarf sequence . most of these sources are probably behind the cloud . the color excess of the second sample can not be attributed to the reddening by dusty material in the cloud . several authors have found that circumstellar material ( disks + infalling envelopes ) surrounding the young central star can produce these color excess at the near - infrared wavelengths @xcite . formally speaking we have detected sources lying to the right of the reddening band in our survey with roughly 1/3 are known members of the cloud and denis candidates from @xcite . the other 2/3 are new detections . in we have used starred symbols to indicate new candidates with significant near - infrared excess ( i.e. , @xmath8 @xmath9 0.5 ) . this group comprises @xmath0 50 objects . table 3 gives coordinates and magnitudes for these stars . the rest of the sources either lie very close to the reddening line or to the bulge defined by the background stars . finally , our survey has also produced a list of @xmath0 50 faint objects ( i.e. , @xmath2 @xmath63 14 16 ) , with large near - infrared excess ( i.e. , @xmath8 @xmath9 0.8 ) . figure 7 shows the location of these stars in the color - color diagram . they are located well to the right of the main sequence locus . in this manner even allowing typical photometric uncertainties for these magnitudes ( i.e. , @xmath0 0.3 mag for @xmath2 @xmath0 15 ) the sources still have color excesses . table 4 gives coordinates and magnitudes for these stars . for these faint sources , we checked each of the images visually to avoid possible confusion with cosmic ray events , bad pixels , plate artifacts , or contamination with close stars . a few of the objects listed in the table 3 and table 4 were only detected in two of the filters ( @xmath2 and @xmath4 ) and therefore @xmath5 must be @xmath9 18.0 . these are extremely red objects and thus good candidate members of the cloud . some of the objects display extended or fuzzy images principally at @xmath2 . four candidates were also detected by the isocam survey of the chamaeleon i dark cloud and proposed as new low - mass candidate members based on their position on the mid - infrared color - color diagram @xcite . we also have in common two stars with @xcite ( see table 3 and table 4 ) . sources with near - infrared excess emission are either young stars in the cloud , galactic sources such as planetary nebulae ( e.g. , @xcite ) , or galaxies . extragalactic source counts predict 3060 galaxies in an area similar in size to our survey region for @xmath64 14 ( e.g. , @xcite and the references therein ) . the near - infrared extinction towards cha i is significant @xcite ; we thus expect @xmath0 30% to 50% fewer galaxies in our survey compared to unobscured regions . galaxies also tend to have smaller near - infrared excesses compared to young stars . most galaxies have colors similar to red main sequence stars , @xmath7 @xmath65 0.60.8 and @xmath8 @xmath65 0.250.45 , with modest near - infrared excesses of 0.20.3 mag in h k ( e.g. , @xcite ) . unpublished 2mass data support this conclusion for nearby galaxies ( see @xcite ) . our survey is not deep enough to detect more distant galaxies which could have redder near - infrared colors than measured for the nearby galaxy samples . several types of galactic sources , such as carbon stars and planetary nebulae , can also lie outside the reddening band in the near - infrared color - color diagram . the surface densities of these objects towards cha i should be small , @xmath35 510 deg@xmath66 for carbon stars and @xmath35 12 deg@xmath66 for planetary nebulae ( e.g. , @xcite ) . the spatial distribution of the near - infrared excess sources provides the best evidence that these objects are pre - main sequence stars . figure 8 compares the spatial distribution of the new candidate objects with the positions of previously known young stars @xcite and the h@xmath27co contour maps @xcite . the new candidates tend to cluster around ced 112 ( hd 97300 ) , ced 110 , and ced 111 ( hd 97048 ) in a manner similar to the previously known members of the cloud ( see figure 1 ) . we expect extragalactic and other galactic sources such as carbon stars and planetary nebulae to avoid regions of high obscuration and see no evidence for this tendency in our candidates . assuming an average inter - cloud extinction of a@xmath67 @xmath0 5 ( * ? ? ? * ; * ? ? ? * i.e. , a@xmath68 @xmath0 0.5 ) , a main - sequence star at the h burning limit and at the distance of the cha i cloud ( * ? ? * pc ) , would have @xmath2 @xmath0 15.5 16.0 @xcite . low mass members of the chamaeleon i dark cloud must be much younger , still contracting toward the main sequence , and thus significantly more luminous than main sequence stars of 0.1 m@xmath69 . our imaging survey is complete to @xmath2 @xmath0 15.0 and sensitive down to 16.5 . the objects detected in this survey ( with @xmath2 @xmath0 12 16 , @xmath8 @xmath9 0.5 1.5 ) are among the lowest luminosity and , presumably , lowest mass members of the cloud . some of them are near or , probably , below the h burning limit . some of our candidates ( with @xmath2 @xmath0 12 14 ) have optical counterparts on the dss plates . we crudely estimate for this subgroup of candidates . moderate resolution optical spectra would allow a search for li i 6707 absorption , an indicator of youth , as well as other atomic ( such as h@xmath10 ) and many forbidden emission lines usually present in the spectra of young stellar objects ( * ? ? ? * for example ) . to investigate further the nature of the rest of our candidate objects we need additional 10 @xmath70 m photometry and/or near - infrared spectroscopic observations . in particular , near - infrared spectra would allow us to place the stars in the hr diagram and thus , adopting recently developed pre - main sequence tracks ( * ? ? ? * ; * ? ? ? * for example ) , determine masses and ages for the individual sources . mass determinations are needed to estimate the lower end of the imf . present determinations of the imf in the cha i cloud are limited to 0.3 m@xmath69 and for masses @xmath62 0.6 m@xmath69 these determinations are uncertain @xcite . age estimations will provide additional support to the concept of a coeval star - formation process in this cloud @xcite . we have used the large number of background stars detected in our survey region to derived a reliable near - infrared extinction law for the chamaeleon i cloud . our analysis yields @xmath38 @xmath30 1.80 @xmath29 0.03 for the restricted sample ( i.e. , eliminating previously known pre - main sequences stars ) . this result differs with the @xcite s determination for the @xmath39 oph cloud ( @xmath38 @xmath30 1.57 @xmath29 0.03 ) at more than 3@xmath28 level . based on preliminary results for other two cloud ( taurus and ic 348 ) , in addition to their determination for @xmath39 oph , @xcite suggest that a real variation in the near - infrared reddening law occurs from region to region . our results for cha i provide additional support for this variation . the ratio of total to selective extinction , r@xmath67 , seems also to change for different clouds ( see @xcite and the references therein ) . in addition , r@xmath67 varies across the cha i dark cloud ( from r@xmath67 @xmath0 3 to r@xmath67 @xmath0 5.5 ; @xcite , see also @xcite ) . for the @xmath39 oph , @xcite obtained a mean value of r@xmath67 @xmath0 4 , although individual determinations vary within a range similar to that found for the cha i dark cloud . @xcite recently found that the a@xmath67 @xmath71 relation also varies among different star - forming clouds and suggest that a@xmath67/@xmath71 may also change across the chamaeleon i dark cloud . these results indicate different dust - grain properties , in particular grain - sizes , for different nearby star - forming regions and caution against using the `` universal '' reddening law to derive extinction estimates for young stars in star - forming regions . previous near - infrared observations of the cloud focussed on iras - selected sources @xcite or already known and candidate young stellar objects detected at optical and x - ray wavelengths @xcite . these investigations were sensitive to objects with k @xmath0 11 and limited to small areas around the target objects . present imaging near - infrared surveys of the cloud have overcome these limitations ( both in sensitivity and spatial coverage ) and greatly increased the number of potential members of the cloud . @xcite detected one new class i source of the cloud with k @xmath0 13 . @xcite proposed 9 new candidates with k @xmath30 13 16 . @xcite and this paper carry out large scale near - infrared survey on the cloud . @xcite proposed 54 new candidates members with k @xmath62 13 . this paper presents an additional set of @xmath0 100 new stars with k @xmath0 12 16 . in addition to these candidates selected on basis of their near - infrared color excesses , @xcite have recently proposed a list of 74 new low mass candidates detected by isocam and selected on basis of their color excesses at the mid - infrared wavelengths . these objects are among the lowest luminosity and , presumably , lowest mass members of the cloud . some of them are near or , probably , below the h burning limit . a spectroscopic follow up of the optically visible candidates should provide additional indications of their pre - main sequence nature . optical and/or near - infrared spectral types of our proposed candidates , in combination with recently developed pre - main sequence evolutionary tracks , will allow us to determine masses and ages . reliable mass determinations are required to calculate the imf , particular towards the sub - solar and probably into the sub - stellar mass regime . ages determination will help to reconstruct the history of the star - formation process of the cloud . we are grateful to the ctio staff , specially to m. fernndez , m. hernndez , and p. ugarte for assistance during the observing runs . we also thank r. elston and r. probst for their help with cirim , d. mink for assistance with the wsctools software and the anonymous referee for suggestions that improved the manuscript . this research was partially supported by the scholarly studies program of the smithsonian institution and the national aeronautics and space administration ( grants : nagw-2919 and go-06132.01.94a ) . m.g . acknowledges support from the national science foundation through grants gf-1001 - 96 and gf-1001 - 97 from the association of universities for research in astronomy , inc . , under nsf cooperative agreement ast-8947990 . appenzeller , i. , jankovics , i. , and krautter , j. 1983 , , 53 , 291 assendorp , r. , wesselius , p. r. , whittet , d. c. b. , and prusti , t. 1990 , , 247 , 624 bessell , m. s. , and brett , j. m. 1988 , , 100 , 1134 boulanger , f. , bronfman , l. , dame , t. m. , and thaddeus , p. 1998 , , 332 , 273 cambrsy , l. 1999 , , 345 , 965 cambrsy , l. , copet , e. , epchtein , n. , de batz , b. , borsenberger , j. , fouqu , p. , kimeswenger , s. , and tiphne , d. 1998 , , 338 , 977 cambrsy , l. , epchtein , n. , copet , e. , de batz , b. , kimeswenger , s. , le bertre , t. , rouan , d. , and tiphne , d. 1997 , , 324 , l5 casali , m. m. , and hawarden , t. g. 1992 , in : jcmt - ukirt newsletter no . 3 , 33 comern , f. , rieke , g. h. , and neuhuser , r. 1999 , , 343 , 477 dantona , f. , and mazzitelli , i. 1997 , mem . ital . , 68 , 807 elias , j. 1978 , , 224 , 453 elias , j. , fogel , j. a. , matthews , k. , and neugebauer , g. 1982 , , 87 , 1029 elias , j. , fogel , j. a. , hyland , a. r. , and jones , t. j. 1983 , , 88 , 1027 epchtein , n. 1997 , in : garzn , f. , epchtein , n. , omount , a. , burton , b. , persi , p. ( eds . ) the impact of large scale near - infrared sky surveys , kluwer academic publishers , 15 frogel , j. a. 1985 , , 298 , 528 gauvin , l. s. , and strom , k. m. 1992 , , 385 , 217 guglielmo , f. , le bertre , t. , and epchtein , n. 1998 , , 334 , 609 hartigan , p. 1993 , , 105 , 1511 hayakawa , t. , mizunon , a. , onishi , t. , yonekura , y. , hara , a. , yamaguchi , r. , and fukui , y. 1999 , , 51 , 919 he , l. , whittet , d. c. b. , kilkenny , d. , and spencer jones , j. h. 1995 , , 101 , 335 impey , c. d. , wynn - willians , c. g. , and becklin , e. e. 1986 , , 309 , 572 jarrett , h. , chester , t. , huchra , j. , and schneider , s. , 1998 , , 30 , 900 henry , t. , and mccarthy , jr . d. w. 1993 , , 106 . 773 hyland , a. r. , jones , t. , and michell , r. m. 1982 , , 201 , 1095 jones , t. j. , hyland , a. r. , robinson , g. , smith , r. , and thomas , j. 1980 , , 242 , 132 kenyon , s. j. , and hartmann , l. 1995 , , 101 , 117 kenyon , s. j. , brown , d. i. , tout , c. a. , and berlind , p. 1998 , , 115 , 2359 kenyon , s. j. , lada , e. a. , and barsony , m. 1998 , , 115 , 252 lada , c. j. , and adams , f. c. 1992 , , 393 , 278 lawson , w. a. , feigelson , e. d. , and huenemoerder , d. p. 1996 , , 280 , 1071 mattila , k. , liljestrn , t. , and toriseva , m. 1989 , in : reipurth , b. ( ed . ) low mass star formation and pre - main sequence objects ( garching : eso ) , 153 mink , d. j. 1997 , in : hunt , g. , payne , h. e. ( eds . ) vi asp conf . series , 125 , astronomical data analysis software and systems , 249 oasa , y. , tamura , m. , and lawson , w. a. 1999 , , 526 , 336 ortiz , r. , and maciel , w. j. 1994 , 287 , 552 palla , f. , and sthaler , s. 1999 , , 525 , 772 persi , p. , marenzini , a. r. , kaas , a. a. , olofsson , g. , nordh , l. , and roth , m. 1999 , , 117,439 persi , p. , marenzini , a. r. , olofsson , g. , kaas , a. a. , nordh , l. , huldtgren , m. , abergel , a. , andr , p. , bontemps , s. , boulanger , f. , burgdorf , m. , casali , m. m. , cesarsky , c. j. , copet , e. , davies , j. , falgarone , e. , montmerle , t. , perault , m. , prusti , t. , puget , j. l. , and sibille , f. 2000 , , 357 , 1020 press , w. h. , flannery , b. p. , teukolsky , s. a. , vetterling , w.t . 1992 , numerical recipes in fortran ( 2d ed : cambridge : cambridge univ . press ) . prusti , t. , clark , f. o. , whittet , d. c. b. , laureijis , r. j. , and zhang , c. y. 1991 , , 251 , 303 schwartz , r. d. 1991 , in : reipurth , b. ( ed . ) eso scientific report no . 11 , low mass star formation in southern molecular clouds , 93 silverman , b. w. 1986 , density estimator for statistics and data analysis ( london : chapman & hall ) szokoly , g. p. , subbarao , m. u. , and connolly , a. j. 1998 , , 492 , 452 tapia , s. 1975 , in : greenberg , j. m. , and van de hulst , h. c. ( eds . ) iau symposium 52 , interstellar dust and related topics , 43 toriseva , m. , and mattila , k. 1985 , , 153 , 207 visnen , p. , tollestrup , e. v. , and cohen , m. 2000 , , 540 , 593 vrba , f. j. , coyne , g. v. , and tapia , s. 1993 , , 105 , 1010 whittet , d. c. b. , prusti , t. , franco , g. a. p. , gerakines , p. a. , kilkenny , d. , larson , k. , and wesselius , p. r. 1997 , , 327 , 1194 whittet , d. c. b. , prusti , t. , and wesselius , p.r . 1991 , , 249 , 319 cccc 8 9 & 0.02 & 0.01 & 0.01 + 9 10 & 0.03 & 0.02 & 0.02 + 10 11 & 0.03 & 0.02 & 0.02 + 11 12 & 0.03 & 0.03 & 0.03 + 12 13 & 0.04 & 0.03 & 0.03 + 13 14 & 0.07 & 0.04 & 0.03 + 14 15 & 0.12 & 0.08 & 0.05 + 15 16 & 0.26 & 0.12 & 0.07 + 16 17 & 0.46 & 0.20 & 0.09 + 17 18 & & 0.38 & 0.15 + cccccccl 1 & 11 04 34.0 & @xmath1378 01 26 & 13.34 & 0.68 & 1.00 & & + 2 & 11 04 41.5 & @xmath1377 38 52 & 13.64 & 0.73 & 1.09 & & + 3 & 11 04 49.2 & @xmath1378 04 46 & 12.48 & 0.75 & 1.03 & & + 4 & 11 04 59.5 & @xmath1377 17 00 & 12.59 & 0.85 & 0.84 & & + 5 & 11 05 27.2 & @xmath1377 33 09 & 11.80 & 2.99 & 1.24 & & + 6 & 11 05 43.0 & @xmath1377 29 13 & 13.79 & 1.34 & 1.74 & & + 7 & 11 05 43.4 & @xmath1377 21 22 & 13.81 & 0.83 & 0.85 & & + 8 & 11 05 48.8 & @xmath1376 40 17 & 13.94 & 0.73 & 0.55 & & + 9 & 11 05 57.8 & @xmath1376 40 56 & 13.06 & 0.84 & 0.79 & & + 10 & 11 06 08.8 & @xmath1376 56 02 & 13.56 & 0.73 & 0.91 & & + 11 & 11 06 18.6 & @xmath1377 57 53 & 13.95 & 0.89 & 1.42 & & + 12 & 11 06 27.7 & @xmath1377 35 54 & 13.26 & 1.10 & 1.65 & & + 13 & 11 06 39.3 & @xmath1377 36 04 & 12.40 & 1.23 & 1.78 & iso - chai 79 & 1 + 14 & 11 06 40.1 & @xmath1376 48 35 & 12.48 & 0.94 & 1.03 & & + 15 & 11 07 06.5 & @xmath1376 37 17 & 10.22 & 1.50 & 0.93 & & + 16 & 11 07 18.8 & @xmath1377 31 56 & 12.63 & 1.93 & 1.49 & & + 17 & 11 07 27.9 & @xmath1376 48 35 & 13.91 & 0.74 & 0.38 & & + 18 & 11 07 46.9 & @xmath1376 15 17 & 12.22 & 0.71 & 0.94 & & + 19 & 11 07 50.0 & @xmath1377 38 07 & 13.50 & 1.55 & 2.23 & & + 20 & 11 08 20.6 & @xmath1376 26 49 & 13.80 & 0.67 & 0.62 & & + 21 & 11 08 22.6 & @xmath1376 49 19 & 12.32 & 1.12 & 0.82 & & + 22 & 11 08 41.9 & @xmath1377 13 22 & 13.78 & 1.47 & 0.83 & & + 23 & 11 08 42.9 & @xmath1377 10 08 & 12.85 & 0.95 & 0.92 & & + 24 & 11 08 44.5 & @xmath1376 13 29 & 12.92 & 0.79 & 0.56 & & + 25 & 11 09 29.4 & @xmath1376 34 46 & 13.68 & 0.95 & 1.35 & nir-14 & 2 + 26 & 11 09 32.3 & @xmath1376 30 17 & 13.00 & 0.94 & 1.16 & & + 27 & 11 09 33.9 & @xmath1377 53 33 & 13.90 & 0.82 & 1.19 & & + 28 & 11 09 50.7 & @xmath1378 04 19 & 13.80 & 0.77 & 1.07 & & + 29 & 11 09 52.0 & @xmath1376 39 12 & 11.67 & 0.69 & 1.02 & iso - chai 217 & 1 + 30 & 11 09 53.4 & @xmath1377 28 36 & 12.48 & 0.85 & 1.24 & iso - chai 220 & 1 + 31 & 11 09 54.1 & @xmath1376 31 11 & 12.75 & 0.98 & 1.11 & iso - chai 225 & 1 + 32 & 11 10 01.6 & @xmath1377 47 47 & 13.97 & 1.23 & 1.46 & & + 33 & 11 10 04.5 & @xmath1377 48 22 & 13.65 & 1.21 & 1.42 & & + 34 & 11 10 06.2 & @xmath1376 40 20 & 13.85 & 1.06 & 1.53 & & + 35 & 11 10 16.8 & @xmath1377 43 44 & 13.01 & 1.38 & 1.28 & & + 36 & 11 10 47.6 & @xmath1376 32 38 & 13.22 & 0.84 & 0.24 & & + 37 & 11 10 49.4 & @xmath1376 38 08 & 13.87 & 1.19 & 0.76 & & + 38 & 11 10 53.2 & @xmath1376 16 53 & 13.96 & 0.78 & 0.18 & & + 39 & 11 10 56.9 & @xmath1377 13 26 & 13.93 & 0.77 & 1.03 & & + 40 & 11 10 58.2 & @xmath1376 17 56 & 13.82 & 0.77 & 0.93 & & + 41 & 11 11 08.9 & @xmath1376 49 12 & 13.21 & 0.95 & 1.54 & & + 42 & 11 11 21.1 & @xmath1378 05 20 & 13.25 & 0.69 & 0.64 & & + 43 & 11 11 27.8 & @xmath1376 50 17 & 13.48 & 0.90 & 1.09 & & + 44 & 11 11 35.7 & @xmath1376 53 26 & 13.19 & 1.16 & 1.79 & & + 45 & 11 11 42.6 & @xmath1378 05 21 & 13.88 & 0.66 & 0.53 & & + 46 & 11 11 48.7 & @xmath1376 12 21 & 13.83 & 0.75 & 0.49 & & + 47 & 11 11 54.5 & @xmath1376 58 29 & 13.71 & 1.29 & 0.98 & & + 48 & 11 11 57.5 & @xmath1377 21 57 & 13.72 & 1.38 & 1.13 & & + 49 & 11 12 03.7 & @xmath1376 51 33 & 13.51 & 1.27 & 1.82 & & + 50 & 11 13 25.3 & @xmath1377 00 23 & 11.55 & 0.98 & 0.67 & & + 51 & 11 13 26.0 & @xmath1377 00 32 & 12.40 & 1.20 & 0.28 & & + 52 & 11 13 42.7 & @xmath1377 55 30 & 13.71 & 0.45 & 0.25 & & + 53 & 11 14 31.4 & @xmath1378 06 44 & 13.03 & 0.98 & 0.75 & & + 54 & 11 14 23.6 & @xmath1377 56 12 & 13.73 & 0.84 & 0.80 & & + 55 & 11 16 13.2 & @xmath1377 25 15 & 11.00 & 0.85 & 0.21 & & + 56 & 11 18 09.0 & @xmath1378 14 11 & 13.71 & 1.03 & 0.72 & & + cccccccl 57 & 10 40 47.6 & @xmath1377 36 49 & 15.12 & 0.95 & 0.73 & & + 58 & 11 04 17.1 & @xmath1377 25 49 & 15.34 & 0.86 & 0.65 & & + 59 & 11 04 27.0 & @xmath1377 28 02 & 14.87 & 1.14 & 1.02 & & + 60 & 11 04 30.5 & @xmath1377 34 41 & 14.33 & 0.96 & 0.89 & & + 61 & 11 04 32.0 & @xmath1377 54 09 & 15.27 & 1.02 & 0.88 & & + 62 & 11 04 35.8 & @xmath1377 29 09 & 15.34 & 0.99 & 0.94 & & + 63 & 11 04 36.2 & @xmath1377 50 12 & 14.76 & 1.21 & 1.26 & & + 64 & 11 04 38.7 & @xmath1377 24 21 & 14.49 & 1.27 & 0.87 & & + 65 & 11 04 39.1 & @xmath1377 40 05 & 14.39 & 1.56 & 1.18 & & + 66 & 11 04 48.4 & @xmath1377 49 20 & 15.45 & 1.69 & 0.85 & & + 67 & 11 05 37.0 & @xmath1376 51 18 & 15.33 & 1.48 & 1.04 & & + 68 & 11 05 41.2 & @xmath1376 38 38 & 15.12 & 0.91 & 0.58 & & + 69 & 11 05 42.6 & @xmath1377 26 30 & 14.33 & 2.09 & 2.67 & & + 70 & 11 05 43.0 & @xmath1377 31 58 & 14.43 & 1.46 & 1.16 & & + 71 & 11 06 33.5 & @xmath1377 52 26 & 14.79 & 0.98 & 0.74 & & + 72 & 11 06 35.3 & @xmath1377 21 10 & 15.94 & 1.20 & 0.95 & & + 73 & 11 06 49.4 & @xmath1377 34 37 & 14.99 & 1.51 & 1.78 & & + 74 & 11 06 51.0 & @xmath1377 11 33 & 15.08 & 1.16 & 0.64 & & + 75 & 11 06 54.5 & @xmath1377 25 54 & 15.30 & 1.36 & 0.70 & & + 76 & 11 06 59.6 & @xmath1377 18 29 & 14.90 & 1.59 & 1.52 & & + 77 & 11 07 59.0 & @xmath1378 00 13 & 15.05 & 1.41 & 0.84 & & + 78 & 11 08 27.3 & @xmath1377 47 20 & 14.75 & 1.44 & 0.93 & & + 79 & 11 08 51.6 & @xmath1377 02 30 & 14.82 & 1.75 & 0.65 & & + 80 & 11 08 51.5 & @xmath1377 06 44 & 15.10 & 0.87 & 0.64 & & + 81 & 11 09 08.3 & @xmath1378 14 25 & 15.09 & 1.31 & 0.69 & & + 82 & 11 09 18.8 & @xmath1377 39 37 & 15.10 & 1.26 & 1.10 & & + 83 & 11 09 24.1 & @xmath1376 34 55 & 14.87 & 1.46 & 1.56 & nir-11 & 2 + 84 & 11 09 24.7 & @xmath1378 13 54 & 15.69 & 0.86 & 0.78 & & + 85 & 11 09 52.7 & @xmath1377 49 08 & 14.87 & 1.80 & 1.37 & & + 86 & 11 10 01.7 & @xmath1377 19 20 & 15.35 & 1.05 & 1.00 & & + 87 & 11 10 04.3 & @xmath1377 27 15 & 15.34 & 2.01 & 0.66 & & + 88 & 11 10 17.7 & @xmath1377 22 18 & 15.69 & 1.31 & 1.03 & & + 89 & 11 10 19.6 & @xmath1377 11 20 & 14.57 & 0.94 & 0.89 & & + 90 & 11 10 24.6 & @xmath1377 23 06 & 15.58 & 1.93 & 0.73 & & + 91 & 11 10 26.2 & @xmath1376 44 09 & 15.99 & 1.18 & 0.60 & & + 92 & 11 10 29.4 & @xmath1376 31 12 & 15.36 & 1.37 & 0.58 & & + 93 & 11 10 29.8 & @xmath1377 21 12 & 15.11 & 1.43 & 0.64 & & + 94 & 11 10 37.9 & @xmath1377 01 39 & 15.29 & 1.21 & 0.75 & & + 95 & 11 10 39.2 & @xmath1377 02 48 & 15.44 & 1.45 & 1.10 & & + 96 & 11 10 46.1 & @xmath1377 17 07 & 14.47 & 0.99 & 0.88 & & + 97 & 11 11 06.3 & @xmath1378 03 57 & 14.74 & 1.00 & 0.98 & & + 98 & 11 11 11.5 & @xmath1376 44 37 & 15.02 & 1.24 & 1.21 & & + 99 & 11 11 12.3 & @xmath1376 47 51 & 15.28 & 1.01 & 0.99 & & + 100 & 11 11 15.6 & @xmath1378 13 22 & 14.71 & 1.28 & 0.87 & & + 101 & 11 11 21.3 & @xmath1377 24 15 & 15.02 & 0.97 & 0.80 & & + 102 & 11 11 28.6 & @xmath1378 05 28 & 15.02 & 1.61 & 1.03 & & + 103 & 11 11 30.2 & @xmath1378 10 10 & 15.06 & 0.89 & 0.72 & & + 104 & 11 11 34.3 & @xmath1376 47 35 & 15.34 & 0.82 & 0.67 & & + 105 & 11 11 47.3 & @xmath1377 08 12 & 14.50 & 1.40 & 1.00 & & + 106 & 11 11 48.1 & @xmath1377 25 46 & 15.06 & 0.95 & 0.81 & & + 107 & 11 12 05.1 & @xmath1377 29 58 & 15.22 & 1.17 & 1.16 & & + 108 & 11 12 13.1 & @xmath1377 22 37 & 14.47 & 1.07 & 1.02 & & + 109 & 11 13 02.6 & @xmath1377 22 08 & 14.72 & 1.13 & 0.74 & & + 110 & 11 13 23.8 & @xmath1377 25 28 & 15.64 & 1.55 & 1.55 & & + 111 & 11 13 53.1 & @xmath1377 37 01 & 14.62 & 1.93 & 1.91 & & + 112 & 11 13 56.0 & @xmath1377 06 35 & 14.28 & 1.13 & 0.72 & & + 113 & 11 14 17.3 & @xmath1376 48 18 & 15.11 & 0.95 & 0.86 & & + 114 & 11 15 00.3 & @xmath1378 03 13 & 15.04 & 0.96 & 0.89 & & + 115 & 11 16 30.3 & @xmath1378 05 08 & 15.30 & 1.12 & 0.77 & & + 116 & 11 16 46.0 & @xmath1377 37 40 & 15.06 & 0.89 & 0.74 & & + 117 & 11 17 36.8 & @xmath1377 45 35 & 15.18 & 1.03 & 0.68 & & + 118 & 11 18 02.2 & @xmath1377 41 30 & 15.61 & 0.86 & 0.58 & & +	we describe a near - infrared imaging survey covering @xmath0 1 deg@xmath1 of the chamaeleon i dark cloud . the survey is complete for @xmath2 @xmath3 15.0 , @xmath4 @xmath3 16.0 , and @xmath5 @xmath3 16.5 , roughly two magnitudes more sensitive than previous large scale surveys . we use the large number of background stars detected to derive an accurate near - infrared extinction law for the cloud and select new candidate members with near - infrared color excesses . we list @xmath0 100 candidates of the cloud with @xmath2 @xmath6 12.0 , based on their positions in the @xmath7 , @xmath8 color - color diagram . these new stars have low luminosities ( @xmath2 @xmath0 12 16 , @xmath8 @xmath9 0.5 1.5 ) and may have masses close to or even below the hydrogen burning limit .
@xcite ( henceforth kr ) and shahbaz , charles & king ( 1998 ) ( henceforth sck ) examined the x - ray light curves of transient low mass x - ray binaries ( lmxbs ) in decay , characterising them as simple exponential or linear decays . in some cases , however , the decays show a distinct knee . as shown in fig . [ fig:1808 ] , this is not a secondary maximum in the sense that the x - ray count rate declines throughout . in x - ray bright disc - accreting systems , including transient lmxbs during outburst , the temperature of the accretion disc is dominated by x - ray heating from the inner accretion regions across most of the disc . for a disc in which the scale height can be described as a function of radius by @xmath0 for constant @xmath1 and @xmath2 , the temperature , @xmath3 , is given by @xcite as @xmath4 where @xmath5 is the disc albedo , @xmath6 is the central x - ray luminosity heating the disc , assumed to be emanating radially from a small spherical source . only if @xmath7 is the whole surface of the accretion disc illuminated ; if @xmath8 ( [ eq : tofr ] ) predicts a surface temperature of zero which is invalid because of local viscous heat production . consistent with kr we adopt a value of @xmath9 , although see section [ sec : dm2small ] . ( [ eq : tofr ] ) can be rewritten to give the maximum radius @xmath10 which is heated by x - rays to the temperature @xmath11 needed to remain in the hot , viscous state . thus @xmath12 we abbreviate this equation as @xmath13 we note that especially in the case of a black hole the x - ray source will be the inner accretion disc , and as such disc - like rather than point - like . this introduces an factor of @xmath14 into the r.h.s . of ( [ eq : rhot ] ) , which is equivalent to modified values of @xmath2 and @xmath15 in ( [ eq : rhotl ] ) ; specifically , the index of @xmath16 in the l.h.s . becomes @xmath17 , and @xmath15 gains a factor of @xmath1 . however , @xmath2 and @xmath1 may differ systematically between neutron star and black hole systems , in which case the differences in ( [ eq : rhotl ] ) will be less simple . kr showed that x - ray heating during the decay from outburst causes the light curves of transient lmxbs to exhibit either exponential or linear declines depending on whether or not the luminosity is sufficient to keep the outer disc edge hot . they note that exponential decays must revert to the linear mode when the x - ray flux has decreased sufficiently , but do not analyse this in detail . nor do they consider the effects of mass transfer from the donor star , @xmath18 , during the outburst . whilst in many systems @xmath18 is negligible compared to the mass accretion rate onto the compact object , @xmath19 , during outburst , this is not necessarily the case . in this paper we consider these issues . section [ sec : theory ] contains our theoretical predictions for the decay light curve . in section [ sec : fitting ] we identify some suitable light curves and describe their parametrization . we compare the results from the fitting in section [ sec : comparison ] . the mass of the accretion disc during the exponential decay is given by equation 3 of kr ; @xmath20 where @xmath21 is some measure of viscosity , taken by kr to be the viscosity near the outer disc edge with a value of @xmath22 . in the case of significant mass transfer into the disc from the donor star the central accretion rate is given by @xmath23 using this to replace @xmath24 in ( [ eq : mdisc ] ) we obtain @xmath25 therefore the mass of the hot disc can be written @xmath26 where @xmath27 is the constant of integration , and the x - ray luminosity is proportional to @xmath28 at some time , @xmath29 , the temperature of the outer disc edge , while still dominated by x - ray heating , will be only just sufficient to remain in the hot , viscous state , i.e. @xmath30 . we denote the corresponding x - ray luminosity as @xmath31 , with the central accretion rate @xmath32 . the x - ray luminosity at earlier times is @xmath33 where @xmath34 , and @xmath35 is the efficiency with which rest mass is liberated as x - rays . when @xmath36 the outer part of the disc is no longer kept in the hot , viscous state by central irradiation . the radius of the hot part of the disc will decrease with the decreasing x - ray luminosity , corresponding to the late linear decline of kr . this , too , may be modified by the donor star mass loss rate , depending on the mass accretion rate through the now cold outer disc . designating @xmath37 as the rate at which mass is transported inwards through the cold disc , we examine the two extreme behaviours of @xmath38 . for the minimally efficient cold disc @xmath39 for all @xmath40 , and none of @xmath18 reaches the hot disc when @xmath36 . at the other extreme @xmath41 and the cold disc does not increase in surface density with time . the occurrence of outbursts suggests that the latter case is false ; with stable cold - state mass transfer , outbursts would never occur . the mass in the hot disc changes due to three factors : accretion from its inner edge , the loss of material into the encroaching cold disc , and mass transfer from the cold disc . hence , @xmath42 using the approximate substitution from ( [ eq : rhotl ] ) that @xmath43 ( [ eq : mhotlin ] ) becomes @xmath44 additionally , equation 2 of kr allows @xmath45 to be written as @xmath46 equating ( [ eq : dmhotlin ] ) to @xmath47 ( [ eq : kr2 ] ) , @xmath48 so using equation 2 of kr , @xmath49 we now substitute the two cases of @xmath50 outlined above . when mass transfer into the hot disc is negligible , @xmath51 and the decay is the linear decline predicted by kr as expected . when @xmath52 , @xmath53 while @xmath54 this reduces to the same form . as @xmath19 decreases it will approach the limit of @xmath55 , at which point @xmath56 . the exact form of the decay will depend on the unreliably known form of @xmath37 , and during the late linear decline viscous heating near @xmath10 becomes non - negligible . we do not predict the form of the late decay other than to note that if there is some radius @xmath57 satisfying @xmath58 then the l.h.s . of ( [ eq : lqui ] ) will give the asymptotic limit of the decay and the accretion luminosity at the beginning of quiescence . as a final point on the linear decay we note that the viscous instability may cause a range of annuli within the cold disc to enter the hot state and rapidly transfer mass inward . some of this may enter the inner hot disc , or else may increase the surface density of the subsequent cold disc such that the inner hot disc receives more material at its outer edge . consequently , small rebrightenings or other artefacts potentially complicate the linear decay . to examine the observed x - ray light curves of transient lmxbs it is useful to examine the constraints on the transition . the gradient of the exponential decay at time @xmath29 is @xmath59 whilst the gradient of the linear decline is given by ( [ eq : ddmc ] ) , @xmath60 from ( [ eq : rhotl ] ) we adopt @xmath61 where it is assumed that at the outer disc edge @xmath62 . we therefore obtain @xmath63 in the case that @xmath64 is small relative to @xmath6 the gradient of the exponential decay at time @xmath29 is equal to that of the subsequent linear decline ; the first derivative of the x - ray light curve is expected to vary smoothly throughout the decay . if however @xmath64 can not be neglected then the linear decline is steeper than the exponential decay at @xmath29 and a discontinuity in gradient is expected . whether this is detectable depends on the value of @xmath18 and the quality of the x - ray data . some x - ray light curves , including most of those discussed in this paper , include a knee feature in the decay from outburst . where this can be explained by an exponential to linear transition with non - negligible @xmath64 we label it as a ` brink ' . in general the decay light curve may contain multiple knees due to different effects , but we do not expect more than one brink unless the source rebrightens between them such that each occurs at approximately the same luminosity , corresponding to a disc not varying greatly in radius . x - ray light curves from the all - sky monitor ( asm ) and proportional counter array ( pca ) of the _ rxte _ space telescope were examined to find examples of decay light curves which fit the template predicted above . we use our analysis to determine distances and accretion disc radii . system parameters from the literature for the objects analysed are collected in table [ tab : syspar ] for comparison with quantities we derive . since we will be calculating the radius of the accretion disc we calculate from the system parameters the orbital separation @xmath65 . the count rates @xmath66 as a function of time @xmath67 were fitted with @xmath68 for the exponential phase , where @xmath69 is the limit of the exponential decay , @xmath70 is the count rate at the brink , and @xmath71 is the time - scale of the decay . the linear decline is given by @xmath72 where @xmath73 is the time after the brink at which the count rate would become zero if the linear decline continued . using the assumed distance to each source with a hydrogen column density of @xmath74 it is possible to convert the count rates @xmath70 and @xmath69 into absolute x - ray luminosities @xmath31 and @xmath75 , the x - ray luminosity at the brink and the limit of the exponential decay ( corresponding to @xmath18 ) respectively . for the two systems for which we have simultaneous asm and pca light curves , and , the ratio of count rates is approximately 31 . we assume this conversion to be valid for all sources for comparison purposes , and present luminosities in the 1.512@xmath76 range corresponding to the asm spectral range using an assumed power law spectrum with photon index of @xmath77 . the luminosities thus determined are given in table [ tab : fitpar ] . we note that if the spectrum differs between systems our deduced @xmath31 is invalid , though our deduced exponential timescale will be unaffected . using ( [ eq : rhotl ] ) and ( [ eq : roftau ] ) it is possible to calculate the disc radius independently from the measured values of @xmath70 and @xmath71 ; the result based on @xmath71 is of particular importance as it does not depend on the conversion of count rate to flux but only on a directly observable timescale . it is also possible to determine @xmath18 from @xmath69 . results are given in table [ tab : calcpar ] ; for comparison we have also listed the circularization radius , @xmath78 , and the distance between the compact object and the inner lagrange point , @xmath79 , for each source based on the values in table [ tab : syspar ] . discussion of the choice of values for @xmath15 and @xmath21 is given in section [ sec : comparison ] . [ [ section ] ] ~ fig . [ fig:1543 ] shows that the exponential - then - linear fit to the decay from the 2002 outburst of is significantly better than the least - squares regression line . because no discontinuity is seen in the gradient , the fitting routine used was based on the continuous derivative model ( see section [ sec : cont ] ) . previous outbursts were reported in 1971 , 1983 and 1992 . the source therefore spends only a small fraction of its time in outburst , suggesting that @xmath18 is small compared with the central accretion rate during outburst , consistent with the source s failure to display a broken gradient . [ [ section-1 ] ] ~ was first detected in outburst in 1996 and has subsequently shown outbursts in 1998 , 2000 , 2002 , and 2005 . hence it is expected that the ratio of @xmath18 to the peak central accretion rate is much higher than in . this is consistent with the appearance of a broken decay in the two well observed outbursts , shown in figs . [ fig:1808 ] and [ fig:1808_2002 ] . each light curve is well fitted by a decay that is exponential towards a positive limit followed by a sharp gradient discontinuity ( brink ) leading to a linear decline . the fit to the 1998 outburst includes all points that can be confirmed to occur during the decay ; the observation at day 16 of fig . [ fig:1808 ] may belong to a period of constant luminosity or even an unsteady rise to outburst , while the observations around day 33 of fig . [ fig:1808 ] may involve a rebrightening . even if this is not the case the late linear decline is expected to become increasingly complex , as we noted in section [ sec : linear ] . comparison with fig . [ fig:1808_2002 ] shows that the brink occurs at approximately the same count rate in both cases . this is expected if @xmath80 remains constant ( i.e. @xmath31 is the same ) . the dashed line in fig . [ fig:1808 ] indicates the fit proposed by sck , in which they identified the surplus around the brink as a secondary maximum . we disagree and view their fit as inadequate in having a negative limit to the exponential decay . it is possible that features in other decay light curves identified as secondary maxima but having monotonically decreasing x - ray flux have the same explanation . a second issue in fig . [ fig:1808_2002 ] is the broken nature of the light curve prior to the brink . whilst we do not have a detailed explanation for this , we speculate that it may be associated with some regions of the disc being self - shadowed and initially remaining in the cool state . 3d hydrodynamic simulations ( e.g. * ? ? ? * ) show self - shadowing is likely , due to spiral density waves and irradiated disc warping ( c.f . consequently the effective area of the hot disc , the quantity probably corresponding to our measurement of @xmath81 , increases during the decay in an apparently stepwise manner , introducing new material into the hot disc and therefore resetting the central luminosity to a higher value . this view is supported by the fact that fitting the region around day 65 of fig . [ fig:1808_2002 ] indicates a shorter exponential time - scale than from day 70 to the brink , corresponding to a smaller outer radius of the hot region . by day 70 we assume that the entire disc is in the hot state . the growing irradiated area could lead to an actual increase in @xmath6 , hence it is an explanation for a type of secondary maximum . it is possible that the light curve will monotonically decline despite the growing irradiated area , hence producing knees in the light curve . the light curve of around day 67 of fig . [ fig:1808_2002 ] is suggestive of this behaviour . [ [ section-2 ] ] ~ was first detected during its 2002 outburst , shown in fig . [ fig:1751 ] , and a second outburst was observed in 2005 . if this repetition interval is normal for the system , it indicates a ratio of @xmath18 to peak central accretion rate more similar to that of than to . consistent with this , the 2002 decay light curve has a shape similar to those of in that a smooth exponential decay is broken by occasional rebrightenings . therefore only the last portion , appearing to correspond to a single smooth decay , is fitted . the linear portion of this decay terminates at a constant level . in accordance with ( [ eq : ddmc2 ] ) we interpret this as the rate at which mass can be transferred through the mostly cold disc ; this rate will continue during quiescence whilst the disc mass increases . [ [ section-3 ] ] ~ a third observed example of the brink is found in the only observed ( 2002 ) outburst of . here the overall shape before the brink is exponential , but a maximum at about day 10 of fig . [ fig:0929 ] and a minimum at about day 19 are anomalous . possibly the outer edge of the disc remained cold until late in the exponential decay and several secondary maxima of the type already discussed have occurred . [ [ section-4 ] ] ~ is , like , and , one of seven known milli - second pulsar ( msp ) systems . msps have short orbital periods and hence small and therefore relatively simple discs . we therefore expect the x - ray light curve of to also display the brinked decay seen in the first three . [ fig:1807 ] does not appear to lend itself to this type of fitting , and we have been unable to obtain reasonable fitting parameters . one possible explanation is that the linear decay had already begun at the start of the observed light curve ; in this case the brink count rate must be around fifty per cent greater than that seen in . the orbital separation of ( c.f . table [ tab : syspar ] ) is half that of , so the disc must be correspondingly smaller . a high value of @xmath70 can only be accounted for if is at a distance of less than about @xmath82 so the luminosity remains low while the countrate is high . conversely , if the light curve corresponds to the exponential portion with several rebrightenings , the maximum value of @xmath70 may be 15 , with the light curve after day 80 tentatively assumed to be linear . if the disc is half the radius of that of , in proportion to the orbital separation , ( [ eq : dmhotlin ] ) indicates that the brink luminosity will be one quarter that of . the pca count rate at the proposed brink is one quarter that of , indicating a similar source distance of @xmath83 . we deem the latter explanation more likely , and suspect that the exponential decay of is complicated by self - shadowing of the disc . the more extreme mass ratio of than , indicated in table [ tab : syspar ] , suggests that the accretion disc radius in is a larger fraction of the orbital separation , indicating a larger brink luminosity than assumed above , increasing the distance estimates in both cases . a lower value of @xmath70 is possible in the second case , which would also indicate a greater source distance . [ [ section-5 ] ] ~ the earliest part of the _ rxte _ light curve of , shown in fig . [ fig:1744_1996 ] , shows the system already in outburst ; the first pca data are consistent with the outburst decay . the light curve has a brink just after day 170 in fig . [ fig:1744_1996 ] . entered a relatively quiescent period in 1996 april ( @xmath84day 200 of fig . [ fig:1744_1996 ] ) , punctuated by a mini - outburst shown inset in fig . [ fig:1744_1996 ] following day 260 which reached less than half the brink luminosity . a second outburst , smaller than the 1996 outburst , followed during the first four months of 1997 . since then the system has been quiescent . [ fig:1744_1997 ] shows the decay from the 1997 outburst , in which a brink is clearly present . our confidence in the identification of these two knees as brinks is increased by their similarity in luminosity . [ [ section-6 ] ] ~ as shown in fig . [ fig : gx339 - 4 ] , follows an exponential decay to less than 10 per cent of its outburst maximum , a much smaller fraction than is exhibited by the other systems considered so far . the relatively few existing points fit the brinked decay model well . the late decay departs smoothly from the linear decline , in line with expectations ( c.f . section [ sec : linear ] eqn . [ eq : ddmcgeneric ] and following comments ) . [ [ section-7 ] ] ~ although not enough data exist to be able to establish the parameters of the brinked decay , a knee occurs in the pca light curve , which has been marked in fig . [ fig:1550 ] . interpreting this as the brink it is therefore possible to calculate @xmath80 based on @xmath70 but no other system parameters . [ [ section-8 ] ] ~ the asm light curve of shows one outburst , starting in 1996 april and persisting until 1997 august . this outburst is broadly divided into two broad peaks , of which the decay of the second is shown . although the system had been in outburst for a considerable time , @xcite note that no stable hot disc state exists for luminosities less than the eddington luminosity ; it is therefore unclear whether material from the outer disc edge participated in the outburst . as shown in table [ tab : calcpar ] , the accretion disc radius calculated from the brink luminosity and the exponential timescale match each other , but are considerably less than the system s circularization radius . this may indicate that the mechanism examined here is applicable , but acts on a small inner disc region rather than on the disc as a whole . possibly this second peak is caused by a high surface density remnant left by the preceding outburst activity . returned to outburst in 2005 ; if the average flux between the beginning of the two successive outbursts is taken as an indicator of @xmath18 , this value is found to lie below the brink by a factor of three . the mass transfer rate from the secondary star may be enhanced during outburst , but it is likely that the mass flow onto the ( constant radius ) outer edge of the hot disc during the exponential phase is supplied from additional disc structures rather than from the secondary directly . the complexity of the disc in this system does not necessarily invalidate the application of this model during at least some phases of the decay . [ [ section-9 ] ] ~ no pca light curves for are available , so the best evidence for our model being applicable to this system comes from the asm light curve . this shows numerous outbursts , the decays of which are shown in fig . [ fig:1705 ] . a range of smaller outbursts also occur which are not shown in fig . [ fig:1705 ] . arrows mark features in the light curve that we have interpreted as the brink of each decay ; using these as part of the initial conditions for finding a best fit we obtained the best - fitting decays overplotted . clearly there is substantial variation in the height of the brink , which indicates variation in the radius of the disc . in order to explain this we used the asm light curve to assess the fluence of each outburst . a greater fluence suggests a greater initial disc mass , which leads us to expect positive correlation between the fluence and calculated disc radius . relative fluences are given in table [ tab : calcpar ] , and fig . [ fig:1705flu ] shows a positive correlation between outburst fluence and @xmath85 . if at least some of the knees marked in fig . [ fig:1705 ] actually correspond to the disc radius mechanism being considered ( i.e. they are brinks ) , then substantial variation in the radius is seen . furthermore , the limit of the exponential decay is clearly in several cases below the mean long term count rate of 12 counts per second , based on the numerical integration of the asm light curve . we account for this by the additional theory given in section [ sec : dm2small ] . [ cols="<,^,^,^,^,^,^,^",options="header " , ] to force the disc radii based on @xmath71 to lie within the range @xmath86 for the two best described systems , and , the viscosity of @xmath87 was adopted , close to the value of @xmath88 proposed by kr . to make @xmath85 consistent with @xmath89 for most of the neutron star systems a value of @xmath90 was adopted . a better fit was found for the black hole systems by reducing @xmath15 by a factor of ten , consistent with the extra factor of @xmath1 discussed in the introduction . @xmath15 is somewhat small compared with the range of @xmath91 adopted by kr . furthermore , since @xmath92 is based on @xmath93 , our systematic underestimate of @xmath31 by considering only the flux in the 1.512@xmath76 band suggests a yet smaller value of @xmath15 . calculated disc radii using these parameter values are given in table [ tab : calcpar ] . [ fig : radiiratio ] shows the relation between the values of @xmath80 calculated from @xmath31 and @xmath71 . we note that for the results would be more consistent if we adopted the lower value of @xmath15 , appropriate to black hole systems . this may reflect differences in the source spectrum as well as intrinsic differences in @xmath15 since we have assumed the same relation between 1.512@xmath76 count rate and total x - ray flux for all systems . in addition to this effect , we note that the results for in fig . [ fig : radiiratio ] do not lie on a straight line . to examine this issue we first determined a best - fitting value of @xmath94 from decays _ a _ and _ d _ , which we deem the strongest results by inspection of the light curve . we then demanded that this value of @xmath94 apply to each decay of . in fig . [ fig:1705res ] we show the effect this has on the fits , plotting the residuals for the individual best - fits , and for the constrained fits . if the original fits were strongly determined by the data , we would expect the residuals to appear significantly worse in the constrained fits . in contrast , we find that in some cases the residuals appear to improve , and in none is the fit substantially worse ; the reduced @xmath95 for the whole set of fits increases from 11.4 when @xmath31 and @xmath71 are optimized individually to 12.1 for the constrained fitting . we note that technique is only applicable when comparing decays of the same source , where variation in the quantity @xmath94 is minimized . the large value of @xmath85 for may be again the result of spectral differences , or in this case may be the result of an erroneous source distance estimate . has been assumed to lie at a distance of @xmath96 , being possibly associated with the galactic centre ( e.g. * ? ? ? * ) ; if the true distance is more like @xmath97 then the two values of @xmath80 will be approximately equal . the increasing value of @xmath89 for the 2002 decay of and the erratically increasing luminosity between days 65 and 70 of fig . [ fig:1808_2002 ] are both consistent with the interpretation that the outer disc edge was not in the hot state at the beginning of the decay . as such , the disc radius measured corresponds at first to a hot disc smaller than the disc as a whole , which is therefore depleted on a shorter timescale . at some point the outer disc is heated and enters the hot state . as well as increasing the timescale of the decay , this introduces extra material into the hot disc , causing rebrightening . we have assumed in the initial model that the limit of the exponential decay is equal to the assumed constant mass transfer rate from the donor star . this requires that the hot disc is capable of supporting the additional mass flux @xmath18 at all times and radii . however , the exponential nature of the decay itself implies that as the hot disc is depleted in surface density , its mass transfer rate falls . the mass transfer equation of the hot disc is linear in surface density , which can therefore be decomposed into two components ; one representing the decay analogous to an unfed disc , the other corresponding to the effects of feeding the hot disc at the rate @xmath18 . the former component we denote @xmath98 and the latter @xmath99 . because the viscosity of the accretion disc depends on temperature , and the temperature is declining during the decay , it is necessary for @xmath99 to increase with time if it is to sustain the same mass flux , implying that the corresponding flux @xmath100 must decrease inwards . thus the limit of the central mass accretion rate does not correspond to @xmath18 , but rather @xmath101 . the relevance of this issue is made clear by the light curves of , in which the limit of the exponential decay is often below the long term mean count rate of @xmath102 asm counts @xmath103 . the standard derivation of the diffusion equation ( e.g. * ? ? ? * ) can be used to give the mass transfer rate inward through the disc @xmath104 as @xmath105 using the relation @xmath106 we derive equation 9 of @xcite as expected . we require the form given in ( [ eq : mu ] ) to prevent a constant of integration arising . to make use of ( [ eq : mu ] ) we must adopt an appropriate form for the viscosity @xmath107 . we use the @xmath108-viscosity form @xmath109 where @xmath110 is the sound speed , and the scale height is given by equation 2 of @xcite ; @xmath111 using ( [ eq : tofr ] ) to substitute for @xmath3 and assuming that @xmath112 is a power law we obtain @xmath113 for a mean particle mass @xmath114 . we write ( [ eq : hpower ] ) in terms of constant @xmath1 as @xmath115 where @xmath9 is derived and @xmath116 is adopted from kr . @xcite note that if the x - ray flux is assumed to be absorbed at the mid - plane , the power law index is changed to @xmath117 ; we therefore consider values of @xmath118 as valid to simplify subsequent derivation . using this form for @xmath119 in ( [ eq : tofr ] ) to derive the viscosity we find that @xmath120 where @xmath121 is defined as @xmath122 we simplify ( [ eq : nu ] ) by adopting @xmath123 , so ( [ eq : mu ] ) becomes @xmath124 if we impose @xmath125 for all radii we obtain @xmath126 corresponding to a total contribution to mass in the hot disc of @xmath127 where the radius of the inner disc edge is much smaller than @xmath92 . as @xmath6 decreases , a greater surface density and total mass is required to transport the same @xmath18 inward , so some of this material must be absorbed to increase the surface density . for a typical luminosity at the beginning of the exponential phase 2.5 times greater than @xmath31 the change in mass required is @xmath128 examining , for which we have the largest mean luminosity relative to @xmath31 , we find a mean long term count rate of @xmath129 , whereas the limit of the decay _ a _ for instance is @xmath130 . taken over an exponential phase duration of 57 days this makes the l.h.s . of ( [ eq : deltam ] ) equivalent to @xmath131 asm units of mass . using the conversion factor from webpimms of @xmath132 @xmath133 , at a distance of @xmath134 this integrated flux is equivalent to @xmath135 . using @xmath136 ( as in table [ tab : calcpar ] ) we find @xmath137 in strong agreement with the values given in table [ tab : calcpar ] despite the weakness of the derivation . in principle it is necessary to simultaneously derive the time and radius dependencies of @xmath104 given the outer boundary condition of @xmath138 , but this is beyond the scope of this paper . a knee in the light curve of the decay from outburst of an transient lmxb is a natural consequence of mass transfer onto the outer edge of the disc , since this supply is effectively cut off from the compact object when the outer disc enters the cool low - viscosity state . when the knee can be interpreted in this way we refer to it as a brink , since knees of other types may be seen in the decay outburst . the x - ray luminosity at which the brink occurs is that at which the outer disc edge is just kept hot by central illumination , allowing this radius to be calculated . in addition , the exponential time - scale of the decay gives a second measure of the disc radius ; these two estimates are in good agreement . by examining systems with well constrained disc radii we deduced values of the constants @xmath15 and @xmath21 required to perform the calculations . this allows more accurate results to be obtained in other systems . where the source distance is poorly constrained , the timescale @xmath71 allows the absolute luminosity @xmath31 to be estimated , providing a measure of source distance . as well as ` secondary maxima ' associated with the brink , we have identified maxima whose natural explanation is the discontinuous increase in the radius of the hot disc during outburst . this increase arises from initially self - shadowed disc regions becoming irradiated . this process can occur until the whole disc is in the hot state . it is characteristic for the time - scale of the exponential decay to increase with each such maximum , and it is expected that the limit of each exponential decay will also increase . although we do not have sufficient data to examine the matter in detail , it is probable that some secondary ` maxima ' of this type will also involve monotonic decay , and therefore probably will appear as a knee in the x - ray light curve . current instrumentation is capable of measuring the light curves of x - ray transients in m31 , for example @xcite show part of a 2004 july decay of . while this light curve , with only 4 data points over 3 consecutive days , is too sparse to allow us to deduce any parameters from fitting the decay , our method could be applied to better - sampled extragalactic transient decays . since the distance is relatively well - known for objects in external galaxies , we could use our method to deduce their accretion disc radii . this would contribute to constraints on the fundamental system parameters , such as orbital separation and mass ratio , which might otherwise be impossible to determine . we thank andrew king for comments on an earlier version of this work . crp was supported by a pparc studentship , and thanks the ou dept . of physics and astronomy for funding during the final write - up of this paper . mf acknowledges the cnrs for financial support . this research has made use of data obtained through the high energy astrophysics science archive research center online service , provided by nasa s goddard space flight center ( gsfc ) . asm results were provided by the asm and rxte teams at mit and at the rxte sof and gof at nasa s gsfc . 99 burderi l. , king a. r. , szuszkiewicz e. , 1998 , apj , 509 , 85 campana s. , ravasio m. , israel g. l. , mangano v. , belloni t. , 2003 , apj , 594 , l39 chakrabarty d. , morgan e. h. , 1998 , nat , 394 , 346 falanga m. et al . , 2005 , a&a , 436 , 647 foulkes s. b. , haswell c. a. , murray j. r. , 2006 , mnras , 366 , 1399 galloway d. k. , chakrabarty d. , morgan e. h. , remillard r. a. , 2002 , apj , 576 , l137 gierliski m. , poutanen j. , 2005 , mnras , 359 , 1261 haberl f. , titarchuk l. , 1995 , a&a , 299 , 414 hynes r. i. et al . 1998 , mnras , 300 , 64 hynes r. i. , steeghs d. , casares j.,charles p. a. , obrien k. , 2004 , apj , 609 , 317 de jong j. a. , van paradijs j. , augusteijn t. , 1996 , a&a , 314 , 484 king a. r. , ritter h. , 1998 , mnras , 293 , l42 ( kr ) king a. r. , 1998 , mnras , 296 , l45 markwardt c. b. , swank j. h. , strohmayer t. e. , in t zand j. j. , marshall f. e. 2002 , apj , 575 , l21 nishiuchi m. et al . , 1999 , apj , 517 , 436 orosz j. a. , bailyn c. d. , 1997 , apj , 477 , 876 orosz j. a. , jain r. k. , bailyn c. d. , mcclintock j. e. , remillard r. a. , 1998 , apj , 499 , 375 orosz j. a. et al . , 2002 , apj , 568 , 845 park s. q. et al . , 2004 , apj , 610 , 378 pringle j. e. , 1981 , ara&a , 19 , 137 pringle j. e. , 1996 , mnras , 281 , 357 rappaport s. , joss p. c. , 1997 , apj , 486 , 435 shahbaz t. , charles p. a. , king a. r. , 1998 , mnras , 301 , 382 ( sck ) trudolyubov s. , priedhorsky w. , cordova f. , 2006 , apj , 645 , 277 in t zand j. j. m. et al . , 2001 , a&a , 372 , 916	the outbursts of low mass x - ray binaries are prolonged relative to those of dwarf nova cataclysmic variables as a consequence of x - ray irradiation of the disc . we show that the time - scale of the decay light curve and its luminosity at a characteristic time are linked to the radius of the accretion disc . hence a good x - ray light curve permits two independent estimates of the disc radius . in the case of the milli - second pulsars and the agreement between these estimates is very strong . our analysis allows new determinations of distances and accretion disc radii . our analysis will allow determination of accretion disc radii for sources in external galaxies , and hence constrain system parameters where other observational techniques are not possible . we also use the x - ray light curves to estimate the mass transfer rate . the broken exponential decay observed in the 2002 outburst of may be caused by the changing self - shadowing of the disc . [ firstpage ] accretion , accretion discs , binaries : close , x - rays : binaries , stars : individual : xte j1808.4 - 3658 , stars : individual : 4u 1705 - 44
complex dynamical systems in science and engineering often involve multiple time scales , such as slow and fast time scales , as well as uncertainty caused by noisy fluctuations . for example , aerosol and pollutant particles , occur in various natural contexts ( e.g. , in atmosphere and ocean coasts @xcite ) and engineering systems ( e.g. spray droplets ) , are described by coupled system of differential equations . some particles move fast while others move slower , and they are usually subject to random influences , due to molecular diffusion , environmental fluctuations , or other small scale mechanisms that are not explicitly modeled @xcite . invariant manifolds are geometric structures in state space that help describe dynamical behaviors of dynamical systems . a slow manifold is a special invariant manifold , with an exponential attracting property and with the dimension the same as the number of slow variables . the reduced system on a slow manifold thus characterizes the long time dynamics in a lower dimensional setting , facilitating geometric and numerical investigation . existence for slow manifolds of stochastic dynamical systems with slow - fast time scales has been investigated recently @xcite . however , stochastic slow manifolds are difficult to depict or visualize . therefore , in this paper , we approximate these random geometric invariant structures in the case of large time scale separation . we derive an asymptotic approximation for these stochastic manifolds , and illustrate the random slow manifold reduction by considering the motion of aerosol particles in a random cellular fluid flow . the reduced slow system , being lower dimensional , facilitates our understanding of particle settling . we comment that approximations for individual solution paths ( not a stochastic slow manifold ) for stochastic slow - fast systems have been well investigated @xcite . approximations for deterministic slow manifolds have also been considered @xcite . this paper is organized as follows . an approximation method for random slow manifolds is considered in [ slow888 ] , and the dynamics of aerosol particles in a random flow field is investigated in [ settle888 ] . we first examine the existence of a random slow manifold for a slow - fast stochastic dynamical system , then devise an approximation method for this slow manifold , and thus obtain a low dimensional , reduced system for the evolution of slow dynamics . we consider the following slow - fast system of stochastic differential equations ( sdes ) @xmath0 here @xmath1 and @xmath2 are respectively @xmath3 and @xmath4 matrices . the nonlinear functions @xmath5 and @xmath6 are @xmath7-smooth and lipschitz continuous with lipschitz constants @xmath8 and @xmath9 , respectively . the parameter @xmath10 is a positive number and the parameter @xmath11 is small ( representing scale separation ) . the stochastic process @xmath12 is a two - sided @xmath13-valued wiener process . when @xmath14 and @xmath15 are locally lipschitz but the system has a bounded ( i.e. , in mean square norm ) absorbing set , a useful trick is to cut - off the nonlinearities to zero outside the absorbing set , so that the new system has global lipscitz nonlinearities and has the same long time random dynamics as the original system . the existence of a random slow manifold for this system has been considered in @xcite but we adopt a method from our earlier work @xcite . we recall the definition of a random dynamical system ( rds ) in a probability space @xmath16 . let @xmath17 be a @xmath18 , @xmath19measurable flow , i.e. , @xmath20 additionally , the measure @xmath21 is supposed to be an invariant measure for @xmath22 , i.e. @xmath23 , for all @xmath24 . for a wiener process driving system , we take @xmath25 consisting of all continuous sample paths of @xmath26 on @xmath27 with values in @xmath13 and @xmath28 . on @xmath29 , the flow @xmath22 is given by the wiener shift @xmath30 a measurable map @xmath31 is said to satisfy the cocycle property if @xmath32 a random dynamical system consists of a driving system @xmath33 and a measurable map with the cocycle property . introduce a banach space @xmath34 as our working space for random slow manifolds . for @xmath35 , define + @xmath36\rightarrow r^n : \nu\ ; \text { is continuous and } \sup\limits_{t\leq 0 } \|e^{\lambda t}\nu(t)\|_{r^n } < \infty\big\},\]]and@xmath37\rightarrow r^m : \nu\ ; \text { is continuous and } \sup\limits_{t\leq 0 } \|e^{\lambda t}\nu(t)\|_{r^m } < \infty\big\},\ ] ] + with the following norms respectively @xmath38 and @xmath39 let @xmath34 be the product banach space @xmath40 , with the norm @xmath41 for matrices @xmath1 and @xmath2 , we make the following assumptions : @xmath42 : there are constants @xmath43 , @xmath44 and @xmath45 , satisfying @xmath46 and @xmath47 , such that for every @xmath48 and @xmath49 , the following exponential estimates hold : @xmath50 @xmath51 : @xmath52 . + in order to use the random invariant manifold framework @xcite , we transfer an sde system into a random differential equation ( rde ) system . introduce the following linear langevin system @xmath53 it is known @xcite that the following process @xmath54 is the stationary solution of the linear system @xmath55 moreover , @xmath56 similarly , @xmath57 is the stationary solution of the following linear sde system @xmath58 with @xmath59 and @xmath60 denoting @xmath61 , which is also a wiener process @xcite , and has the same distribution as @xmath62 , with @xmath63 . therefore , by a transformation @xmath64 at the second equal sign and then omitting the prime in @xmath65 , we have @xmath66 and @xmath67 moreover , by defining @xmath68 at the second equal sign below , we get @xmath69 the equations and indicate that @xmath70 and @xmath71 are identically distributed with @xmath72 and @xmath73 , respectively . and by and , @xmath74 and @xmath73 have the same distribution . we then introduce a random transformation @xmath75 where @xmath76 satisfies system . + then the sde system is transferred into the following rde system , @xmath77 by the variation of constants formula , this rde system is further rewritten as @xmath78 as @xmath79 , we have the following estimation , @xmath80 for @xmath81 satisfying @xmath82 . + letting @xmath83 , we get the expression of the rde system , @xmath84 we rescale the time by letting @xmath85 , from system and by we get , @xmath86 , \,\,\,x\in \mathbb{r}^n,\\\label{rde - var - y } y'(\tau\varepsilon ) & = & by + g(x(\tau\varepsilon ) , y(\tau\varepsilon)+\sigma\eta(\theta_\tau\psi_\varepsilon\omega)),\,\ , y\in \mathbb{r}^m,\end{aligned}\ ] ] where @xmath87 . + we can rewrite these as the integral form below , @xmath88\,ds,\\ y(\tau\varepsilon ) & = & \int_{-\infty}^\tau e^{b(\tau - s ) } g(x(s\varepsilon ) , y(s\varepsilon)+\sigma\eta(\theta_s\psi_\varepsilon\omega))\,ds.\end{aligned}\ ] ] we now recall some basic facts about random slow manifolds and dimension - reduced systems , when the scale separation is sufficiently large . a random set @xmath89 is called a random slow manifold ( a special random inertial manifold ) for the system , if it satisfies the following conditions @xcite : + @xmath90 @xmath91 is invariant with respect to a random dynamical system @xmath92 , i.e. @xmath93 @xmath94 @xmath95 is globally lipschitz in @xmath96 for all @xmath97 and for any @xmath98 the mapping @xmath99 is a random variable . + @xmath100 the distance of @xmath101 and @xmath102 tends to @xmath103 with exponential rate , for @xmath104 , as @xmath105 tends to infinite . a random slow manifold @xmath91 , which is lower dimensional , retains the long time dynamics of the original system , when @xmath106 is sufficiently small @xcite . in @xcite , a random hadamard graph transform was used to prove the existence of a random inertial manifolds , here we use lyapunov- perron method to achieve our result as in @xcite . assume that @xmath42 and @xmath51 hold and that there exists a @xmath81 such that @xmath82 . then , for sufficiently small @xmath106 , there exists a random slow manifold @xmath107 for the random slow - fast system . this proof is adapted from @xcite for our finite dimensional setting . for completeness , we include the essential part here . for a @xmath108 , we use the banach space @xmath109 as defined in the beginning of this section . + denote a nonlinear mapping @xmath110 note that @xmath111 is well - defined from @xmath112 . we will show that for every initial data @xmath113 , have a unique solution in @xmath109 . for @xmath114 , @xmath115 , we have that + @xmath116\,ds\big\|_{\mathbb{r}^n}\\{}&&+\sup\limits_{t\leq 0}e^{\lambda t}\big\|\frac{1}{\varepsilon}\int_{-\infty}^te^{b\frac{t - s}{\varepsilon}}\big[g(x(s ) , y(s)+\sigma\eta^\varepsilon(\theta_s\omega))-g(\bar{x}(s ) , \bar{y}(s)+\sigma\eta^\varepsilon(\theta_s\omega))\big]\,ds\big\|_{\mathbb{r}^m}\\ & \leq & \big\|(x , y)-(\bar{x } , \bar{y})\big\|_{c_\lambda } \sup\limits_{t\leq 0}\big\{e^{(\lambda+\alpha ) t}kl_f\int_t^0e^{-(\lambda+\alpha ) s}\,ds\big\}\\{}&&+\big\|(x , y)-(\bar{x } , \bar{y})\big\|_{c_\lambda } \sup\limits_{t\leq 0}\big\{\frac{1}{\varepsilon}e^{(\lambda-\frac{\beta}{\varepsilon } ) t}kl_g\int_{-\infty}^te^{(\frac{\beta}{\varepsilon}-\lambda)s}\,ds\big\}\\ & \leq & \big(\frac{kl_f}{\alpha+\lambda } + \frac{kl_g}{\beta-\varepsilon\lambda}\big)\big\|(x , y)-(\bar{x } , \bar{y})\big\|_{c_\lambda } .\end{aligned}\ ] ] the first inequality is by @xmath42 and the lipschitz continuity of @xmath14 and @xmath15 , while the second inequality comes from direct calculation . taking @xmath117 , which satisfies @xmath118 , we conclude that @xmath119 by the assumption @xmath51 , @xmath120 , and @xmath121 as @xmath122 . therefore , for @xmath106 small enough , @xmath123 . the contraction map theorem implies that for every @xmath124 , @xmath125 has a fixed point @xmath126 which is the unique solution of the differential equation system . moreover , the fixed point has the property @xmath127 denoting @xmath128 , we obtain @xmath129 with the help of inequality , we further have @xmath130 thus , @xmath131 is lipschitz continuous . by the fact that @xmath132 if and only if there exists @xmath133 and satisfies and , it follows that @xmath134 if and only if there exists @xmath135 such that @xmath136 therefore , there exists a random slow manifold @xmath137 by the random transformation and noting that @xmath74 and @xmath73 have the same distribution , the dynamics on the slow manifold is now described by the following dimension - reduced system in @xmath138 ( from equation ( [ sde ] ) ) , for @xmath106 sufficiently small : @xmath139 we now approximate the slow manifolds for sufficiently small @xmath106 . expand the solution of system as @xmath140 and the initial conditions as @xmath141 and @xmath142 . with the help of and , we have the expansions @xmath143\,ds\\{}&&+f_y(\xi , y_0(\tau)+\sigma\eta(\theta_\tau\psi_\varepsilon\omega))\cdot\big[\varepsilon y_1(\tau)+\cdots\big]+\cdots\\&=&f(\xi , y_0(\tau)+\sigma\eta(\theta_\tau\psi_\varepsilon\omega))\\{}&&+f_x(\xi , y_0(\tau)+\sigma\eta(\theta_\tau\psi_\varepsilon\omega))\cdot \varepsilon\int_0^\tau\big[a \xi+f(\xi , y_0(s)+\sigma\eta(\theta_s\psi_\varepsilon\omega))\big]\,ds\\{}&&+f_y(\xi , y_0(\tau)+\sigma\eta(\theta_\tau\psi_\varepsilon\omega))\cdot\varepsilon y_1(\tau)+\cdots,\\ g(x(\tau\varepsilon ) , y(\tau\varepsilon)+\sigma\eta(\theta_\tau\psi_\varepsilon\omega))&=&g(\xi , y_0(\tau)+\sigma\eta(\theta_\tau\psi_\varepsilon\omega))\\{}&&+g_x(\xi , y_0(\tau)+\sigma\eta(\theta_\tau\psi_\varepsilon\omega))\cdot \varepsilon\int_0^\tau\big[a \xi+f(\xi , y_0(s)+\sigma\eta(\theta_s\psi_\varepsilon\omega))\big]\,ds\\{}&&+g_y(\xi , y_0(\tau)+\sigma\eta(\theta_\tau\psi_\varepsilon\omega))\cdot\varepsilon y_1(\tau)+\cdots.\end{aligned}\ ] ] inserting into , expanding and then matching the terms of the same power of @xmath106 , we get @xmath144 and @xmath145y_1(\tau)\\ \quad\quad\quad\quad+g_x(\xi , y_0(\tau)+\sigma\eta(\theta_\tau\psi_\varepsilon\omega))\big\{a\tau\xi+\int_0^\tau f(\xi , y_0(s)+\sigma\eta(\theta_s\psi_\varepsilon\omega))\,ds\big\ } , \\ y_1(0)=\tilde h^{(1)}(\xi , \omega ) . \end{cases}\end{aligned}\ ] ] solving the two equations for @xmath146 and @xmath147 , we obtain @xmath148 and @xmath149\,ds.\nonumber\\\end{aligned}\ ] ] with the help of and , the expression can be calculated as follows @xmath150 + g_y(\xi , y_0(s)+\sigma\eta(\theta_s\psi_\varepsilon\omega))\varepsilon y_1(s)\big\}\,ds+\mathcal{o}(\varepsilon^2)\\ & = & \int_{-\infty}^0 e^{-bs}g(\xi , y_0(s)+\sigma\eta(\theta_s\psi_\varepsilon\omega))\,ds \\{}&&+ \varepsilon\int_{-\infty}^0 e^{-bs}\big\{g_x(\xi , y_0(s)+\sigma\eta(\theta_s\psi_\varepsilon\omega))\big[as\xi+\int_0^s\big(f(\xi , y_0(r)+\sigma\eta(\theta_r\psi_\varepsilon\omega))\big)\,dr\big]\\{}&&\quad\quad\quad\quad\quad\quad + g_y(\xi , y_0(s)+\sigma\eta(\theta_s\psi_\varepsilon\omega))y_1(s)\big\}\,ds+\mathcal{o}(\varepsilon^2).\end{aligned}\ ] ] to get the second equation , we used @xmath151 and then used @xmath152 to replace @xmath153 . thus the zero and first order terms in @xmath106 , of @xmath131 in the random slow manifold @xmath154 for , are respectively @xmath155 and @xmath156\\{}&&\quad\quad\quad\quad\quad\quad + g_y(\xi , y_0(s)+\sigma\eta(\theta_s\psi_\varepsilon\omega))y_1(s)\big\}\,ds.\end{aligned}\ ] ] that is , the slow manifolds @xmath157 of up to the order @xmath158 is represented by @xmath159 . this produces an approximation of the random slow manifold . therefore , we have the following result . [ slow999 ] assume that @xmath42 and @xmath51 hold and assume that there is a @xmath81 such that @xmath82 . then , for sufficiently small @xmath106 , there exists a slow manifold @xmath160 where @xmath161 with @xmath162 and @xmath163 expressed in and , respectively . with the approximated random slow manifold @xmath164 we obtain the following dimension - reduced approximate random system in @xmath138 ( from equation ( [ slowdynamics ] ) ) , for @xmath106 sufficiently small : @xmath165 for the motion of aerosol particles in a cellular flow field , stommel once observed that , ignoring particle inertial ( @xmath166 ) , some particles follow closed paths and are permanently suspended in the flow . rubin , jones and maxey @xcite showed that any small amount inertial ( small @xmath167 ) will cause almost all particles to settle . via a singular perturbation theory @xcite , jones showed the existence of an attracting slow manifold . by analyzing the equations of motion on the slow manifold , especially heteroclinic orbits , they established the presence of mechanisms that inhibit trapping and enhance settling of particles . consider a model for the motion of aerosol particles in a cellular flow field , under random environmental influences @xcite @xmath168 where @xmath169 and @xmath170 are position and velocity , respectively , of a particle in the horizontal- vertical plane ( positive @xmath171 axis points to the settling / gravitational direction ) , @xmath172 is a velocity scale , and @xmath173 is the settling velocity in still fluid . moreover , @xmath174 are independent scalar wiener processes , @xmath10 is a positive parameter , and @xmath106 is the inertial response time scale of the particle . note that @xmath175 is the so - called cellular flow field velocity components ( horizontal and vertical ) on the domain ( a ` cell ' ) @xmath176 . as in section [ slow888 ] , this four dimensional sde system can be converted to the following rde system @xmath177 where @xmath178 denoting @xmath179 and @xmath180 , and we examine the motion of the particle @xmath181 . by using and , we get @xmath182 and @xmath183 owing to , @xmath184\,ds\\{}&= & -a^2 \sin \xi_1 \cos \xi_1 + a v \sin \xi_1 \sin \xi_2 + a \sigma \cos \xi_1 \cos \xi_2 \int_{-\infty}^0 s e^s \,dw_s^1 \\ & & { } -a \sigma \sin \xi_1 \sin \xi_2 \int_{-\infty}^0 s e^s \,dw_s^2,\\ \tilde h_2^{(1)}(\xi_1,\xi_2,\omega ) & = & -a^2 \sin \xi_2 \cos \xi_2 + a v \cos \xi_1 \cos \xi_2 + a \sigma \sin \xi_1 \sin \xi_2 \int_{-\infty}^0 s e^s \,dw_s^1 \\ & & { } -a \sigma\cos \xi_1 \cos \xi_2 \int_{-\infty}^0 s e^s \,dw_s^2.\end{aligned}\ ] ] therefore , from ( [ slowdynamics ] ) , the dynamics on the random slow manifold is described by the following dimension - reduced system : @xmath185 @xmath186 note that @xmath181 is the particle position . for random slow manifold reduction , it is customary to use a notation different from the original one @xmath169 . the positive @xmath187 direction points toward the bottom of the fluid . in this section , we conduct numerical simulations for this reduced or slow system - . when @xmath188 , @xmath189 , this reduced system becomes the classical system for the motion of particles in the cellular flow . when @xmath190 , @xmath189 indicates no noise , while a non - zero @xmath10 means noise is present . motivated by understanding the settling of particles as in @xcite , we first calculate first exit time of particles , described by the random system - , from the domain @xmath191 and then examine how particles , exit or escape the fluid domain @xmath192 . to this end , we introduce two concepts : first exit time and escape probability . the first exit time is the time when a particle , initially at @xmath193 , first exits the domain @xmath192 : @xmath194 let @xmath195 be a subboundary . the escape probability @xmath196 , for a particle initially at @xmath181 , through a subboundary @xmath197 , is the likelihood that this particle first escapes the domain @xmath192 by passing through @xmath197 . we will take @xmath197 to be one of the four sides of the fluid domain @xmath192 . the escape probability of a particle through the top side @xmath198 means the likelihood that this particle settles directly to the bottom of the fluid ( note that the positive @xmath187 direction points to the bottom of the fluid ) . to compute the first exit time from the domain @xmath192 , we place particles on a lattice of grid points in @xmath192 and on its boundary , and set a large enough threshold time @xmath199 . as soon as a particle reaches boundary of @xmath192 , it is regarded as ` having exited ' from @xmath192 . if a particle leaves @xmath192 before @xmath199 , then the time of leaving is taken as the first exit time , but if it is still in the domain at time @xmath199 , we take @xmath199 as the first exit time . when a particle s first exit time is @xmath199 , we can see it as trapped in the cell . in order to calculate the escape probability of a particle under noise through a subboundary @xmath197 , one of the four sides of the domain , we calculate a large number , @xmath200 , of paths for each particle to see how many ( say @xmath201 ) of them exit through the subboundary @xmath197 , and then we get the escape probability @xmath202 . we do this for particles placed on a lattice of grid points in @xmath192 and on its boundary . when a particle reaches or is on a side subboundary , it is regarded as ` having escaped through ' that part of the boundary . , @xmath203 and @xmath189 ( no noise ) : @xmath188 ( top ) ; @xmath204 ( bottom).,title="fig:",height=226 ] + , @xmath203 and @xmath189 ( no noise ) : @xmath188 ( top ) ; @xmath204 ( bottom).,title="fig:",height=226 ] and @xmath203 : @xmath188 and @xmath189 ( top , no noise ) ; @xmath204 and @xmath189 ( middle , no noise ) ; @xmath204 and @xmath205 ( bottom , with noise).,title="fig:",height=226 ] + and @xmath203 : @xmath188 and @xmath189 ( top , no noise ) ; @xmath204 and @xmath189 ( middle , no noise ) ; @xmath204 and @xmath205 ( bottom , with noise).,title="fig:",height=226 ] + and @xmath203 : @xmath188 and @xmath189 ( top , no noise ) ; @xmath204 and @xmath189 ( middle , no noise ) ; @xmath204 and @xmath205 ( bottom , with noise).,title="fig:",height=226 ] ( a ) , @xmath203 and @xmath205 : ( a ) @xmath206 , ( b ) @xmath207 , ( c ) @xmath208 and ( d ) @xmath209 . , title="fig:",height=226 ] ( b ) , @xmath203 and @xmath205 : ( a ) @xmath206 , ( b ) @xmath207 , ( c ) @xmath208 and ( d ) @xmath209 . , title="fig:",height=226 ] + ( c ) , @xmath203 and @xmath205 : ( a ) @xmath206 , ( b ) @xmath207 , ( c ) @xmath208 and ( d ) @xmath209 . , title="fig:",height=226 ] ( d ) , @xmath203 and @xmath205 : ( a ) @xmath206 , ( b ) @xmath207 , ( c ) @xmath208 and ( d ) @xmath209 . , title="fig:",height=226 ] ( a ) , @xmath204 , @xmath203 and @xmath205 : ( a ) escape through @xmath210 ( settling direction or physical bottom boundary ) , ( b ) escape through @xmath211 ( physical top boundary ) , ( c ) escape through @xmath212 ( left boundary ) , and ( d ) escape through @xmath213 ( right boundary).,title="fig:",height=226 ] ( b ) , @xmath204 , @xmath203 and @xmath205 : ( a ) escape through @xmath210 ( settling direction or physical bottom boundary ) , ( b ) escape through @xmath211 ( physical top boundary ) , ( c ) escape through @xmath212 ( left boundary ) , and ( d ) escape through @xmath213 ( right boundary).,title="fig:",height=226 ] + ( c ) , @xmath204 , @xmath203 and @xmath205 : ( a ) escape through @xmath210 ( settling direction or physical bottom boundary ) , ( b ) escape through @xmath211 ( physical top boundary ) , ( c ) escape through @xmath212 ( left boundary ) , and ( d ) escape through @xmath213 ( right boundary).,title="fig:",height=226 ] ( d ) , @xmath204 , @xmath203 and @xmath205 : ( a ) escape through @xmath210 ( settling direction or physical bottom boundary ) , ( b ) escape through @xmath211 ( physical top boundary ) , ( c ) escape through @xmath212 ( left boundary ) , and ( d ) escape through @xmath213 ( right boundary).,title="fig:",height=226 ] ( a ) , @xmath204 , @xmath203 and @xmath205 : ( a ) escape through @xmath210 ( settling direction or physical bottom boundary ) , ( b ) escape through @xmath211 ( physical top boundary ) , ( c ) escape through @xmath212 ( left boundary ) , and ( d ) escape through @xmath213 ( right boundary).,title="fig:",height=226 ] ( b ) , @xmath204 , @xmath203 and @xmath205 : ( a ) escape through @xmath210 ( settling direction or physical bottom boundary ) , ( b ) escape through @xmath211 ( physical top boundary ) , ( c ) escape through @xmath212 ( left boundary ) , and ( d ) escape through @xmath213 ( right boundary).,title="fig:",height=226 ] + ( c ) , @xmath204 , @xmath203 and @xmath205 : ( a ) escape through @xmath210 ( settling direction or physical bottom boundary ) , ( b ) escape through @xmath211 ( physical top boundary ) , ( c ) escape through @xmath212 ( left boundary ) , and ( d ) escape through @xmath213 ( right boundary).,title="fig:",height=226 ] ( d ) , @xmath204 , @xmath203 and @xmath205 : ( a ) escape through @xmath210 ( settling direction or physical bottom boundary ) , ( b ) escape through @xmath211 ( physical top boundary ) , ( c ) escape through @xmath212 ( left boundary ) , and ( d ) escape through @xmath213 ( right boundary).,title="fig:",height=226 ] ( a1 ) , @xmath204 , @xmath203 and @xmath205 : ( a1 ) and ( a2 ) are figure [ ep - w01](a ) splitting at the deterministic heteroclinic orbit @xmath214 ; ( c1 ) and ( c2 ) are figure [ ep - w01](c ) splitting at the same heteroclinic orbit.,title="fig:",height=226 ] ( a2 ) , @xmath204 , @xmath203 and @xmath205 : ( a1 ) and ( a2 ) are figure [ ep - w01](a ) splitting at the deterministic heteroclinic orbit @xmath214 ; ( c1 ) and ( c2 ) are figure [ ep - w01](c ) splitting at the same heteroclinic orbit.,title="fig:",height=226 ] + ( c1 ) , @xmath204 , @xmath203 and @xmath205 : ( a1 ) and ( a2 ) are figure [ ep - w01](a ) splitting at the deterministic heteroclinic orbit @xmath214 ; ( c1 ) and ( c2 ) are figure [ ep - w01](c ) splitting at the same heteroclinic orbit.,title="fig:",height=226 ] ( c2 ) , @xmath204 , @xmath203 and @xmath205 : ( a1 ) and ( a2 ) are figure [ ep - w01](a ) splitting at the deterministic heteroclinic orbit @xmath214 ; ( c1 ) and ( c2 ) are figure [ ep - w01](c ) splitting at the same heteroclinic orbit.,title="fig:",height=226 ] ( blue solid curve ) and unstable manifold @xmath215 ( blue dashed curve ) for the slow system and : @xmath207 , @xmath204 , @xmath203 , and @xmath189 ( no noise ) . two equilibrium points are @xmath216 and @xmath217.,width=302,height=264 ] when the settling velocity in still fluid is zero , i.e. , @xmath206 ( and also @xmath189 ) , particles are trapped in the cell either in circular motion ( with no inertial , @xmath188 ) or spiralling motion ( with inertial , @xmath11 ) , as shown in figure [ w0 ] ( top ) and ( bottom ) , respectively . when the settling velocity in still fluid is non - zero , i.e. , @xmath220 , in the case with inertial absent ( @xmath188 ) and noise absent ( @xmath189 ) , the particles in the area surrounded by the heteroclinic orbit @xmath221 connecting the equilibrium points @xmath222 and @xmath223 , are trapped inside it , with the equilibrium point @xmath224 as a center . but the particles in the remaining area settle to the bottom of the fluid . with an arbitrarily small inertial effect ( @xmath225 and also @xmath189 ) , the heteroclinic orbit breaks and it leads to the settling of almost all particles , with the equilibrium point @xmath226 becoming an unstable spiral . figure [ stable - unstable manifolds v01 ] is the stable manifold @xmath227 ( blue solid curve ) and unstable manifold @xmath228 ( blue dashed curve ) when inertial presents ( @xmath204 and also @xmath189 ) . for non - zero settling velocity in still fluid ( @xmath230 ) , when noise is absent ( @xmath231 ) , all particles settle to the fluid bottom ; see figure [ w03 ] ( top , middle ) . but when noise is present ( @xmath232 ) , some particles exit the cell not only by settling . figure [ w03 ] ( bottom ) shows that , with small noise ( @xmath205 ) , some particles indeed exit the cell from the vertical side boundary @xmath212 . in fact , when noise is present , all particles will exit , no matter the settling velocity in still fluid @xmath173 is zero or non - zero . figure [ first exit time ] indicates that with noise , particles will all exit from a fluid cell in finite time , almost surely . in the following we only consider the case with noise . figures [ ep - w0 ] [ ep - w01 ] plot the escape probability through four side boundaries , for zero or non - zero @xmath173 ( settling velocity in still fluid ) values . when a particle reaches or is on a side boundary , it is regarded as ` having escaped through ' that part of the boundary . in other words , particles on a side boundary have escape probability @xmath233 ( you see this in these figures ) . when @xmath206 , the particles escape the cell through each of the four side boundaries with similar or equal likelihood ( figure [ ep - w0 ] ) , as there is no preferred direction for particles due to zero settling velocity @xmath206 ( in still fluid ) . with non - zero @xmath173 , particles almost surely do not escape through the right side boundary @xmath213 . in fact , the inertial particles either settle to the physical bottom or exit from the left side boundary @xmath212 . figure [ ep - w01 ] displays the escape probability for @xmath207 , through each of the four side boundaries of the fluid cell . although most particles settle ( figure [ ep - w01 ] ( a ) ) , some particles escape the fluid cell through the left side boundary ( figure [ ep - w01 ] ( c ) ) . see figure [ ep - heter - w01 ] for a split view of this phenomenon . to examine this phenomenon more carefully , we draw the stable manifold @xmath227 and unstable manifold @xmath228 for the deterministic system ( @xmath189 ) in figure [ stable - unstable manifolds v01 ] . as shown in figure [ plot3 + mesh v01 ] , inertial particles with significant likelihood of escaping through the left side boundary are near or on the stable manifold @xmath227 . in other words , some ( but not all ) inertial particles near or on this stable manifold are resistent to settling in the stochastic case . this resistance is quantified by the escape probability for a particle to get out of the fluid cell through the left side boundary . more specifically , the difference between the inertial particle settling times for deterministic case ( @xmath189 ) and a random case ( @xmath205 ) is shown in figure [ difference ] . we observe that the inertial particles near or on the stable manifold @xmath234 could have either a longer or shorter settling time , compared with the deterministic case . this indicates that the noise could either delay or enhance the settling ( although we do not know the reason ) , and the stable manifold is an agent facilitating this behavior . however , the overall impact of noise appears to delay the settling , as the averaged difference over the cell @xmath235 is @xmath236 for noise intensity @xmath205 , while this averaged value is @xmath237 for a stronger noise with @xmath238 . let the settling velocity in still fluid be non - zero ( i.e. , @xmath230 ) . + ( i ) in the classical case ( no inertial : @xmath188 and no noise : @xmath239 ) , the particles surrounded inside a heteroclinic orbit are trapped inside it and all the other particles settle to the bottom of this cellular fluid flow . + ( ii ) in the case with only small inertia influence ( @xmath240 ) , the heteroclinic orbit breaks up to form a stable manifold @xmath241 and an unstable manifold @xmath242 , the trapped particles then settle , i.e. , all inertial particles settle . + ( iii ) however , when the noise is present ( @xmath243 ) , although most inertial particles still settle , some particles near or on the deterministic stable manifold @xmath241 escape the fluid cell through the left side boundary , with non - negligible likelihood . thus , inertial particle motions occur in two adjacent fluid cells in random cases , but confine in single cells in the deterministic case . in fact , noise could either delay settling for some particles or enhance settling for others , and the deterministic stable manifold is an agent to facilitate this phenomenon . overall , noise appears to delay the settling in an averaged sense .	a method is provided for approximating random slow manifolds of a class of slow - fast stochastic dynamical systems . thus approximate , low dimensional , reduced slow systems are obtained analytically in the case of sufficiently large time scale separation . to illustrate this dimension reduction procedure , the impact of random environmental fluctuations on the settling motion of inertial particles in a cellular flow field is examined . it is found that noise delays settling for some particles but enhances settling for others . a deterministic stable manifold is an agent to facilitate this phenomenon . overall , noise appears to delay the settling in an averaged sense . * key words : * random slow manifolds , dimension reduction , stochastic differential equations ( sdes ) , approximation under big scale - separation , inertial particles in flows * mathematics subject classifications ( 2010 ) * : 37h10 , 37m99 , 60h10
the birnbaum saunders ( @xmath1 ) distribution has received considerable attention in the last few years . it was proposed by @xcite and is also known as the fatigue life distribution . it describes the total time until the damage caused by the development and growth of a dominant crack reaches a threshold level and causes a failure . the random variable @xmath2 is said to have a @xmath1 distribution with parameters @xmath3 and @xmath4 , denoted by @xmath5 , if its probability density function is given by @xmath6 where @xmath7 , @xmath8 , @xmath9 ( shape parameter ) and @xmath10 ( scale parameter ) . it is positively skewed , the skewness decreasing with @xmath3 . for any constant @xmath11 , it follows that @xmath12 . it is also noteworthy that the reciprocal property holds : @xmath13 , which is in the same family of distributions @xcite . there are several recent articles considering the @xmath1 distribution ; see for example , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , among others . @xcite introduced a log - linear regression model based on the @xmath1 distribution by showing that if @xmath5 , then @xmath14 has a sinh - normal distribution with shape , location and scale parameters given by @xmath3 , @xmath15 and @xmath16 , respectively , say @xmath17 . the regression model proposed by the authors is given by @xmath18 where @xmath19 is the logarithm of the @xmath20th observed lifetime , @xmath21 contains the @xmath20th observation on @xmath22 covariates ( @xmath23 ) , @xmath24 is a vector of unknown regression parameters , and @xmath25 . diagnostic tools for the @xmath1 regression model can be found in @xcite , @xcite and @xcite . in the @xmath1 regression model hypothesis testing inference is usually performed using the likelihood ratio , rao score and wald tests . a new criterion for testing hypothesis , referred to as the _ gradient test _ , has been proposed by @xcite . its statistic shares the same first order asymptotic properties with the likelihood ratio , wald and score statistics and is very simple when compared with the other three classic tests . in fact , @xcite wrote : `` the suggestion by terrell is attractive as it is simple to compute . it would be of interest to investigate the performance of the [ gradient ] statistic . '' to the best of our knowledge , however , there is no mention in the statistical literature on the use of the gradient test in @xmath1 regressions . in this paper we compare the four rival tests from two different points of view . first , we invoke asymptotic arguments . we then move to a finite - sample comparison , which is accomplished by means of a simulation study . our principal aim is to help practitioners to choose among the different criteria when performing inference in @xmath1 regressions . on asymptotic grounds , it is known that , to the first order of approximation , the likelihood ratio , wald , score and gradient statistics have the same asymptotic distributional properties either under the null hypothesis or under a sequence of local alternatives , i.e. a sequence of pitman alternatives converging to the null hypothesis at a convergence rate @xmath0 . on the other hand , up to an error of order @xmath26 the corresponding criteria have the same size properties but their local powers differ in the @xmath0 term . a meaningful comparison among the criteria can be performed by comparing the nonnull asymptotic expansions to order @xmath0 of the distribution functions of these statistics under a sequence of pitman alternatives . in this regard , we can benefit from the work by @xcite , @xcite and @xcite . @xcite derived the nonnull asymptotic expansions up to order @xmath0 for the densities of the likelihood ratio and wald statistics , while an analogous result for the score statistic was obtained by @xcite . recently , the asymptotic expansion up to order @xmath0 for the density of the gradient statistic was derived by @xcite . the expansions obtained by these authors are extremely general but it can be very difficult or even impossible to particularize their formulas for specific regression models . as we shall see below , we have been able to apply their results for the @xmath1 regression model . the rest of the paper is organized as follows . section [ tests ] briefly describes the likelihood ratio , wald , score and gradient tests . in section [ inference_bs ] these tests are applied for testing hypotheses on the parameters of the @xmath1 regression model . in section [ main_result ] we obtain and compare the local powers of the tests . monte carlo simulation results on the finite - sample performance of the tests are presented and discussed in section [ mcsimulation ] . section [ application ] contains an application to a real fatigue data set . finally , section [ conclusions ] discusses our main findings and closes the paper with some conclusions . consider a parametric model @xmath27 with corresponding log - likelihood function @xmath28 , where @xmath29 is a @xmath30-vector of unknown parameters . the dimensions of @xmath31 and @xmath32 are @xmath30 and @xmath33 , respectively . suppose the interest lies in testing the composite null hypothesis @xmath34 against @xmath35 , where @xmath36 is a specified vector . hence , @xmath31 is a vector of nuisance parameters . let @xmath37 and @xmath38 denote the score function and the fisher information matrix for @xmath39 , respectively . the partition for @xmath39 induces the corresponding partitions @xmath40 where @xmath41 is the inverse of @xmath38 . let @xmath42 and @xmath43 denote the maximum likelihood estimators of @xmath29 under @xmath44 and @xmath45 , respectively . the likelihood ratio ( @xmath46 ) , wald ( @xmath47 ) , score ( @xmath48 ) and gradient ( @xmath49 ) statistics for testing @xmath45 versus @xmath44 are given by @xmath50 @xmath51 respectively , where @xmath52 , @xmath53 and @xmath54 . the limiting distribution of @xmath46 , @xmath47 , @xmath48 and @xmath49 is @xmath55 under @xmath45 and @xmath56 , i.e. a noncentral chi - square distribution with @xmath33 degrees of freedom and an appropriate noncentrality parameter @xmath57 , under @xmath44 . the null hypothesis is rejected for a given nominal level , @xmath58 say , if the test statistic exceeds the upper @xmath59 quantile of the @xmath60 distribution . clearly , @xmath61 has a very simple form and does not involve knowledge of the information matrix , neither expected nor observed , and any matrix , unlike @xmath62 and @xmath63 . in what follows , we shall consider the tests which are based on the statistics @xmath46 , @xmath47 , @xmath48 and @xmath49 in the class of @xmath1 regression models for testing a composite null hypothesis . the log - likelihood function @xmath28 for the vector parameter @xmath64 from a random sample @xmath65 obtained from model ( [ eq1 ] ) , except for constants , can be written as @xmath66 where @xmath67/2)$ ] , @xmath68/2)$ ] and @xmath69 , for @xmath70 . it is assumed that the model matrix @xmath71 has full column rank , i.e. , rank@xmath72 . the score function and the fisher information matrix for @xmath64 are , respectively , given by @xmath73 where @xmath74 , @xmath75 , @xmath76 , @xmath77 , @xmath78 with @xmath79 and @xmath80 , @xmath81 denoting the _ error function _ : @xmath82 ( see , for instance , * ? ? ? * ) . from the block - diagonal form of @xmath38 we have that @xmath83 and @xmath3 are globally orthogonal @xcite . the hypothesis of interest is @xmath84 , which will be tested against the alternative hypothesis @xmath85 , where @xmath83 is partitioned as @xmath86 , with @xmath87 and @xmath88 . here , @xmath89 is a fixed column vector of dimension @xmath90 . the partition for @xmath83 induces the corresponding partitions @xmath91 , with @xmath92 and @xmath93 , @xmath94 with the matrix @xmath95 partitioned as @xmath96 . the likelihood ratio , wald , score and gradient statistics for testing @xmath45 can be expressed , respectively , as @xmath97 @xmath98 where @xmath99 , @xmath100 , @xmath101 and @xmath102 . the limiting distribution of all these statistics under @xmath45 is @xmath103 . notice that , unlike the wald and score statistics , the gradient statistic does note involve the error function . now , the problem under consideration is that of testing a composite null hypothesis @xmath104 against @xmath105 , where @xmath106 is a positive specified value for @xmath3 , and @xmath83 acts as a nuisance parameter . the four statistics are expressed as follows : @xmath107 @xmath108 where @xmath109 , with @xmath110 and @xmath111 representing the unrestricted and restricted maximum likelihood estimators of @xmath39 under @xmath44 and @xmath45 , respectively . in this section we shall assume the following local alternative hypothesis @xmath112 , where @xmath113 with @xmath114 for @xmath115 . we follow the notation in @xcite . let @xmath116 where @xmath117 it then follows that @xmath118 , @xmath119 , for @xmath120 , and @xmath121 , where @xmath122 and @xmath123 are the @xmath124 elements of the matrices @xmath125 and @xmath126 , respectively , and @xmath127 and @xmath128 are the @xmath129 elements of the matrices @xmath125 and @xmath126 , respectively . additionally , let @xmath130 where @xmath131 is a @xmath132 identity matrix . the nonnull distributions of the statistics @xmath46 , @xmath47 , @xmath48 and @xmath49 under pitman alternatives for testing @xmath133 in the @xmath1 regression model can be expressed as @xmath134 where @xmath135 is the cumulative distribution function of a non - central chi - square variate with @xmath136 degrees of freedom and non - centrality parameter @xmath57 . here , @xmath137 and the coefficients @xmath138 s ( @xmath139 and @xmath140 ) can be written as @xmath141 @xmath142 @xmath143 @xmath144 @xmath145 @xmath146 @xmath147 @xmath148 where the @xmath149 s are defined as @xmath150 , @xmath151 , @xmath152 , @xmath153 , etc . , and @xmath154 is the ( @xmath155)th element of inverse of @xmath156 . the coefficients @xmath157 are obtained from @xmath158 , for @xmath139 . after some algebra , it is possible to show that , in the @xmath1 regression model , @xmath159 therefore , @xmath160 for @xmath139 and @xmath140 , and we can write @xmath161 this is a very interesting result , which implies that the likelihood ratio , score , wald and gradient tests for testing the composite null hypothesis @xmath133 have exactly the same local power up to an error of order @xmath26 . we now turn to the problem of testing hypotheses on @xmath3 , the shape parameter . the nonnull asymptotic distributions of the statistics @xmath46 , @xmath47 , @xmath48 and @xmath49 for testing @xmath104 under the local alternative @xmath162 , where @xmath163 is assumed to be @xmath164 , is @xmath165 with @xmath166 . the @xmath138 s for the test of @xmath104 are easy to obtain and are given by @xmath167 , @xmath168 , @xmath169 , @xmath170 , @xmath171 , @xmath172 , @xmath173 , @xmath174 , @xmath175 , @xmath176 , @xmath177 and @xmath178 , where @xmath179 . the coefficients @xmath157 are obtained from @xmath158 , for @xmath139 . we have that @xmath180 for @xmath181 . after some algebra the coefficients @xmath138 reduce to @xmath182 @xmath183 @xmath184 it should be noticed that the above expressions depend on the model only through @xmath3 and the rank of the model matrix @xmath95 ; they do not involve the unknown parameter @xmath83 . we now present an analytical comparison among the local powers of the four tests for testing the null hypothesis @xmath104 . let @xmath185 be the power function , up to order @xmath0 , of the test that uses the statistic @xmath186 , for @xmath139 . we have @xmath187 for @xmath188 . it is well known that @xmath189 where @xmath190 is the probability density function of a non - central chi - square random variable with @xmath191 degrees of freedom and non - centrality parameter @xmath57 . from ( [ diff_power ] ) and ( [ diff_g ] ) , we have @xmath192 @xmath193 hence , we arrive at the following inequalities : @xmath194 if @xmath195 , and @xmath196 if @xmath197 . the most important finding obtained so far is that the likelihood ratio , score , wald and gradient tests for testing the null hypothesis @xmath133 share the same null size and local power up to an error of order @xmath26 . to this order of approximation the null distribution of the four statistics is @xmath198 . therefore , if the sample size is large , type i error probabilities of all the tests do not significantly deviate from the true nominal level , and their powers are approximately equal for alternatives that are close to the null hypothesis . the natural question now is how these tests perform when the sample size is small or of moderate size , and which one is the most reliable . in the next section , we shall use monte carlo simulations to put some light on this issue . in this section we shall present the results of a monte carlo simulation in which we evaluate the finite sample performance of the likelihood ratio , wald , score and gradient tests . the simulations were based on the model @xmath199 where @xmath200 and @xmath25 , @xmath201 . the covariate values were selected as random draws from the uniform @xmath202 distribution and for fixed @xmath203 those values were kept constant throughout the experiment . the number of monte carlo replications was 15,000 , the nominal levels of the tests were @xmath58 = 10% , 5% and 1% , and all simulations were performed using the ox matrix programming language @xcite . ox is freely distributed for academic purposes and available at http://www.doornik.com . all log - likelihood maximizations with respect to @xmath83 and @xmath3 were carried out using the bfgs quasi - newton method with analytic first derivatives through maxbfgs subroutine . this method is generally regarded as the best - performing nonlinear optimization method @xcite . the initial values in the iterative bfgs scheme were @xmath204 for @xmath83 and @xmath205 for @xmath3 , where @xmath206 first , the null hypothesis is @xmath207 , which is tested against a two - sided alternative . the sample size is @xmath208 , @xmath209 and @xmath210 . the values of the response were generated using @xmath211 . the null rejection rates of the four tests are presented in table [ tab1 ] . it is evident that the likelihood ratio ( @xmath46 ) and wald ( @xmath47 ) tests are markedly liberal , more so as the number of regressors increases . the score ( @xmath48 ) and gradient ( @xmath49 ) tests are also liberal in most of the cases , but much less size distorted than the likelihood ratio and wald tests in all cases . for instance , when @xmath212 , @xmath213 and @xmath214 , the rejection rates are 10.12% ( @xmath46 ) , 12.77% ( @xmath47 ) , 7.12% ( @xmath48 ) and 7.32% ( @xmath49 ) . it is noticeable that the score test is much less liberal than the likelihood ratio and wald tests and slightly less liberal than the gradient test . the score and gradient tests are slightly conservative in some cases . additionally , the wald test is much more liberal than the other tests . similar results hold for @xmath215 . table [ tab2 ] reports results for @xmath216 and @xmath217 and sample sizes ranging from 15 to 200 . as expected , the null rejection rates of all the tests approach the corresponding nominal levels as the sample size grows . again , the score and gradient tests present the best performances . .null rejection rates ( % ) ; @xmath3 = 0.5 and 1.0 , with @xmath218 . [ cols="^,^,^,^,^,^,^,^,^,^,^,^,^ " , ] we now turn to the finite - sample power properties of the four tests . the simulation results above show that the tests have different sizes when one uses their asymptotic @xmath219 distribution in small and moderate - sized samples . in evaluating the power of these tests , it is important to ensure that they all have the correct size under the null hypothesis . to overcome this difficulty , we used 500,000 monte carlo simulated samples , drawn under the null hypothesis , to estimate the exact critical value of each test for the chosen nominal level . we set @xmath218 , @xmath220 , @xmath216 and @xmath221 . for the power simulations we computed the rejection rates under the alternative hypothesis @xmath222 , for @xmath223 ranging from @xmath224 to @xmath225 . figure [ figpower ] shows that the power curves of the four tests are indistinguishable from each other . as expected , the powers of the tests approach 1 as @xmath226 grows . power simulations carried out for other values of @xmath203 , @xmath22 and @xmath58 showed a similar pattern . , @xmath220 , @xmath216 and @xmath221.,width=377,height=283 ] overall , in small to moderate - sized samples the best performing tests are the score and the gradient tests . they are less size distorted than the other two and are as powerful as the others . we also performed monte carlo simulations considering hypothesis testing on @xmath3 . to save space , the results are not shown . the score and gradient tests exhibited superior behavior than the likelihood ratio and wald tests . for example , when @xmath227 , @xmath228 , @xmath214 and @xmath229 , we obtained the following null rejection rates : 9.99% ( @xmath46 ) , 15.89% ( @xmath47 ) , 5.29% ( @xmath48 ) and 6.99% ( @xmath49 ) . again , the best performing tests are the score and gradient tests . this application focuses on modeling the die lifetime ( @xmath2 ) in the process of metal extrusion . the data were taken from @xcite . according to the authors , `` the estimation of tool life ( fatigue life ) in the extrusion operation is important for scheduling tool changing times , for adaptive process control and for tool cost evaluation . '' the authors noted that `` die fatigue cracks are caused by the repeat application of loads which individually would be too small to cause failure . '' the @xmath1 regression model is then appealing in this context since the main motivation for the @xmath1 distribution is the fatigue failure time due to propagation of an initial crack . we consider the following regression model : @xmath230 where @xmath231 , @xmath25 , @xmath232 and the covariates are @xmath233 ( friction coefficient ) , @xmath234 ( angle of the die ) and @xmath235 ( work temperature ) , for @xmath236 . we wish to test the significance of the interaction effects , i.e. , the interest lies in testing @xmath237 . the likelihood ratio , wald , score and gradient test statistics equal @xmath238 ( @xmath22-value : 0.094 ) , @xmath239 ( @xmath22-value : 0.045 ) , @xmath240 ( @xmath22-value : 0.162 ) and @xmath241 ( @xmath22-value : 0.157 ) , respectively . hence , the null hypothesis is rejected at the 10% nominal level when inference is based on the likelihood ratio or the wald test , but the opposite decision is reached when either the score or the gradient test is used . recall that our simulation results indicated that the likelihood ratio and wald tests are markedly liberal in small samples ( here , @xmath242 ) , which leads us to mistrust the inference delivered by the likelihood ratio and wald tests . therefore , we removed the interaction effects from the model as indicated by the score and gradient tests . the model containing only main effects is @xmath243 for @xmath244 . the null hypothesis @xmath245 is strongly rejected by the four tests at the usual significance levels . all tests also suggest the individual and joint exclusions of the friction coefficient and angle of the die from the model . hence , we end up with the regression model @xmath246 , for @xmath236 . the maximum likelihood estimates of the parameters are ( standard errors in parentheses ) : @xmath247 , @xmath248 and @xmath249 . the @xmath1 regression model is becoming increasingly popular in lifetime analyses and reliability studies . in this paper , we dealt with the issue of performing hypothesis testing concerning the parameters of this model . we considered the three classic tests , likelihood ratio , wald and score tests , and a recently proposed test , the gradient test . for the discussion that follows , let us concentrate on tests regarding the regression parameters , which are , in general , of primary interest . the four tests have the same distribution , under either the null hypothesis or a sequence of local alternatives , up to an error of order @xmath26 , as we showed . our monte carlo simulation study added some important information . it revealed that the likelihood ratio and the wald tests can be remarkably oversized if the sample is small . the score and the gradient tests are clearly much less size distorted than the other two tests . our power simulations suggested that all the four tests have similar power properties when estimated correct critical values are used . overall , this is an indication that the score and the gradient tests should be prefered . at this point , a discussion on small - sample corrections for the classic tests is in order . a bartlett correction for the likelihood ratio statistic and a bartlett - type correction for the score statistic were derived in @xcite and @xcite , respectively ; see also cordeiro and ferrari ( 1991 ) . the corrected statistics have the following interesting properties : ( i ) the uncorrected and corrected statistics have the same asymptotic distribution under the null hypothesis ; ( ii ) the order of the error of the approximation for the distribution of the test statistics by @xmath219 is smaller for the corrected statistics than for the uncorrected statistics ; ( iii ) the corrections have no effect on the @xmath0 term of the local power of the corresponding tests ; ( iv ) simulation results in @xcite and @xcite show that these corrections reduce the size distortion of the tests and that the best performing test in small and moderate - sized samples is the test which uses the corrected score statistic . therefore , the bartlett - type - corrected score test is an excellent alternative to the tests under consideration in the present paper . the slight disadvantage of such an alternative is the extra computational burden involved in computing the bartlett - type correction . we computed the corrected versions of likelihood ratio and score statistics for the hypotheses tested in the real data application presented in section [ application ] . recall that the hypothesis of no interaction effects is rejected by the likelihood ratio and wald tests , but not rejected by the score and gradient tests . it is noteworthy that the decision reached by either the corrected likelihood ratio test or the corrected score test is in agreement with that obtained by the later two tests and in disagreement with the likelihood ratio and wald tests , which tend to reject the null hypothesis much more often than indicated by the significance level . finally , our overall recommendations for practitioners when performing testing inference in @xmath1 regressions are as follows . the score test or the gradient test should be prefered as both perform better than the likelihood ratio and wald tests in small and moderate - sized samples . while the gradient test is a little more liberal than the score test , it is easier to calculate . the bartlett - type corrected score test is a further better option although it requires a small extra computational effort . we gratefully acknowledge grants from fapesp and cnpq ( brazil ) . balakrishnan , n. , leiva , v. , lpez , j. ( 2007 ) . acceptance sampling plans from truncated life tests from generalized birnbaum saunders distribution . _ communications in statistics simulation and computation _ * 36 * , 643656 . leiva , v. , sanhueza , a. , angulo , j.m . a length - biased version of the birnbaum saunders distribution with application in water quality . _ stochastic environmental research and risk assessment _ * 23 * , 299307 . rao , c.r . score test : historical review and recent developments . in _ advances in ranking and selection , multiple comparisons , and reliability _ , n. balakrishnan , n. kannan and h. n. nagaraja , eds . birkhuser , boston .	the birnbaum saunders distribution has been used quite effectively to model times to failure for materials subject to fatigue and for modeling lifetime data . in this paper we obtain asymptotic expansions , up to order @xmath0 and under a sequence of pitman alternatives , for the nonnull distribution functions of the likelihood ratio , wald , score and gradient test statistics in the birnbaum saunders regression model . the asymptotic distributions of all four statistics are obtained for testing a subset of regression parameters and for testing the shape parameter . monte carlo simulation is presented in order to compare the finite - sample performance of these tests . we also present an empirical application . birnbaum saunders distribution , fatigue life distribution , gradient test , lifetime data , likelihood ratio test , local power , score test , wald test .
we denote a hypergraph as @xmath80 where @xmath81 is a set of @xmath82 observables and @xmath83 being a set of contexts of the hypergraph , i.e. the set of subsets of mutually commensurable observables of @xmath17 . a box has an input @xmath84 with cardinality @xmath8 equal to the number of edges of the hypergraph ( number of contexts in a given @xmath37 ) and ( for simplicity we assume ) each output has the same cardinality @xmath85 of dimension equal to multiplication of cardinalities of outputs of @xmath0 which contribute in the corresponding context . the set of such boxes we denote as @xmath86 . we say that a box is _ compatible _ with a hypergraph @xmath37 if it is family of @xmath8 probability distributions such that for each @xmath87 , where @xmath88 is the power of the context @xmath4 , there is a corresponding probability distribution in this family on @xmath89 . we denote it as a family of distributions @xmath90 and @xmath91 . for a given hypergraph @xmath92 , @xmath93 is a _ consistent _ box if for all pairs @xmath94 , and for set of observables @xmath95 there is _ s , _ t pr(s = s , t = t\|x = c ) = _ t pr(s = s , t=t\|x = c ) where @xmath96 and @xmath97 . the set of all consistent boxes compatible with hypergraph @xmath37 that has @xmath8 contexts is denoted as @xmath98 . [ def : consistent ] note , that the well known non - signaling condition is special case of such defined consistency . a _ non - contextual _ box associated with a hypergraph @xmath37 is a consistent box with a property that there exists a common joint probability distribution for all the observables in @xmath17 . the set of all such boxes compatible with @xmath37 , we denote as @xmath99 . all boxes that are consistent but do not satisfy this condition , we call _ contextual_. similarly as in the main text , to specify distributions that belong to box @xmath38 we will denote it as @xmath6 where @xmath4 numbers the contexts running from @xmath100 to @xmath8 . if it is not stated otherwise , in what follows we assume @xmath101 , since for @xmath102 all boxes compatible with any hypergraph @xmath37 , are non - contextual . if a box @xmath19 is non - contextual , we denote it as @xmath103 , and by @xmath7 we will denote the joint probability distribution on @xmath17 ( which exists by definition of non - contextual box ) of which @xmath104 s are appropriate marginals . for short , by @xmath105 for some set of boxes @xmath106 we mean that non - contextual box defined by @xmath7 belongs to @xmath106 where graph @xmath37 with which this box is compatible should be understood from the context . we now make a trivial observation about these boxes : a consistent box on @xmath107 is non - contextual iff it can be written as a convex combination of consistent deterministic boxes , i.e. such that the joint probability distribution of the outputs of all observables @xmath108 equals @xmath109 for some fixed vector @xmath110 . _ proof_. it follows from the definition of noncontextual boxes : the joint probability distribution of all observables @xmath108 is a mixture of the deterministic ones . in this section we present one of the main results of this work - equality of the mutual information of contextuality and the relative entropy of contextuality . in this and the next section , for the sake of proof , we will use also a quantity defined on box @xmath111 as @xmath112 , which is a version of relative entropy of contextuality for fixed distribution @xmath15 . for any box @xmath113 , there holds @xmath114 . [ thm : equivalence ] _ proof_. to show the equality , we introduce another measure of contextuality @xmath115 , i_max(b ) _ \{p(c ) } _ \{g(\|c):g(_c\|c)=g(_c)},p ( ) _ c p(c ) d(g(\|c)\|\|p ( ) ) , [ eq : equiv ] and prove @xmath116 . the proof will not involve optimality of distribution @xmath27 over which in all quantities we take supremum , so we show the equality for @xmath117 , @xmath118 and @xmath119 , from which desired equality follows . we then fix @xmath27 and @xmath120 arbitrarily from now on , and show that @xmath121 . we prove now the first of these equalities . it is easy to see that @xmath122 since relative entropy does not increase under partial trace . to see the converse inequality , consider the optimal classical probability in @xmath123 , call it @xmath124 ( see [ app : not - equal ] for the proof , that such @xmath125 exists ) with marginals @xmath126 , then find a conditional probability distributions @xmath127 such that @xmath128 , where @xmath129 , and define @xmath130 . it is easy to check , that such a choice saturates the inequality @xmath131 giving equality . to see that @xmath132 , we use the following fact : i(_c p(c ) \|cc\|_c ) _ c p(c ) d ( g(\|c ) \|\| _ c p(c ) g(\|c ) ) = _ p ( ) _ cp(c ) d(g(\|c)\|\|p ( ) ) , [ eq : prime ] where @xmath18 has distribution @xmath133 , which is proven in lemma [ lem : prime ] below , stated in more general - quantum case ( where in place of @xmath133 there is a quantum state @xmath134 and minimization is over some states @xmath135 ) . if we set minimization over @xmath133 having marginals @xmath5 of a box b , we get desired equality . summarizing the results we get @xmath136 for arbitrary @xmath27 and @xmath19 , hence taking supremum over this distribution proves @xmath114 for arbitrary consistent box @xmath19 . before proving equality ( [ eq : prime ] ) , we need another result , stated in the lemma below . we need it only for random variables , but we state it for quantum states , since it is valid for quantum states in general , and use the fact that quantum relative entropy and relative entropy distance coincide for classical distributions : for a quantum state @xmath137 with subsystems @xmath138 and @xmath139 _ _ s(\|\|_a_b ) = s(\|\|_a_b ) where @xmath140 ( @xmath141 ) denotes the partial trace over system @xmath142 ( @xmath19 ) , and @xmath106 is quantum relative entropy distance @xcite . [ lem : mutual - prod ] _ proof . _ we first note , that @xmath143 , where @xmath144 and @xmath145 are identity operators on systems @xmath142 and @xmath19 respectively . thus @xmath146 + s(\rho_b ) + [ - s(\rho_b ) - tr\rho_b\log\sigma_b ] = i(\rho ) + s(\rho_a\|\|\sigma_a ) + s(\rho_b\|\|\sigma_b)\end{gathered}\ ] ] where @xmath147 is quantum mutual information @xcite . the last equality proves that @xmath148 because the relative entropy terms @xmath149 and @xmath150 are non - negative , but @xmath151 , hence the equality . we prove now the lemma needed in proof of theorem [ thm : equivalence ] . we state it again for quantum states , since it is valid not only for probability distributions : for arbitrary ensemble of quantum states @xmath152 , there holds i(_c p(c ) \|cc\|_c ) = _ _ c p(c ) s(_c\|\| ) . [ lem : prime ] _ proof_. let us note that lhs can be rewritten as @xmath153 . then , we use the fact that denoting @xmath154 as @xmath137 , by lemma [ lem : mutual - prod ] we have s(\|\|(_c p(c)\|cc\| ) ( _ c p(c ) _ c ) ) = _ _ a,_b s(\|\|_a_b ) . knowing that @xmath155 , i.e. the subsystem of @xmath137 , is the best @xmath156 in the above minimization , we can fix it , having s(\|\|(_c p(c)\|cc\| ) ( _ c p(c ) _ c ) ) = _ s(\|\|(_c p(c)\|cc\| ) ) it is then easy to check that the rhs of above equals just @xmath157 , and the assertion follows . in order to compute @xmath31 for the isotropic xor - boxes and the kcbs box @xcite , we first observe that these boxes have numerous symmetries , i.e. they are invariant under some non - contextuality preserving operations . in this paragraph we specify groups of such operations and a map which applies them at random , called _ twirling_. this leads us to the definition of isotropic boxes and the main result of this section ( theorem [ thm : iso - optimal ] ) which shows that for these boxes it is enough to minimize in the definition of @xmath31 only over non - contextual isotropic boxes . to be more precise , consider any hypergraph @xmath37 with @xmath8 contexts and a box @xmath38 . a non - contextuality preserving operation satisfying @xmath158 we call _ non - contextuality preserving automorphism of @xmath19_. for any finite set of non - contextuality preserving automorphisms @xmath159 , if the group generated by the set @xmath159 ( denoted as @xmath160 ) is finite of order @xmath161 , then the map defined on @xmath19 as b _ _ 1\|g_\| l(b ) , we call _ b-@xmath159-twirling _ and denote as @xmath162 . the image of the set of all boxes through b-@xmath159-twirling we call the set of _ b-@xmath159-isotropic states _ : ^_b : = \ { dc^(n)_g : _ f c^(n)_g d = ^_b(f ) } . note , that there may be different twirlings depending on the set of generators @xmath159 of @xmath160 . however , when the results are true for any fixed choice of @xmath159 , or the set @xmath159 is known from the context , we will omit it in notation , denoting the introduced objects as b - twirling ( @xmath163 ) , and a set of b - isotropic boxes ( @xmath164 ) . we observe that to find the set of b - isotropic boxes we need not to apply @xmath162 . by theorem [ thm : image ] , which we prove below , the set @xmath165 is equal to the set of boxes invariant under elements of @xmath159 . this theorem is true for any subset of linear space , but for clarity , we state it for the set of consistent boxes . for a hypergraph @xmath37 and the set of consistent boxes @xmath98 compatible with this graph , let @xmath166 be a finite group of linear maps @xmath167 and @xmath168 a subset of its elements such that each of them have its inverse @xmath169 in @xmath166 . let us define a family of boxes @xmath19 invariant under transformations @xmath170 : : = \{b c^(n)_g : _ i h_i ( b ) = b } , and a subgroup @xmath171 generated by @xmath172 . we then have the following : _ _ ( c^(n)_g ) : = \{d c^(n)_g : _ b c^(n)_g _ f _ f(b ) = d } = . [ thm : image ] _ proof . _ let @xmath173 . then for each @xmath174 we have : h_i(b ) & = & h_i(_f _ f(b ) ) + & = & _ f _ h_if(b ) + & = & _ = h_if _ ( b ) = b , where in first step we use linearity of the maps @xmath170 and in the last we use the fact that @xmath175 runs through the whole group @xmath176 since each @xmath170 has its inverse . from the above we see that @xmath177 , and so @xmath178 . on the other hand , for each box @xmath179 we have : [ eq : b ] b = _ f _ f(b ) , because for all @xmath180 , @xmath181 ( @xmath182 ) , and so @xmath183 , from which we arrive at eq.([eq : b ] ) . thus , we showed that @xmath184 which , jointly with the opposite inclusion , proves the theorem . consider now specific set of non - contextuality preserving automorphisms @xmath185 which is any set of compositions of two types of linear maps : ( i ) @xmath44 - permutations of observables , and ( ii ) @xmath45 - negations of outputs of observables . for this set we have general theorem which allows for easier evaluating the relative entropy of contextuality . for any box @xmath38 and a set of b-@xmath185-isotropic boxes @xmath186 we have : [ xu ] x_u(b)= _ p()^_0_b _ c 1n d(g(_c ) \|\| p(_c ) ) , where the minimum is taken over all probability distributions @xmath7 which give rise to non - contextual box from the set of b-@xmath185-isotropic boxes @xmath186 . [ thm : iso - optimal ] _ proof . _ let @xmath7 be optimal for @xmath187 , and denote the non - contextual box defined by this distribution as @xmath188 . because of the choice of @xmath185 , for any element @xmath189 in group @xmath190 generated by this set , there is x_u ( b ) = _ c 1n d(g_f(_c ) \|\| p_f(_c ) ) , where @xmath191 and @xmath192 are distributions of context @xmath4 of a box @xmath193 and a box @xmath194 respectively . to see this , we note that , by definition of @xmath185 , @xmath189 is a composition of permutation of observables and bit - flips of their outputs . it is then enough to prove separately that the above equality holds , for @xmath189 being one of them . consider first @xmath189 to be a permutation of observables . since @xmath183 , it is also an automorphism of @xmath37 with which @xmath19 is compatible , hence it is special permutation of observables which induces permutation of the contexts and in turn of elements @xmath195 . it means that applying @xmath189 induces just change of the order of summation in the definition of @xmath31 . second , if @xmath189 is a bit - flip , since it is applied to both @xmath196 and @xmath197 , it does not change the relative entropy which is invariant under doubly applied reversible operations @xcite . thus we have : x_u ( b ) = _ c 1n _ f__0 d(g_f(_c ) \|\| p_f(_c ) ) _ c 1n d ( _ f__0 g_f(_c ) \|\| _ f__0 p_f(_c ) ) , [ xineq ] where in the second line we used the joint convexity of relative entropy . what we obtain is the fact that such process of symmetrization can not increase the relative entropy . what is more , since @xmath189 is an automorphism of @xmath19 , we have that for each context @xmath4 : _ f _ _ 0 g_f(_c ) = g(_c ) . we observe now , that since @xmath198 preserves non - contextuality , the box @xmath199 has a context @xmath4 equal to @xmath200 and is a non - contextual box . since @xmath188 is optimal for @xmath31 and , when we substitute the box @xmath199 in place of @xmath188 , we can not increase the quantity @xmath31 due to inequality ( [ xineq ] ) , the box @xmath199 must also be optimal for @xmath31 , which proves desired equality in ( [ xineq ] ) . we have @xmath201 hence the assertion follows . in this section we specify twirling operations @xmath185 for the xor - boxes , hence showing that one can obtain the isotropic xor - boxes by operations that are non - contextuality preserving . this is crucial , since then we can use theorem [ thm : iso - optimal ] to compute @xmath31 for these boxes , which is done in theorem [ thm : iso - xu - values ] . we first define the peres - mermin s ( pm ) and mermin s ( m ) box below : the @xmath65 is a box on @xmath202 with @xmath203 for first 5 contexts , and @xmath204 for the 6th one @xcite . the @xmath66 is a box on @xmath205 with @xmath206 for first 4 contexts , and @xmath207 for the 5th one @xcite . to begin with , we introduce twirling for @xmath65 box by specifying the set @xmath185 which leads to one - parameter family of isotropic boxes . * for pm box we choose @xmath208 , where the elements @xmath170 are in general compositions of maps ( i ) and ( ii ) : @xmath209 - are 3 ! permutations of contexts @xmath210 i.e. the rows on fig . [ figurynka : pm ] , @xmath211 - a swap of @xmath212 and @xmath213 , i.e. a swap of two solid columns on fig . [ figurynka : pm ] , @xmath214 - is a composition of permutation defined by mappings : @xmath215 , @xmath216 , @xmath217 ( the rest of the observables are mapped to themselves ) , composed with a bit - flip of the output of observable @xmath218 . this operation is a reflection of the hypergraph w.r.t . to the diagonal with appropriate bit - flip . on fig . [ figurynka : pm ] . the set @xmath219 we call the set of isotropic @xmath65 boxes . the reason for this is stated in lemma below : there holds : ^_0_pm= \{pm + ( 1- ) pm\| } , where @xmath220 is an opposite version of the box @xmath65 , i.e. @xmath65 with @xmath221 exchanged with @xmath222 and vice versa . [ lem : iso : equal - pm ] _ proof_. to see the above statement , we will use theorem [ thm : image ] . due to this theorem it is enough to argue that invariance of a box under @xmath185 implies that it belongs to @xmath219 . in the proof we will refer to fig . [ figurynka : pm ] . first , due to invariance under @xmath209 the rows need to have the same probability distribution . second , using @xmath214 we obtain that middle row and middle column has the same distributions . third , by @xmath211 we get that all solid lines has the same distributions with 8 probabilities @xmath223 of string @xmath224 where @xmath225 are binary . due to invariance under @xmath209 , both the solid columns and the dashed column are permutationally symmetric , i.e. are described only by @xmath226 , @xmath227 , @xmath228 and @xmath229 ( @xmath230 , @xmath231 , @xmath232 and @xmath233 for dashed column ) . invariance under operation @xmath214 imposes @xmath234 and @xmath235 which equalizes @xmath226 and @xmath228 because of @xmath236 . thus @xmath237 for some parameter @xmath238 $ ] . similarly , we have @xmath239 , @xmath240 , which implies that @xmath241 . exchanging @xmath242 with @xmath243 , we get that also @xmath244 and @xmath245 , which ends the proof . the argument given in the above lemma is analogous in the case of other xor - boxes considered in this paper , where in particular we have : * for @xmath66 box we choose @xmath246 , where : @xmath247 - reflection of the star with respect to the @xmath248 symmetry line , @xmath249 - reflection of the star with respect to the @xmath250 symmetry line with bit flip on the node @xmath4 , @xmath251 - reflection of the star with respect to the @xmath252 symmetry line with bit flip on the node @xmath85 , @xmath253 - reflection of the star with respect to the @xmath254 symmetry line with bit flip on the node @xmath255 , @xmath256 - reflection of the star with respect to the @xmath257 symmetry line with bit flip on the node @xmath19 , @xmath258 - bit flips on three nodes that form any triangle on the hypergraph ( @xmath259 , @xmath260 , etc . ) . for such defined @xmath185 there holds : ^_0_m = \{m + ( 1- ) m \| } , [ lem : iso : equal - m ] and the set of these boxes we call isotropic @xmath66 boxes . * for @xmath9 box we choose @xmath261 , where : @xmath262 is a composition of cyclic permutation of contexts such that all @xmath263 with bit flips on the observables @xmath264 . for such defined @xmath185 there holds : ^_0_ch^(n)=\{ch_(n ) + ( 1- ) ch_(n)\| } , [ lem : iso : equal - ch ] and the set of these boxes we call isotropic @xmath9 boxes . let us now fix a contextual box @xmath265 and denote by @xmath266 the outcome @xmath174 of distribution @xmath5 under a measurement on the box @xmath19 the context @xmath4 . for a box @xmath267 compatible with the same hypergraph @xmath37 as @xmath19 , we define the quantity @xmath268 which measures how contextual is box @xmath269 w.r.t . to box @xmath19 : _ b ( ) : = _ c _ i ( g(_c ) ) ( _ c)_i , where @xmath5 are probabilities of outcomes within a given context , and @xmath270 is the support of the distribution @xmath5 , i.e. the set of the outcomes of a measurement of the context @xmath4 which have nonzero probability in distribution @xmath5 . we will need some properties of @xmath271 , which are collected in the lemma below , where we treat boxes as vectors of probabilities . for any box @xmath272 and an xor - box @xmath38 with all contexts of the same cardinality @xmath273 and all observables of the same cardinality @xmath62 , there holds : _ b ( ) = 2^(m-1)\|b , where @xmath274 is euclidean scalar product of vectors . moreover , for a twirling @xmath275 there holds : _ b ( ) = _ b(_b^_0 ( ) ) . [ lem : bet - prop ] _ proof_. the first statement is easy , as @xmath276 is a vector of @xmath100 s for probabilities for which @xmath277 where @xmath14 , hence the scalar product sums the probabilities of box @xmath269 from the support of box @xmath19 . to see the next consider the following chain of equalities : @xmath278 where in the second equality we use linearity of scalar product , in the third we use the fact that by definition of twirling @xmath189 is an automorphism of @xmath19 and in the fourth we use the fact , that each @xmath189 is a composition of elements from @xmath279 , i.e. permutations of observables and bit flips of outputs , hence it is a permutation , which does not change the scalar product . based on @xmath268 we can build naturally a contextuality inequality , which for @xmath65 box is equivalent to that given in @xcite , for pr box that given in @xcite and for @xmath9 box that from @xcite ( see also @xcite ) . for an xor - box @xmath38 with a single context with distribution @xmath280 , such that each vertex from @xmath17 belongs to even number of contexts and for a non - contextual box @xmath272 , there holds : _ b ( ) n-1 , [ eq : beta - n-1 ] and the bound is tight . [ thm : beta - n-1 ] _ proof . _ in what follows , we generalize the argument of n.d . mermin @xcite , with the use of which he proved that @xmath66 box is contextual . since any noncontextual box is a mixture of deterministic boxes , and by lemma [ lem : bet - prop ] , @xmath281 is linear , it suffices to prove the above inequality for deterministic ones . surely , deterministic boxes can attain only discrete values of lhs . suppose then , that for noncontextual box @xmath282 , i.e. all constraints of a contextual box are satisfied , meaning that for @xmath283 contexts the sum of outputs @xmath284 ( @xmath285 ) and for @xmath100 context @xmath286 ( @xmath287 ) , which gives a total sum over all contexts 1 . on the other hand , for deterministic assignment , summing all the values for the whole hypergraph we get @xmath288 since each @xmath289 ( the number of contexts to which the observable @xmath0 belongs to ) is an even number by the assumption . this gives desired contradiction . the value of @xmath290 can be attained deterministically , e.g. by putting all the outcomes equal 0 , which simultaneously tighten the inequality . for an xor - box @xmath38 with even @xmath8 and a simple context with distribution @xmath280 , such that each vertex from @xmath17 belongs to even number of contexts and for a non - contextual @xmath272 , we have : _ b()1 . [ thm : beta-1 ] moreover if the number of vertices in each context is odd then the bound is tight . _ the argument is analogous to the proof of theorem [ thm : beta - n-1 ] . again , we only need to consider deterministic assignments . suppose there is a deterministic assignment of outcomes with @xmath291 . then the box would satisfy all the constraints of contextual opposite version of a box b. for this box , @xmath283 contexts has the sum of outputs equal to @xmath286 ( @xmath287 ) and for @xmath100 context @xmath284 ( @xmath285 ) , which gives a total sum over all contexts 1 . this however is in contradiction with the fact that the sum over all vertices is 0 since each vertex appears an even number of times in the sum . hence @xmath292 . to see the tightness in a special case , we observe that setting each vertex value 1 constitutes a deterministic assignment that has @xmath268 equal to 1 . indeed , since each context has an odd number of vertices , each edge has distribution @xmath280 , and exactly one of them is in accordance with box @xmath19 . we note , that the assumption about evenness of @xmath8 in the above theorem is necessary : for @xmath293 with odd @xmath8 , there exists a non - contextual box @xmath269 such that @xmath294 . [ obs : zero ] _ proof . _ there exists a deterministic assignment which sets @xmath268 to zero : first , we set all observables to 1 , and then change into 0 @xmath82 of those which does not belong to context which has @xmath280 in b , but such that each belong to disjoint pair of contexts . such an assignment creates an opposite version of a box @xmath19 , hence giving @xmath295 . as we have seen , all examples of sets of isotropic xor - boxes considered so far are one parameter . we now prove the lemma which bounds this parameter for non - contextual boxes . for a non - contextual box @xmath296 with @xmath8 contexts , there holds : . [ eq : r - bound ] for even @xmath8 there holds additionally : , [ eq : l - bound ] while for odd @xmath8 there is 0 , [ eq : z - bound ] and the bounds are tight . [ lem : bounds ] _ proof . _ by the definition of @xmath268 , for any xor - box @xmath19 we have @xmath297 where @xmath298 is an opposite version of box @xmath19 ( with @xmath299 in place of @xmath300 and vice versa ) . this implies that for any isotropic xor - box @xmath301 we have _ b(b _ ) = n , [ eq : alpha - beta ] and in particular , for non - contextual isotropic boxes @xmath301 by theorem [ thm : beta - n-1 ] we have . [ eq : upper - alpha ] to prove the second inequality , we observe , that @xmath65 and @xmath9 for even @xmath8 satisfies the assumption of theorem [ thm : beta-1 ] , which gives the inequality ( [ eq : l - bound ] ) in analogous way . the last inequality follows from the dependence ( [ eq : alpha - beta ] ) and observation [ obs : zero ] . to see that the boundary values of @xmath302 are attained by non - contextual boxes , we first observe that by theorem [ thm : beta - n-1 ] , there exists a non - contextual box @xmath269 with @xmath303 . now by lemma [ lem : iso : equal - pm ] equalities ( [ lem : iso : equal - m ] ) and ( [ lem : iso : equal - ch ] ) after twirling @xmath304 @xmath269 belongs to @xmath305 , hence it has a form @xmath306 where @xmath298 is opposite version of b. by lemma [ lem : bet - prop ] , @xmath307 . now , by equation ( [ eq : alpha - beta ] ) , we have @xmath308 proving that @xmath309 attains the value @xmath310 of @xmath302 . this box is clearly non - contextual , since @xmath304 is application at random some permutation of observables composed with bit - flips on outputs of observables , hence preserving non - contextuality . analogous argument , by use of theorem [ thm : beta-1 ] and observation [ obs : zero ] , proves the tightness of the bounds ( [ eq : l - bound ] ) and ( [ eq : alpha - beta ] ) respectively . we can state the main theorem of this section : for @xmath311 with @xmath101 number of contexts and @xmath312 there holds x_u(b _ ) = -h ( ) , [ eq : xu - n-1 ] while for even @xmath8 and @xmath313 there holds x_u(b _ ) = - h ( ) , [ eq : xu-1 ] where @xmath52 . [ thm : iso - xu - values ] _ proof . _ we first note , that in both cases we want to consider , the isotropic boxes satisfy assumptions of theorem [ thm : iso - optimal ] , hence we need to optimize only over appropriate isotropic xor - boxes : x_u(b_)= _ p()^_0_b _ _ c 1n d(g(_c ) \|\| p(_c ) ) , [ xu2 ] where for short by @xmath314 we mean some isotropic non - contextual box @xmath315 which is defined by distribution @xmath7 . since @xmath315 has only two kinds of distributions @xmath25 , @xmath316 and @xmath317 we can write : x_u(b _ ) = _ _ 0 ( ( n-1 ) d(p^_even \|\| p^_0_even ) + d(p^_odd @xmath315 is noncontextual , and @xmath318 and @xmath319 are defined analogously to @xmath320 and @xmath321 , respectively . now , it is easy to check that if @xmath322 for @xmath323 , the assumption of the lemma [ lem : one - term ] are satisfied with @xmath324 being identity operations for every @xmath4 for @xmath283 contexts for which @xmath325 has the same distribution and bit - flip on one of the observables on a distribution of the remaining context , giving : x_u(b _ ) = _ _ 0 d(p^_even\|\| p^_0_even ) , [ eq : xu - one - term ] where @xmath326 is bounded by the fact that @xmath320 is a distribution of a non - contextual box which is isotropic . more specifically , it is bounded according to lemma [ lem : bounds ] , i.e. we have @xmath327 . it is easy to check that for @xmath328 the function ( [ eq : xu - one - term ] ) is decreasing with @xmath326 . lemma [ lem : bounds ] shows that the boundary value @xmath329 is attained by non - contextual isotropic xor - box , hence the function attains minimum for this value of @xmath326 , which proves ( [ eq : xu - n-1 ] ) . for @xmath330 , this function is increasing , and again by lemma [ lem : bounds ] attains minimal value at @xmath331 , which proves ( [ eq : xu-1 ] ) . we note here , that due to the above theorem , @xmath49 for even @xmath8 , @xmath332 equals the value of @xmath333 for @xmath334 , in correspondence with the fact that @xmath325 can be changed by bit - flips into @xmath335 which does not change the relative entropy distance . in particular , for considered examples of xor - boxes i.e. in the case @xmath336 we have : x_u(pr ) = 0.4150 , + x_u(pm ) = 0.2630 , + x_u(m ) = 0.3219 , + x_u(ch_(n ) ) = . according to the above formula , @xmath31 tends to zero in asymptotic limit for maximally contextual chain boxes . interestingly , if we do not take average over number of contexts , i.e. consider a measure @xmath337 , it will equal to @xmath338 and tend asymptotically to @xmath339 where @xmath340 is the euler number . in other words , if we consider natural logarithm in definition of relative entropy , @xmath69 tends to 1 with increasing @xmath8 . it means that , although `` average '' contextuality of chain box - per number of contexts - vanishes with increasing @xmath8 , the `` total '' contextuality is bounded by 1 from below . remarkably , the same result holds for quantum maximally contextual chain boxes : @xmath341 based on natural logarithm tends to 1 for both odd and even @xmath8 where @xmath302 for each @xmath8 is given in description of the fig [ fig : miary_chn ] . for comparison , in the case of maximal violation of chsh inequality we have @xmath342 with @xmath343 which gives : x_u(b_chsh ) 0.0463 . [ app : not - equal ] as it was mentioned , @xmath344 . in this section we prove that these two measures are equal for isotropic xor - boxes but are different in general . more precisely , we find their values on _ direct sum _ of two boxes , in terms of their values of the boxes themselves in theorem [ thm : direct - sum ] . using this result , we show among others that these measures differ on any direct sum of contextual box and non - contextual one in corollary [ cor : xu - c - nc - differ ] . to begin with , we show that in equation [ eq : xu ] we can indeed write minimum instead of infimum . we will show more , proving that in definition of @xmath117 one can also consider minimum . taking than uniform distribution , we get the thesis for @xmath79 . we also prove that both measures are faithful and are monotonous under certain subclass of noncontextuality preserving operations . recall that for a box @xmath14 , x_\{p(c)}(b)= _ \{p ( ) } _ c e_g p(c ) d(g(_c ) \|\| p(_c ) ) which depends on both the box @xmath19 and probability distribution over contexts @xmath345 . recall here also , that the minimization is taken over all joint probability distributions @xmath7 defined on outputs of observables , and @xmath25 are marginals , restricted to the context @xmath4 . this quantity can be written as ( using quantum notation for classical distributions ) x_\{p(c)}(b)=_s(\| ) where @xmath346 and @xmath347 with @xmath348 representing probability distributions @xmath25 , and @xmath134 - the distributions @xmath5 of the box @xmath19 . one finds that the set of states @xmath135 is convex and compact ( note , that @xmath349 is fixed here ) . therefore , since relative entropy is lower semicontinuous @xcite , there exists state @xmath350 , which achieves the infimum . this finishes the proof . we first argue , that both @xmath31 and @xmath32 are _ faithful _ i.e. that are nonzero iff the box is contextual . indeed , the relative entropy is lower bounded by square of variational distance between the probability distributions , which is zero only if all @xmath5 are equal to @xmath25 . this however can not hold for there is no single joint probability distribution with marginals @xmath5 . we now prove monotonicity of @xmath31 and @xmath32 under operations which lie within the set of non - contextuality preserving ones . given a box @xmath351 , we define an operation @xmath352 as a mixture with some probabilities @xmath353 of independent channels @xmath354 acting on observable @xmath355 . it is easy to see , that such defined set of operations is a convex subset of non - contextuality preserving operations . to see that @xmath79 and @xmath32 are monotonous under these operations we need to use the fact that both measures can be expressed in terms of mutual information ( equation [ eq : mutual - form ] ) . application of @xmath352 is then a local action on variable @xmath18 , and monotonicity of both @xmath79 and @xmath32 follows directly from data processing inequality @xcite . we will now need another fact that apart from theorem [ thm : iso - optimal ] , simplifies computation of @xmath31 . let @xmath23 and x_u(b ) = _ p ( ) s _ c 1n d(g(_c)\| p(_c ) ) , for some @xmath356 . if for any @xmath105 there exist reversible operations @xmath324 satisfying @xmath357 for all contexts @xmath4 and simultaneously @xmath358 for all contexts @xmath4 , then x_u(b ) = _ p ( ) s d(g(_c_0)\| p(_c_0 ) ) . [ lem : one - term ] _ proof . _ the proof boils down to the observation that relative entropy is invariant under bilateral reversible operations @xcite . we observe , that for all isotropic xor - boxes introduced in previous section , the assumption of the above lemma are satisfied . this enables us to prove the following important result : for isotropic boxes @xmath359 , there is @xmath360 . [ thm : xmax - xu ] _ proof . _ let us calculate the quantity @xmath361 defined for a given @xmath125 , which is optimal for the measure @xmath187 . according to the assumptions of lemma [ lem : one - term ] valid for isotropic boxes , we can write x^(b ) & = & _ c p(c ) d ( g(_c_0)\|\|p(_c_0 ) ) + & = & d ( g(_c_0)\|\|p(_c_0 ) ) = x_u(b ) . on the other hand , from the definition of @xmath362 we have x _ \{p(c ) } ( b ) x^(b ) , hence x _ \{p(c ) } ( b ) x_u(b ) , and taking the supremum gives @xmath363 while combining this with the fact @xmath364 we obtain the equality of the two measures for isotropic boxes . we now introduce definition of _ direct sum _ of hypergraphs and boxes . for two hypergraphs @xmath365 and @xmath366 , a direct sum of @xmath367 and @xmath368 is @xmath369 with @xmath370 and @xmath371 . for any two boxes @xmath372 and @xmath373 compatible with hypergraphs @xmath367 and @xmath368 respectively , their direct sum is a box @xmath374 . in the next part of this section , we use the following notation . by @xmath375 $ ] we mean any joint probability distribution of the outputs of observables from set @xmath3 , and by @xmath376 the marginal probability distribution of @xmath7 , defined on the outputs of observables of set @xmath3 . moreover , by @xmath377 we mean the relative entropy distance between distribution of the output of variables from context @xmath4 of box @xmath6 and that from context @xmath4 of non - contextual box defined by distribution @xmath7 . we will now need a lemma , which simplifies computation of @xmath31 and @xmath32 of direct sum of boxes , as it states , that it is enough to take minimization in both quantities only over product distributions . for any two hypergraphs @xmath378 and @xmath379 and boxes @xmath380 and @xmath381 there holds : x_u(b_1b_2 ) = _ p()[v_g_1]p()[v_g_2 ] _ c e_g_1g_2 1n_1 + n_2 d(g(_c)\|\|p(_c ) ) , [ eq : xu - prod ] and x_max(b_1b_2 ) = _ \{p(c)}_p()[v_g_1]p()[v_g_2 ] _ c e_g_1g_2 p(c ) d(g(_c)\|\|p(_c ) ) . [ eq : xmax - prod ] [ lem : xu - xmax - opt - prod ] _ proof . _ to see both equalities we observe that for any distribution @xmath15 and any distribution @xmath382 $ ] , there holds @xmath383}\right . \\ = \sum_{c \in e_{g_1 } } p(c ) d(g(\lambda_c)\|\|p(\lambda_c))\left\|_{p(\lambda)[v_{g_1\oplus g_2}]\|_{v_{g_1}}}\right . + \sum_{c \in e_{g_2 } } p(c ) d(g(\lambda_c)\|\|p(\lambda_c))\left\|_{p(\lambda)[v_{g_1\oplus g_2}]\|_{v_{g_2}}}\right . \\ = \sum_{c \in e_{g_1 } \cup e_{g_2 } } p(c ) d(g(\lambda_c)\|\|p(\lambda_c))\|_{p(\lambda)[v_{g_1\oplus g_2}]\|_{v_{g_1}}p(\lambda)[v_{g_1\oplus g_2}]\|_{v_{g_2}}}\end{gathered}\ ] ] this is because by definition of @xmath384 contexts from @xmath385 depend only on variables from @xmath386 , similarly as contexts from @xmath387 depend only on @xmath388 . hence @xmath389p(\lambda)[v_{g_2 } ] } \sum_{c \in e_{g_1 } \cup e_{g_2 } } d(g(\lambda_c)\|\|p(\lambda_c))$ ] , which for @xmath390 for all @xmath4 implies ( [ eq : xu - prod ] ) . taking supremum over @xmath15 , we obtain ( [ eq : xmax - prod ] ) . we are ready to show our main tool , interesting on its own , which is the following theorem that expresses @xmath31 and @xmath32 of a direct sum of two boxes in terms of these functions of these boxes . for any two hypergraphs @xmath367 and @xmath368 and boxes @xmath380 and @xmath381 there holds : x_u(b_1b_2 ) = n_1 x_u(b_1 ) + n_2 x_u(b_2 ) , [ eq : xu - oplus - sum ] and x_max(b_1b_2 ) = max \ { x_max(b_1 ) , x_max(b_2 ) } . [ eq : xmax - oplus - max ] [ thm : direct - sum ] _ proof . _ to see both the above statements , we observe that for a distribution @xmath15 such that @xmath391 and @xmath392 , and for any two distributions @xmath393 $ ] and @xmath394 $ ] , we have @xmath395p(\lambda)[v_{g_2}]}\right . \\ = w \sum_{c \in e_{g_1 } } { p(c)\over w } d(g(\lambda_c)\|\|p(\lambda_c))\left\|_{p(\lambda)[v_{g_1}]}\right . + ( 1-w)\sum_{c \in e_{g_2 } } { p(c)\over { 1- w } } d(g(\lambda_c)\|\|p(\lambda_c))\left\|_{p(\lambda)[v_{g_2}]}\right.\end{gathered}\ ] ] this immediately gives : @xmath396p(\lambda)[v_{g_2 } ] } \sum_{c \in e_{g_1 } \cup e_{g_2 } } p(c ) d(g(\lambda_c)\|\|p(\lambda_c))\left\|_{p(\lambda)[v_{g_1}]p(\lambda)[v_{g_2}]}\right . = \\ = w \min_{p(\lambda)[v_{g_1}]}\sum_{c \in e_{g_1 } } { p(c)\over w } d(g(\lambda_c)\|\|p(\lambda_c ) ) + ( 1-w)\min_{p(\lambda)[v_{g_2 } ] } \sum_{c \in e_{g_2 } } { p(c)\over { 1- w } } d(g(\lambda_c)\|\|p(\lambda_c ) ) . \label{eq : w - minimas}\end{gathered}\ ] ] substituting @xmath390 in the above equation , we obtain by lemma [ lem : xu - xmax - opt - prod ] that @xmath397 } { 1\over n_1 } \sum_{c \in e_{g_1 } } d(g(\lambda_c)\|\|p(\lambda_c ) ) ] + { n_2\over { n_1 + n_2}}[\min_{p(\lambda)[v_{g_2 } ] } { 1\over n_2 } \sum_{c \in e_{g_2 } } d(g(\lambda_c)\|\|p(\lambda_c ) ) ] \\ = { n_1\over { n_1 + n_2 } } x_{u}(b_1 ) + { n_2\over { n_1 + n_2}}x_u(b_2),\end{gathered}\ ] ] which proves the statement ( [ eq : xu - oplus - sum ] ) . we now pass to prove the statement ( [ eq : xmax - oplus - max ] ) . we can assume w.l.g . that @xmath398 . by definition of @xmath32 , for any @xmath399 , there exists @xmath400 such that @xmath401 . we will argue now , that for any @xmath82 and decreasing @xmath402 , there holds @xmath403 , which will prove desired upper bound in limit @xmath404 . the proof then follows from the fact , that this upper bound can be attained by taking @xmath405 and such @xmath400 that attain supremum in @xmath406 in limit of large @xmath82 . we need to consider only two cases : @xmath400 is such that @xmath407 ( case 1 ) or @xmath408 ( case 2 ) . in the first case we have @xmath409 which we aimed to prove . thus , it is enough to show that in the second case @xmath410 is upper bounded by @xmath411 , since then the first case yields optimal value of @xmath410 . suppose then , that @xmath400 satisfies @xmath412 . this implies that @xmath413 and @xmath414 are valid distributions , hence from ( [ eq : w - minimas ] ) , by lemma [ lem : xu - xmax - opt - prod ] , we have x_max(b_1b_2 ) w_k x_\{p_k(c)w_k}_ce_g_1(b_1 ) + ( 1-w_k ) x_\{p_k(c)1- w_k}_ce_g_2(b_2 ) + _ k. by definition of @xmath32 we have @xmath415 and @xmath416 which gives from the above equality @xmath417 hence , as we explained , the assertion follows . the above theorem can be easily generalized to any finite direct sum of boxes , giving that @xmath31 is the average value of the @xmath31 of particular boxes from the direct sum ( with weights according to cardinality of their number of contexts ) , and @xmath32 is the maximal value of @xmath32 on particular boxes . we can state now the main application of this theorem . for any two hypergraphs @xmath367 and @xmath368 , and a boxes @xmath380 and @xmath381 with @xmath418 , such that @xmath419 , there holds x_u(b_1b_2 ) < x_max ( b_1b_2 ) . [ cor : xu - diff ] _ proof_. since @xmath419 , we can w.l.g . assume @xmath420 . this implies , by theorem [ thm : direct - sum ] , @xmath421 which proves the corollary . from the above corollary we obtain immediately another one : for any two hypergraphs @xmath367 and @xmath368 , a contextual box @xmath422 and a non - contextual box @xmath423 with @xmath424 , there holds x_u(bb_nc ) = n_1 x_u(b ) < x_max ( bb_nc ) = x_max(b ) . [ cor : xu - c - nc - differ ] _ proof_. it is enough to observe , that @xmath31 is faithful , hence @xmath425 and the corollary [ cor : xu - diff ] applies . the above corollary states that @xmath79 and @xmath32 differ on certain direct sums of boxes . exemplary can be @xmath426 , since @xmath427 is maximally mixed box , which is clearly non - contextual and @xmath428 is contextual . there are also quantum boxes , i.e. that originate from performing certain measurements on a quantum state . exemplary is defined as follows : consider the maximally entangled state @xmath429 and consider the hypergraph with @xmath430 where @xmath431 @xmath432 @xmath433 , @xmath434 , @xmath435 } and @xmath436 @xmath437 @xmath438 . @xmath69 and @xmath439 are pauli matrices and @xmath440 represents the rotation of pauli matrix around @xmath441 axis by angle @xmath442 . we then consider a box @xmath19 obtained via measuring observables from @xmath17 on the state @xmath443 in groups defined by contexts . it is easy to see , that this box is a direct sum of the most non - local quantum box with two binary inputs and two binary outputs defined by first 4 observables and first 4 contexts , and a box with a single context , which is by definition non - contextual . a given hypergraph @xmath37 yields a definition of a large family of consistent boxes @xmath98 . e.g. one can consider only those boxes which emerges as measurements of some observables on quantum state . one can narrow the latter family even more , by considering in @xmath17 only observables that are rank-1 projectors ( see e.g. @xcite and references therein ) . then commensurability of 2 projectors turns to be mutual orthogonality , and the hypergraph can be interpreted as _ orthogonality graph_. this is the case for the klyachko et al . result @xcite , where a quantum state of qutrit is found , and measurements which give rise to maximal violation of the so called pentagon inequality . such a pair : a quantum state and the set of measurements defines naturally a box called further kcbs box , denoted as k. using similar techniques to that for xor - boxes , we find now the value of @xmath31 and @xmath32 for this box . to be more precise the measurements given in @xcite form the set of 5 projectors @xmath444 , designed in a such way , that they violate the so called pentagram inequality . this setup corresponds to a hypergraph that is pentagon , so that projectors commute in pairs : @xmath445 with @xmath446 , @xmath446 with @xmath447 in a circle so that the last commutation is @xmath448 with @xmath445 , so the graph is @xmath449 , but there are additional restrictions , since observables are one - dimensional projectors , so that their commutation implies their orthogonality . this implies that , e.g. the probability of the result @xmath450 which corresponds to obtaining an outcome 1 for both projectors as a result of measurement , is zero . more specifically , the klyachko box is given by the same 5 distributions of the form @xmath451 , @xmath452 and @xmath453 . this implies that there are less symmetries than in the xor - boxes . after corresponding twirling , which we show below , there are 2 parameters left . fortunately , we can make use of the fact that @xmath453 in different way : it means that the corresponding classical probability distribution should have @xmath454 for all contexts , since the formula minimizes over the classical probabilities . to describe the twirling consider the set of joined probability distributions @xmath7 with its @xmath455 extremal points . as in case of xor - boxes , we apply twirling determined by group @xmath456 which turns to be dihedral group @xmath457 consisting of @xmath458 permutations . due to symmetrization the marginal probability distributions calculated for the extremal points turns out to be context independent @xmath459 and to posses additional symmetry @xmath460 . as a consequence @xmath461 extremal points , which under the action of group @xmath457 form @xmath462 orbits ( subsets of 32 extremal points invariant under @xmath457 ) the box can be characterized by only two parameters , e.g. @xmath463 , and @xmath464 and conveniently visualized by a @xmath462 points on a triangle plot as it is shown in fig.[triangle ] . thus a set of all symmetrized non - contextual distributions is a convex combination of @xmath465 distributions , namely @xmath466 , @xmath467 , @xmath468 and @xmath469 , and finally @xmath470 and @xmath469 . now our goal is to find minimum in @xmath471 which , due to all mentioned symmetries , is given simply by x_u(k)= _ d(g(_c_0)\|\| ) . note that @xmath453 , so that looking for a minimum we can restrict to the case of @xmath472 . in this way the problem of calculating @xmath31 has been reduced to finding minimum over a single parameter @xmath463 in the range between @xmath473 and @xmath100 x_u(k)=_0.2(00)1 ( g(00),(00 ) ) , where ( x , y)=x ( ) + ( 1-x ) ( ) . @xmath474 is strictly increasing function of argument @xmath441 provided that @xmath475 which is seen from equation = . so finally we get @xmath476 . for reasons similar to that given for xor - boxes , we have in this case @xmath477 . . the bold points denote orbits , i.e. subsets of 32 extremal points invariant under @xmath457.,scaledwidth=40.0% ] in this section we prove that for exemplary xor - boxes @xmath31 is additive , and that it is @xmath62-copy additive for isotropic xor - boxes considered in this paper . we begin with definition and necessary lemmas . the main results are theorems [ thm:2add ] and [ thm : ex - add ] , and their main application is stated in corollary [ cor : iso-2add ] . for any two hypergraphs @xmath378 and @xmath379 , we define their tensor product to be the hypergraph g_1 g_2:=(v_g_1 g_2 , e_g_1 g_2 ) where @xmath478 and @xmath479 and @xmath480 . for two boxes @xmath481 and @xmath482 compatible with hypergrahps @xmath367 and @xmath368 respectively , their tensor product is a box compatible with @xmath483 given by @xmath484 i.e. such that the distribution of its context @xmath485 is a product of distributions @xmath486 and @xmath487 . [ def : tensor ] to see this , consider an extremal point of @xmath488 . it is equal to a box with joint distribution over @xmath490 observables @xmath491 for some @xmath110 . such a distribution is a product of distributions @xmath492 and @xmath493 where @xmath494 and @xmath495 are output strings of outputs @xmath496 and each @xmath497 . @xmath498 and @xmath499 are some fixed output strings . @xmath494 and @xmath495 when written in a system with basis @xmath85 ( assuming that all of them are equal , otherwise one has to consider a multibase system ) and concatenating yields @xmath500 . hence any extremal point of @xmath488 is a product of extremal points of the set @xmath39 . after any linear operation @xmath501 on any convex set @xmath502 , the set of extremal points of the image set @xmath503 is the subset of the set of images of extremal points of @xmath502 through @xmath501 . [ lem : tau ] consider any point @xmath504 which is an image of non - extremal point @xmath19 in @xmath502 . since @xmath501 is linear , we have @xmath505 hence it can not be extremal in @xmath503 . thus , any extremal point in @xmath503 must be an image of extremal point in @xmath502 . by lemma [ lem : tau ] the only extremal points in @xmath509 are within the set of images of extremal points through @xmath501 . we know from observation [ obs : nonc ] that extremal points of @xmath510 are of the form @xmath511 where @xmath512 are extremal points of @xmath39 . now if @xmath513 is not extremal in @xmath508 i.e. can be decomposed into @xmath514 then clearly the image @xmath515 for any @xmath516 is not an extremal in @xmath509 , as it can be decomposed into nontrivial mixture @xmath517 . the same argument holds for @xmath518 : it can not be non - extremal in @xmath508 if the pair @xmath511 is extremal in @xmath509 . hence , the only extremal points in @xmath509 are the tensor products of extremal points in @xmath508 . let box @xmath523 be invariant under some linear operation @xmath501 , which maps all boxes on @xmath98 into interval @xmath524 $ ] and maps @xmath39 into interval @xmath525 \subset [ b_e , b_e']$ ] . let also some of @xmath5 be equal to @xmath526 and the rest of the @xmath5 be equal to @xmath527 for some reversible operation @xmath528 . then , there holds : x_u(b^2)=_p_nc(nc_g^2 ) d(g(_c_0)g(_c_0)\|\|p_\|1 ) where @xmath529 is a product of distributions @xmath526 and @xmath530 is the distribution of some fixed context number @xmath531 of @xmath532 . _ proof : _ let @xmath533 be the number of the contexts of @xmath19 with the same distribution @xmath526 and @xmath534 the number of the remaining contexts with distribution @xmath527 . in what follows , we identify @xmath526 with @xmath243 and @xmath527 with @xmath535 for short . we know that x_u(b^2 ) = _ p_nc ( nc_g^2)1n^2_c , c d(g(_c)g(_c ) \|\| p_cc ) where @xmath536 is total no . of contexts and @xmath537 . from the observations [ obs : nonc ] and [ lem : span ] , the box @xmath532 can be written as p_nc= p_1 ll + p_2 lb_e + p_3 b_el + p_4 b_eb_e. we switch now from equality for boxes to equality for contexts , using for short the notation @xmath538 meaning that @xmath539 is the context number @xmath4 of a box @xmath19 and similarly @xmath540 and @xmath541 , @xmath542 . the above equality gives for each @xmath4 and @xmath543 : p_cc = p_1 l_c l_c + p_2 l_c e_c+ p_3 e_c l_c + p_4 e_c e_c now , consider the following 4 cases , where due to @xmath525\subset [ b_e , b_{e'}]$ ] we can set @xmath544 for some @xmath545 $ ] . * case 2 . * @xmath548 @xmath549 applying reversible operations @xmath550 to both distributions in the above relative entropy term , we get d(qq\|\|p_1 ( sq+(1-s)\|q)(sq+(1-s)\|q ) + p_2 ( sq+(1-s)\|q ) \|q + p_3 \|q(sq+(1-s)\|q ) + p_4 \|q\|q ) which is exactly relative entropy term in ( [ eq : re - first ] ) . similarly , by considering other two cases where @xmath551 and @xmath552 we get the same equality after applying reversible operations @xmath553 and @xmath554 respectively , and the assertion follows @xmath555 let box @xmath556 and let the image of @xmath98 through @xmath557 be the interval @xmath524 $ ] and the image of set @xmath39 be @xmath525 \subset [ b_e , b_e']$ ] such that @xmath544 with @xmath558 and @xmath559 with @xmath560 . let also @xmath538 and @xmath540 , such that @xmath539 has disjoint support from @xmath561 , then there holds x_u(bb ) = 2 x_u(b ) . [ thm:2add ] we first note that by theorem [ thm : iso - optimal ] , with the set of automorphisms @xmath562 being the set of all tensor products of automorphisms from @xmath185 with themselves , we have : x_u(b^2 ) = _ p_nc(nc_g^2 ) 1n^2_c , c d(g(_c)g(_c ) \|\| p_cc ) now , by lemma [ obs : single_term2 ] , we have x_u(b^2)=_p_nc(nc_g^2 ) d(q q\|\|p_ij ) where @xmath563 ( also @xmath564 ) ( @xmath560 by assumption ) and indices @xmath565 represent a fixed context of @xmath532 such that all distributions of @xmath532 are transformable into it , by operations which at the same time transform all distributions of @xmath566 into @xmath567 . by theorem [ thm : iso - optimal ] and using the fact that @xmath567 is invariant under swap ( it can be achieved by local or global swap operations depending on the hypergraph under consideration ) , it is equivalent to the quantity : x_u(b^2)=_p_1+p_2+p_3=1 d(q q\|\| p_1 l_i l_j + p_22(l_i l_j + l_i l_j ) + p_3 l_i l_j ) note , that we can relax the minimization and hence we have the following lower bound : [ eq:2copy ] x_u(b^2)_p_1+p_2+p_3=1 d(q q\|\| p_1 l_i l_j + p_22(l_i e_j + e_i l_j ) + p_3 e_i e_j ) if we are lucky to find the solution that is non - contextual , then we will find solution to our initial minimization problem . as we will see , this will be the case . using the fact that @xmath568 and @xmath569 have disjoint supports , decomposing @xmath570 and @xmath571 we get that : @xmath572\end{gathered}\ ] ] where @xmath573 is the distribution of @xmath574 . for @xmath575 i.e. when @xmath576 which is the case for pr - box , pm - box , mermin s star and ch - box , we have that x_u(b^2 ) _ p_1+p_2+p_3=1 where lhs is clearly minimal for @xmath577 , which means that the closest distribution in our set is non - contextual , equal to @xmath578 , hence x_u(b^2 ) = = 2 x_u(b ) consider now @xmath579 . here we are able to prove additivity for 2 copies , by using lagrange multipliers approach . we need to find infimum of x_u(b^2 ) _ p_1+p_2+p_3=1 we first check if the infimum is attained in the interior of the simplex of the boundary conditions . using mathematica v07 we obtain , that there is only 1 solution of the set of lagrange equations : p_1 = r^2s^2 , + p_2 = 2 r ( -r + s)s^2 , + p_3 = ( r - s)^2s^2 however , we have that @xmath243 is contextual , so @xmath580 , which gives that @xmath581 of the above solution is negative . hence the function does not have infimum in the interior , in the considered region of parameters @xmath582 and @xmath583 . it suffice to consider boundaries , i.e. cases @xmath584 , @xmath585 and @xmath586 ( other cases are excluded by the fact that @xmath587 ) . again , using lagrange multipliers method , we solve the first two cases . the first one has two solutions , which has @xmath588 if @xmath589 . in case @xmath584 we observe that @xmath590 , which finally proves that the only solution that attributes to infimum is @xmath591 , which is non - contextual solution as in case of extremal q , that yields additivity i.e. @xmath592.@xmath555 in the calculation of relative entropy for this case , we will have similar terms as in eqn . ( [ eq:2copy ] ) but with n copies . by observation [ obs : n - one - term ] , we have : [ eq : ncopy ] x_u(b^n)_p_i=1 d(qq ... \|\| p_1 l_1 l_2 ... +t ) where @xmath595 is all the other possible terms of @xmath596s & @xmath597s with weights @xmath598 . note here that , @xmath596s & @xmath597s are all some fixed context . since , @xmath599 for all @xmath8 we have , x_u(b^n)_p_i=1 d ( r^n e_1 e_2 ... + t_1\|\| p_1 s^n e_1 e_2 ... +t_1 ) where @xmath600 and @xmath601 are all the other possible terms of @xmath597s and @xmath602s with @xmath600 having powers @xmath603 while @xmath601 have different weights @xmath604 . x_u(b^n ) _ p_i=1 _ * a * r^n p_e_1 e_2 .. ^(a ) ( r^n p^(a)_e_1 e_2 .. p_1 s^n p^(a)_e_1 e_2 .. ) + t_2 where @xmath605 contains terms with powers of @xmath603 . for extremal points @xmath575 therefore , x_u(b^n ) _ p_i=1 _ * a * p_e_1 e_2 .. ^(a ) ( p^(a)_e_1 e_2 .. p_1 s^n p^(a)_e_1 e_2 .. ) x_u(b^n)_p_1 _ * a * p_e_1 e_2 .. ^(a ) ( 1p_1 s^n ) since @xmath606 and minimum is attained at @xmath607 which gives us desired proof x_u(b^n)n = nx_u(b ) analogously we can show additivity of @xmath31 on @xmath594 by exchanging @xmath608 to @xmath609 and @xmath610 to @xmath611 . @xmath555 for a box @xmath612 @xmath613 . for @xmath614 @xmath31 is additive .	contextuality is central to both the foundations of quantum theory and to the novel information processing tasks . although it was recognized before bell s nonlocality , despite some recent proposals , it still faces a fundamental problem : how to quantify its presence ? in this work , we provide a framework for quantifying contextuality . we conduct two complementary approaches : ( i ) bottom - up approach , where we introduce a communication game , which grasps the phenomenon of contextuality in a quantitative manner ; ( ii ) top - down approach , where we just postulate two measures - relative entropy of contextuality and contextuality cost , analogous to existent measures of non - locality ( a special case of contextuality ) . we then match the two approaches , by showing that the measure emerging from communication scenario turns out to be equal to the relative entropy of contextuality . we give analytical formulas for the proposed measures for some contextual systems . furthermore we explore properties of these measures such as monotonicity or additivity . _ introduction : _ non - locality is one of the most interesting manifestations of quantumness of physical systems @xcite . it exhibits the strength of correlations that comes out of a quantum state when measured independently by distant parties that share it , which is sometimes higher than that coming from classical resources , and can be even higher for super - quantum but non - signaling resources @xcite . nonlocality has been formulated in terms of boxes i.e. families of probability distribution , and has been studied both qualitatively through bell inequalities as well as quantitatively through measures of non - locality such as cost of non - locality , distillable nonlocality @xcite or recently as its ( anti)robustness @xcite . there is however another phenomenon known even earlier than bell s non - locality , called quantum contextuality @xcite . namely , for certain sets of observables , some of which may be commensurable , their results could not preexist prior to the measurements , or otherwise one would obtain logical contradiction sometimes called as kochen - specker paradox @xcite . in recent years , this phenomenon has been studied in depth . new examples of kochen - specker proofs of contextuality has been found @xcite ( see also @xcite and references therein for recent results ) , and the counterparts of bell inequalities have been introduced , however in a state independent fashion @xcite i.e. that are violated by any quantum state ( see also state dependent attempts of @xcite and @xcite for more recent achievements ) . the fact that quantum theory is contextual has been also treated experimentally @xcite , see also @xcite and references therein for recent results . in fact the phenomenon of non - locality is special case of contextuality : the commensurability relations are provided by the fact that observables are measured on separate systems . yet it is not vice versa : the phenomenon of contextuality is more basic , as can hold in single partite systems . since the discovery of quantum contextuality there has been a basic problem : _ how to quantify contextuality ? _ only recently there were interesting attempts to quantify contextuality in terms of memory cost @xcite and the ratio of contextual assignments @xcite . there were also some measures of non - locality , which is a special case of contextuality such as non - locality cost @xcite and relative entropy of non - locality @xcite . in this paper , we propose a program of quantifying contextuality based on two complementary approaches : ( i ) bottom - up approach , where we introduce a communication game , which grasps the phenomenon of contextuality in a quantitative manner ( ii ) top - down approach , where we just postulate two measures - contextuality cost and relative entropy of contextuality , analogous to the above mentioned non - locality measures . we then match the two approaches , by showing that the measure emerging from communication scenario turns out to be equal to the relative entropy of contextuality . we further study properties of the measures such as faithfulness , additivity or monotonicity , which are analogous to that of entanglement measures . we also compute it for some systems that possess high symmetries . _ how to quantify contextuality : _ quantum contextuality clearly manifests that quantum mechanical world which can not be described by a joint probability distribution over a single probability space : there are systems where statistics of observables ( some of which are jointly measurable - form a _ context _ ) , can not be described by a common joint probability distribution . in other words , joint probability distribution that reproduces statistics of some contexts , see fig . [ context ] a ) , at the same time can not reproduce statistics of other contexts - see fig . [ context ] b ) . for this reason , if we would like to simulate such a system we need _ at least two _ common joint probability distributions - see fig . [ context ] c ) where each of them has to fail in reproducing statistics of some context . thus , for a _ contextual _ systems there are inevitable correlations between the contexts and the common joint probability distributions , while for _ non - contextual _ the `` which context information '' is inaccessible via the joint probability distribution . we will quantify these correlations by means of mutual information since they vanish iff the system is non - contextual . this quantity will be called the _ mutual information of contextuality _ ( mic ) . we further show , that it equals another quantity , that can be viewed as an analogue of relative entropy of entanglement , that we call _ relative entropy of contextuality_. we study properties of this measure , showing it s _ additivity _ for some systems , as well as _ monotonicity _ under some set of operations . we then compute it for some known systems , developing technique of symmetrization . finally , we introduce the measure called _ cost of contextuality _ and compute it for some systems . ) a ) contexts ( here neighboring @xmath0 ) : observables within each context are jointly measurable , so that we can ascribe joint probability within context . b ) ascribing single common joint probability distribution which has marginals equal to that ascribed in a ) is not possible . c ) exemplary possible description of the system : by means of two different common joint probability distributions , each of which does not reproduce statistics of some context : the left that of @xmath1 the right that of @xmath2 . , scaledwidth=35.0% ] to formalize the above ideas , we consider a set of observables @xmath3 some of which are commensurable . each set of mutually commensurable observables we call a _ context _ , and assign to it a number @xmath4 . with each context its joint probability distribution over observables that form it , denoted as @xmath5 . the set of such contexts @xmath6 we call a _ box_. the box is _ non - contextual _ if there exists a joint probability distribution @xmath7 of all observables in @xmath3 , such that it has marginal distributions on each context @xmath4 that are equal to @xmath5 . otherwise we call it _ contextual_. for illustration , the family of contextual boxes we describe here the so called _ chain boxes_. the @xmath8-th chain box , denoted as @xmath9 is based on @xmath8 dichotomic observables @xmath10 , with the @xmath8 contexts defined as neighboring pairs of observables @xmath11 . the distributions of these contexts are fully correlated for all but last context and fully anti - correlated for the last one i.e. @xmath12 @xcite . note that @xmath13 is the well known popescu - rohrlich ( pr ) box . the boxes which have only two types of distributions of contexts : equally weighted strings with parity 0 and equally weighted bit - strings of parity 1 we call _ xor - boxes_. the pair : set of observables and set of contexts , form a _ hypergraph_. the hypergraphs of exemplary xor - boxes that we consider in the paper are depicted on fig . [ grafy ] . _ the `` which context '' game._- to formalize introduction of the mic measure , we consider the following game with three persons : alice and bob ( the sender and receiver ) and charlie ( adversary ) . let the parties preagree on some a priori fixed box @xmath14 in hands of alice . the goal of alice is to communicate a number of a context @xmath4 to bob , through hands of charlie . to this end she chooses the best probability distribution @xmath15 , and sends @xmath4 drawn according to it as a challenge to charlie . charlie is bounded to do the following : create a distribution @xmath16 over all variables in @xmath17 , such that it is compatible with @xmath5 on observables that form context @xmath4 , and send it to bob . the goal of charlie is opposite : to disallow communication of @xmath4 in this way . bob distinguishes between @xmath16 s the best he can . the amount of correlations between alice and bob , given alice s choice of distribution @xmath15 achievable in this game is i_\{p(c)}(b):=__c i(_c p(c)\|cc\|_c ) , [ eq : mutual - form ] which is the _ mutual information of contextuality given a priori statistics @xmath15 _ of a box b. we use here dirac notation only for convenience , meaning a classically correlated system of variables @xmath18 correlated with register holding value @xmath4 . optimizing over strategies of alice , we obtain the _ mutual information of contextuality for a box b _ ( mic ) i.e. the following quantity : i_max(b ) = _ \{p(c ) } i_\{p(c)}(b ) . [ eq : imax ] which reports how much correlations alice and bob can obtain in this game . which have context _ c _ as that of a chosen box b such that he minimizes communication from sender ( s ) to receiver ( r),scaledwidth=35.0% ] we will argue now , that this quantity reports how much contextual is box @xmath19 . suppose first that @xmath19 is _ non - contextual_. then by definition _ there exists a single joint probability distribution _ @xmath20 over all observables in @xmath17 with marginals @xmath5 on contexts @xmath4 , hence @xmath21 . however in case of _ contextual _ box @xmath19 , by definition charlie has to use _ at least two _ joint probability distributions of all observables in @xmath17 , so that on observables of context @xmath4 , the distribution is @xmath5 . thus , by compactness argument , the value @xmath22 is strictly positive . _ ( uniform ) relative entropy of contextuality._- we introduce now another measure based directly on the notion of relative entropy distance , in analogy to measure of non - locality introduced in @xcite . the first variant , called _ relative entropy of contextuality _ is defined on any box @xmath23 as follows : x_max(b):= _ p(c ) _ \{p ( ) } _ c e_g p(c ) d(g(_c ) \|\| p(_c ) ) [ eq : xmax ] where @xmath24 is the relative entropy distance between distributions @xmath5 and @xmath25 @xcite , . the minimization is taken over all distributions @xmath7 over @xmath26 with marginal distribution on context @xmath4 equal to @xmath25 , and supremum is taken over probability distributions @xmath27 on the set of numbers of contexts @xmath28 . a natural quantity is also the one which does not distinguish the contexts , i.e. instead of maximization we set @xmath29 for all @xmath4 : x_u(b):= _ p ( ) _ c e_g 1n d(g(_c ) \|\| p(_c ) ) [ eq : xu ] where @xmath8 is number of contexts . we call it the _ uniform relative entropy of contextuality_. by definition we have @xmath30 but in general these measures are not equal since they differ on _ direct sum _ of a contextual and non - contextual boxes ( see section [ app : not - equal ] ) . at first it seems that mutual information of contextuality and relative entropy of contextuality are different , and it is not clear how they are related . interestingly , one can show that they are _ equal _ to each other ( see theorem [ thm : equivalence ] ) , that is : x_max = i_max . we note here , that @xmath31 and @xmath32 ( and hence @xmath33 according to the above result ) are faithful . _ analytical formulas._- we calculate now the value of @xmath31 and @xmath32 for the boxes called _ isotropic xor - boxes_. to give example of isotropic xor boxes we consider here the isotropic chain boxes : ch^(n ) _ = ch_(n ) + ( 1-)ch_(n) [ eq : fam ] where @xmath34 is the @xmath9 box with correlations and anti - correlations replaced with each other . we just give idea of how to calculate the ( uniform ) relative entropy of contextuality for @xmath35 which is isotropic popescu - rohrlich box denoted as @xmath36 , the detailed proof for other xor - boxes is shown in section [ app : computing ] and [ app : beta ] . the techniques employed are analogous to those used in entanglement theory , including twirling @xcite as well as using symmetries to compute measures based on distance from the set of separable states @xcite , and they were applied in the case of nonlocality e.g. in @xcite . we first compute the value of @xmath31 and then argue , that it equals @xmath32 for the isotropic boxes . the first step is to observe , that for isotropic boxes , in definition of @xmath31 the minimum can be taken only over those probability distributions @xmath7 which give rise to an isotropic box , and @xmath25 is marginal of @xmath7 . to show this , we consider @xmath37 such that @xmath38 , and a group of automorphisms of @xmath19 which can be achieved by operations that transforms @xmath39 into @xmath39 i.e. preserve non - contextuality , call it @xmath40 . the idea is to apply to a box @xmath19 a twirling operation : @xmath41 where @xmath42 is number of different automorphisms @xmath43 which in our case are permutations of contexts @xmath44 , composed with appropriate negations of outputs of observables @xmath45 ( see theorem [ thm : iso - optimal ] ) . let us consider an example of @xmath36 box ( the other examples of isotropic xor - boxes , follow similar lines , see setion [ app : beta ] ) , for which x_u(pr _ ) = _ p()= pr_ 14 _ c d(g(_c)\|\|p(_c ) ) , where @xmath7 runs over distributions which are from the family of isotropic boxes @xcite that are non - contextual . since any non - contextual box compatible with @xmath46 has to satisfy the inequality which is equivalent to chsh inequality @xmath47 ( see section [ app : beta ] ) next step is to observe , that relative entropy does not change under reversible operations such as bit - flip of an output of an observable , ( see lemma [ lem : one - term ] ) , which gives : @xmath48 because all isotropic xor - boxes has the above property , that @xmath49 equals a _ single _ term of relative entropy no matter how many contexts the box @xmath19 has , we have that for these boxes @xmath50 ( see theorem [ thm : xmax - xu ] ) . it is then easy to show , that for @xmath51 there holds x_max(pr_)=x_u(pr _ ) = ( 43^ ) - h ( ) , where @xmath52 is the binary shannon entropy . for @xmath53 , @xmath54 equals the value of @xmath55 according to the above equation . on fig . [ fig : miary_chn ] we present values of measure @xmath31 for chosen chain boxes @xmath56 ( quantum ones provided in @xcite and maximally contextual ones ) . for @xmath56 boxes for @xmath57 : maximally contextual boxes ( upper points , @xmath58 ) ; maximally contextual quantum boxes ( lower points ) with ( i ) odd @xmath8 @xmath59 , ( ii ) even @xmath8 , @xmath60.,scaledwidth=40.0% ] analogous considerations gives @xmath61 for the klyachko et al . @xcite ( kcbs ) box see subsection [ app : computing : k ] . one of the most welcome properties of the measure would be its _ additivity_. in theorem [ thm:2add ] , we show that for families of isotropic xor - boxes @xmath31 and @xmath32 are @xmath62-copy additive i.e. @xmath63 for @xmath64 . for boxes which are extremal within the family of isotropic xor - boxes ( such as @xmath9 , @xmath65 , @xmath66 ) @xmath31 and @xmath32 are additive i.e. that the latter statement is true for any natural @xmath67 . we conjecture however , that proposed measures are additive for all isotropic xor - boxes . another welcome property would be monotonicity of @xmath31 and @xmath32 under operations which preserve contextuality . we answer partially this question showing in subsection [ app : props - of - measures ] that they are non - increasing under a natural subclass of contextuality preserving operations . _ the contextuality cost._-we would like to note , that there is an obvious way to quantify contextuality using strength of violation of some kochen - specker ( ks ) inequality . this approach however is not universal , since there are boxes that are contextual but do not violate this specific ks - inequality . does not violate inequality ( [ eq : beta - n-1 ] ) in with @xmath68 . ] thus we demand that our measure of contextuality @xmath69 should be _ faithful _ i.e. nonzero iff the box is contextual . another approach is to base on some known measures of non - locality and define it properly for all ( also one - partite ) boxes . this leads us to the _ contextuality cost _ , which we define as follows : c(b ) : = \{p\| b = p b_c + ( 1-p)b_nc } where infimum is taken over all decompositions of box @xmath19 into mixture of some non - contextual box @xmath70 and some contextual box @xmath71 . this measure inherits after nonlocality cost the property that it is _ not increasing _ under operations that _ preserve non - contextuality _ , which are the operations satisfying the following axioms : ( i ) transform boxes into boxes ( ii ) are linear ( iii ) preserve consistency ( iv ) transform non - contextual boxes into non - contextual ones . this holds for the same reason for which the anti - robustness of nonlocality is non - increasing under class of _ locality preserving operations _ as it is shown in @xcite . we note also that this measure is by definition faithful , and one can easily compute it using linear programming @xcite , it is however not _ extensive _ i.e. is not proportional to dimension of the system . for the families of isotropic boxes , it can be found analytically namely that @xmath72 , @xmath73 and @xmath74 ( in the same way as it is shown in @xcite that @xmath75 ) . _ conclusions._-we have proposed a framework to quantify contextuality . in particular we have introduced measures of state dependent / independent contextuality which are valid for both the single and many party scenarios . our approach can be developed in different ways . first , one can define analogous measures to @xmath31 and @xmath32 setting variational distance in place of relative entropy . one can also consider a measure defined as @xmath76 , i.e. with changed order of min and sup in ( [ eq : imax ] ) which for non - local boxes has been studied in @xcite . this measure have more communicational meaning than @xmath32 , it is minimal capacity of the channel from sender to receiver under adversary s attack . note , that another way of defining relative entropy of contextuality , would be to consider a quantity defined on a box @xmath19 compatible with graph @xmath37 as @xmath77 , where @xmath78 denotes relative entropy of the boxes @xmath19 and @xmath70 defined operationally via distinguishability of box @xmath19 from box @xmath70 in @xcite . it would be interesting to relate such defined measure with @xmath32 and @xmath79 . note also , that following @xcite it is easy to define and study notion of ( anti)robustness of contextuality . this measure will be used in @xcite . it would be also interesting to investigate possible connection between our measures and entropic tests of contextuality put forward in @xcite ( which have their roots in entropic bell inequalities @xcite ) . finally , we note that our measures can be useful for description of experimental results as they are based on correlations between measurement outcomes rather than on mutual exclusiveness of observables . it is important , since in practice it is very difficult to satisfy the latter condition in experiment . we thank m.t . quintino for helpful comments . a.g . and a.w . thank pawe kurzyski for useful discussion . this work is supported by erc grant qolaps . kh also acknowledges grant bmn 538 - 5300 - 0975 - 12 . pj is supported by grant mpd/2009 - 3/4 from foundation for polish science . part of this work was done in national quantum information centre of gdask . 44 natexlab#1#1bibnamefont # 1#1bibfnamefont # 1#1citenamefont # 1#1url # 1`#1`urlprefix[2]#2 [ 2][]#2 , * * , ( ) . , * * , ( ) . , , , , * * , ( ) , . , * * , ( ) , . , , , , , , * * , ( ) , . , * * , ( ) , . , , , , , ( ) , . , * * , ( ) . , * * , ( ) . , * * , ( ) . , * * , ( ) . , * * , ( ) . , * * , ( ) , . ( ) , . , * * , ( ) , . , , , , * * , ( ) , . ( ) , . ( ) , . , , , , * * , ( ) , . , , , , , * * , ( ) . , , , , , , , , * * , ( ) , . , , , , , , , * * , ( ) , . , , , , , , * * , ( ) , . , , , , , , , ( ) , . , , , , , , ( ) , . , , , ( ) , . , , , , , * * , ( ) , . ( ) , . , , , , ( ) , . , _ _ ( , ) . , * * , ( ) . , * * , ( ) , . , * * , ( ) , . , , , * * , ( ) , . , * * , ( ) , . , _ _ , . , , , * * , ( ) , . , . , * * , ( ) , . , , _ _ , . , , , * * , ( ) , . , * * , ( ) , . , * * , ( ) . , * * , ( ) , . , _ _ ( , ) . , , , , * * , ( ) . , * * , ( ) . , * * , ( ) . , , , , , * * , ( ) , .
the compact radio source sagittarius ( sgr ) a * , residing at the galactic center , is believed to be the best candidate of suppermassive black hole @xcite . based on observations of the orbiting stars , the mass of its central object is measured as over 4 million @xmath4 @xcite . however , it remains to be conclusively proved that sgr a * is a black hole , and the explanations of the broad - band emission are even controversial . various morphology of the emission region , torus , jet , and their combinations , are provided by different models . the observations at millimeter or longer wavelengths can not distinguish them due to the interstellar scattering . however , the interstellar scattering is becoming less severe as the observational wavelength decreases to millimeter band , so that the intrinsic size of sgr a * has been measured at 7 mm and 3 mm @xcite . according to the extrapolated scattering size , the intrinsic structure of sgr a * would be dominant , i.e. , the sub - millimeter vlbi promises to produce unadulterated images of its intrinsic structure . however , lack of baselines results in poor uv - coverage , making the imaging impossible . thus , theoretical modelling of sgr a * will be critical to both planning and interpreting sub - millimeter vlbi observaions @xcite . recently , @xcite reported observations at 1.3 mm with a very - long - baseline interferometry ( vlbi ) array consisting of the arizona radio observatory sub - millimeter telescope ( smt ) , combined array for research in millimeter - wave astronomy ( carma ) , and james clerk maxwell telescope ( jcmt ) . robust detections are obtained on baseline carma - smt at @xmath0 500 km and baseline jcmt - smt at @xmath0 4500 km . an upper limit is yielded on baseline carma - jcmt at @xmath0 3000 km . since these visibility measurements are too few to produce an image , a circular gaussian modelling results in a full width at half maximum ( fwhm ) of @xmath5 , with the scattering effects removed @xcite . however , also pointed out by them , this size is somewhat unexpected , because it is smaller than the black hole shadow size of @xmath6 regardless of the black hole spin @xcite . it might be possible that the emission really comes from the inside of black hole shadow . on the other hand , a structure with a central black hole shadow itself , deviating from the gaussian , might be implied by the visibility measurements . in this paper , we present a general relativistic accretion flow model with plasma wave heating mechanism ( sec.[mod ] ) . the polarization observations can be well explained with appropriate viewing angle and other model parameters set for arbitrary black hole spin . we then tend to use eight potential stations for future 1.3 mm vlbi measurements to perform visibility analysis ( sec.[cov ] ) . we perform a visibility analysis using eight potential stations for future 1.3 mm vlbi measurements . we test the images with recent 1.3 mm vlbi measurements and investigate the implications for the properties of the black hole and accetion flow ( sec.[vis ] ) . comparisons to the earlier work and evaluations on the potential @xmath7 coverage are given in the final section . we follow the magneto - rotational - instability ( mri ) driven accretion flow @xcite where the primary mechanism for generation of turbulence and viscous stress in accretion flows supported by magnetohydrodynamics ( mhd ) simulations @xcite . when deriving the basic equations of the general relativistic accretion flow , we use the energy momentum tensor @xmath8 given as @xmath9 where @xmath10 , @xmath11 , @xmath12 , @xmath13 , @xmath14 , @xmath15 , @xmath16 and @xmath17 are the rest number density , four - velocity , heat - flux four - vector , viscous tensor , pressure , relativistic enthalpy , the mass of the proton and the mass of the electron , respectively . the basic equations for the relativistic hydrodynamics are then derived from the baryon - mass conservation @xmath18 , the energy - momentum conservation @xmath19 and an equation of state . detailed derivations can be found in @xcite . in this study , the boyer - lindquist coordinate is used for the description of the kerr metric . following @xcite and @xcite , we introduce model parameters as the ratio of the total stress to the magnetic field energy density @xmath20 , the ratio of the magnetic field energy density to the gas pressure @xmath21 , dimensionless constant for electron heating rate @xmath22 , constant mass accretion rate @xmath23 , inclination angle @xmath24 , and position angle @xmath25 . an additional parameter of black hole spin @xmath26 is also introduced . the mean black hole mass of @xmath27 , measured from the orbit of the short - period star so-2 for the distance of @xmath28 kpc @xcite , is used . we adopt the simplified form for the viscous tensor component @xmath29 @xcite @xmath30 where @xmath31 , @xmath32 , @xmath33 is the lorentz factor , @xmath34 is the angular velocity of the accretion flow , and @xmath35 is the kinematic viscosity coefficient . we then assume the @xmath36-viscosity as @xmath37 where the product of @xmath20 and @xmath21 represents the @xmath36 coefficient . instead of assuming a keplerian accretion flow co - rotating with the central black hole as was done in previous work , we include the radial momentum conservation in the equation set . i.e. , at any radius , the radial velocity is solved by @xmath38 . the angular velocity is a little lower than the keplerian velocity outside about ten gravitational radii to the center . in small radii , the radial velocity increases significantly and the angular velocity becomes much lower than the keplerian velocity . the accreted plasma passes through a sonic point at a radius inside the innermost stable circular orbit , becomes super - sonic and finally crosses the event horizon of the black hole at a speed of light . we mention that we do nt consider the causality of the accretion flow @xcite . we also adopt the viscous tensor component derived from thin disk and the @xmath36-viscosity coefficient for simplicity . such treatments are enough to generate rough dynamical structure for the accretion flow , which can be used for observational predictions at 1.3 mm wavelength . for future observations at much shorter wavelengths and with higher resolution , causal viscosity should be calculated carefully to make more realistic images of the vicinity of the black hole . here , the viscous heating rate is assumed to be equal to the turbulence cascade rate , @xmath39 , where @xmath40 is the magnetic field amplitude determined by the gas pressure and assumed @xmath21 , @xmath41 is the sound speed , and @xmath42 is the angular half - thickness of the disk . the radiative cooling is considered for electrons only and the viscous heating rate for electrons is assumed to be different from that for protons , and the radiative cooling for electrons is much more effective . we adopt the heating mechanism by turbulent plasma wave @xcite for electrons as @xmath43 , where the timescale @xmath44 , @xmath45 $ ] with @xmath46 as the @xmath24th order modified bessel function of dimensionless temperature @xmath47 , and @xmath48 is the mean electron speed . therefore the energy equation for electrons becomes @xmath49 and the energy equation for both protons and electrons derived from local energy conservation @xmath50 is @xmath51 = \gamma_{\rm vis}^+ - \lambda_{\rm rad}^- , \label{energyeqn}\ ] ] where @xmath52 is the coulomb energy exchange rate between electrons and protons and @xmath53 is the radiative cooling rate calculated from the sum of the energy losses due to the synchrotron radiation , the synchrotron self - comptonization of the soft photons and the bremsstrahlung . the temperatures of protons and electrons can be solved with the eq.[energyeqn ] and eq.[energyeqne ] combined . the dynamical structure of the accretion flow is determined by solving all the physical quantities on the equatorial plane and the scale height from the outermost radius of @xmath54 schwarzschild radii to the event horizon . we assume homogeneous distribution in the vertical direction at any radius and a configuration of magnetic field to be parallel to the velocity field in the upper side of the accretion flow and reversed in the lower side , which forms from the initial field lines in vertical structure and the shearing of the keplerian accretion flow . considering synchrotron emission of only thermal population of electrons together with self - absorption and birefringence effects , full general relativistic radiative transfer along ray trajectories is then performed to give polarized images at any given frequency . corresponding equations can be found in @xcite and @xcite . the model parameters are set in the following procedure . first of all , @xmath20 is fixed to be 0.7 when the turbulence saturates @xcite . secondly , the inclination angle @xmath24 is fixed to be @xmath55 as a typical value to reproduce high linear polarization in sub - millimeter band @xcite . the position angle @xmath25 is chosen as @xmath56 , which can make the accretion flow reproduce the observed position angle of electric vector ( evpa ) from millimeter to near - infrared band @xcite with the mean external rotation measure of @xmath57 adopted @xcite . next , we arbitrarily set a value for spin @xmath26 and choose an appropriate value for @xmath21 . in practice , if @xmath26 is set , @xmath21 should be small enough to keep very low circular polarizations in the sub - millimeter band reported by @xcite . finally , we adjust values for @xmath23 and @xmath22 and calculate the synchrotron radiation of thermal electrons to reproduce the linear polarization degree of @xmath58 in sub - millimeter band @xcite and overall spectrum from centimeter to near - infrared band , fixing the flux density at 1.3 mm to 2.4 jy . in practice , the linear polarization is most sensitive to @xmath23 and the flux density is affected by the combination of @xmath23 , @xmath22 , and @xmath21 . we adopt @xmath59 , which can explain the polarization observations well , as our fiducial model , then choose three sets of model parameters corresponding to three different spins . in detail , those parameters are @xmath60 for a non - rotating black hole , @xmath61 for a mildly - rotating black hole , and @xmath62 for a fast - rotating black hole . we show the global solutions of the three cases in fig.[dynam ] and the spectra predicted by them in fig.[spec ] . the observed emission bump and high linear polarization degree in the sub - millimeter band are predicted by the accretion flow . another jet / outflow component @xcite , which is not shown in these figures , contributes to the emission and depolarization in millimeter and longer wavelengths . for the wavelength of 1.3 mm we are interested in , we assume the emission is dominated by the accretion flow and the jet / outflow component may produce a weak depolarization and a certain change in its evpa . we consider the following eight potential mm - vlbi stations to simulate the 1.3 mm vlbi observations in this paper : hawaii ( h ) , including jcmt and submillimeter array ( sma ) ; smto ( s ) ; carma ( c ) ; the large millimeter telescope ( lmt , l ) on sierra negra , mexico ; chilean station ( a ) , consisting of atacama submillimeter telescope experiment ( aste ) and atacama large millimeter array ( alma ) ; the iram plateau de bure ( pdbi , p ) , france ; the iram pico veleta , spain ( pv , v ) ; and a proposed station locating on mount cook in new zealand ( mt - cook , m ) , one of the candidate positions for telescope establishment in the future . the @xmath7 coverage from these stations is shown in the top - left panel of fig.[uv ] . these stations offer a @xmath7 coverage in almost all the direction and various baseline lengths . however , it takes about 11 hours to complete these tracks due to earth rotation . here we divide the total observing time into six parts , with the interval of @xmath63hr . in fig.[uv ] , they are shown in square(black ) , cross(blue ) , triangle(magenta ) , circle(violet ) , rhombus(green ) , and plus(red ) symbols , respectively . from the beginning to the end . the @xmath7 coverage in the same or neighbour symbol , is considered for a simultaneous observation . we choose three typical sub - coverages from the total observation . _ sub - coverage i _ is a sub - coverage during @xmath64hr ( ut ) , produced by the hawaii ( jcmt included ) , smto , and carma , shown in triangle and circle in the top - right panel of fig.[uv ] . this sub - coverage includes the coverage obtained in the observations by @xcite . _ sub - coverage ii _ is during @xmath65hr ( ut ) from all the stations considered , shown in cross and triangle in the bottom - left panel in fig.[uv ] . in this sub - coverage , a wide range of baseline length of @xmath66 in northwest - southeast direction is included . especially , lengths of @xmath67 are covered by baseline group ac / al / as . in almost its perpendicular direction , i.e. the northeast - southwest direction , comparable lengths are also covered by baselines ch , hs , and hl . simultaneous detections on both perpendicular directions are very important to determine the geometry of the assumed accretion flow , as shown in @xcite and as follows . _ sub - coverage iii _ is also from all the eight stations , but during @xmath68hr ( ut ) , shown in rhombus and plus in the bottom - right panel . in this sub - coverage , a wide range of baseline length of @xmath69 in near east - west direction is covered by baselines cs , ch , hs , hl , and am . the baselines al , as , ac , ah , hm , cm , ms , and lm cover a wide range of directions apart from the east - west , but only in lengths of @xmath70 . for the fiducial model described in sec.[mod ] , with inclination angle @xmath71 , position angle @xmath72 , but with three different black hole spins , @xmath73 , @xmath74 , and @xmath75 , we convolve the images with the scattering ellipse . we adopt mean fwhms of the scattering screen of @xmath76 for the major angular size in milli - arc - second and @xmath77 for the minor one , and the orientation of @xmath78 for the major axis , which is derived from size measurements in centimeter band @xcite . we then perform 2-dimensional fourier transform with special @xmath7 coverage provided by those potential stations . with the sampling of _ sub - coverage i _ , the corresponding visibility profiles are shown in fig.[vispa ] . for each case , the scatter - broadened image is shown in the left . its black hole shadow structure is shown as a white region at the center . the dark - red region beside the shadow shows the accretion flow emission boosted by doppler effect . the predicted visibilities are shown in triangle and circle , with the @xmath7 coverage in @xcite included . as mentioned in @xcite , sgr a * could be considered as observed in its quiescent state since it exhibited the total flux density lower than other measurements . these data can be compared with our predictions within 4 hours , although they were measured on two consecutive days . their measurements of total flux density , correlated flux density on baseline cs , and correlated flux density on baseline hs are marked by filled circle , filled square , and filled triangles , respectively . the upper limit on baseline ch is marked by a downward arrow . the models with @xmath73 and @xmath79 provide good explanations to the visibility measurements . notice that the predicted visibilities in triangle at baselines ch and hs , in similar directions , can not be fitted by a gaussian profile . clear structures of null point , or valley point , are predicted at @xmath80 , which are due to the existence of the black hole shadow @xcite . we note that this structure may be absent in a lot of cases ( see in the following ) . in our fiducial model , however , the appropriate lengths and directions make these two baselines , ch and hs , potential to detect the important structure , if better sensitivity can be achieved for carma . when the black hole is assumed to be fast rotating , the predicted correlated flux densities on baselines ch and hs also increase . this is because that the emission region become more compact with large spin . when the black hole spin increases to @xmath75 , the predicted visibilities on baseline ch nearly exceed the observed upper limit . furthermore , the structure of the valley point can not be recognized . the emission region becomes so compact and totally dominated by the blue - shifted side . therefore , for the fiducial model with @xmath59 , determined by polarization data , the 1.3 mm visibility measurements prefer a black hole spin parameter of @xmath81 . the fiducial model with a relatively mildly rotating black hole can reproduce recent 1.3 mm visibility measurements . the orientation of the accretion flow is constrained by polarization observations , based on the assumption of a large - scale magnetic field in toroidal configuration . however , this specific orientation is not conclusive without direct imaging . in spite of the polarizations , the measurements on baselines ch and hs can be explained by many appropriate combinations of @xmath26 , @xmath24 , and @xmath25 , e.g. , highly - inclined disk with @xmath82 for relatively mild spin , or nearly face - on disk with any position angle and spin . our favorite parameters are relatively mildly - rotating black hole , mild inclination angle , and position angle toward southeast . on the other hand , any model in opposite parameters , i.e. , nearly face - on , nearly egde - on , position angle toward northeast , or extremely high spin , can be an extreme alternative to our favorite model . we focus on the disk orientation , i.e. , @xmath24 and @xmath25 , first . choose one fiducial model with @xmath83 , and set three alternative cases with @xmath84 , @xmath85 , and @xmath86 . we sample the predicted visibilities with the _ sub - coverage ii _ mentioned in sec.[cov ] , which covers comparable baseline lengths in almost perpendicular directions by baseline groups ch / hl / hs and ac / al / as . in particular , baseline al can perform simultaneous observations with baseline hl at @xmath87 , and baselines al and as can perform quasi - simultaneous observations ( difference within 4 hours ) with baselines ch and hs at @xmath88 . the corresponding visibilities are predicted in fig.[vispb1 ] , shown in cross and triangle . it can be found that the visibilities predicted by the two baseline groups make big differences among all the possible cases . we can divide the visibility results into three groups : group i , the predicted visibilities yielded by the two orthogonal baseline groups are comparable , e.g. , the case shown in the top - right panel ; group ii , the predicted visibilities yielded by ac / al / as are in higher correlated flux density compared to those yielded by ch / hl / hs , e.g. , cases in the left two panels ; group iii , opposite to group ii , e.g. , the case in the bottom - right panel . these observations are capable of determining the basic geometry of the intrinsic emission region of sgr a. detailed discussions are as follows . if the results in group i are obtained , the emission region is almost symmetrical to the east - west direction . it is probably the nearly face - on case , similar to that shown in the top - right panel of fig.[vispb1 ] . we show its visibilities by the total @xmath7 coverage in the left panel of fig.[vispb ] . in such a case , the visibilities show a similar profile , with a clear valley along all directions . baselines with appropriate length , @xmath89 in this case , in any direction , such as ch , hs , cl , al , have the potential to detect such structure . however , there are exceptions . cases with higher inclination angle can be included in group i , if the position angle is close to north or south . we show an example with @xmath90 in the right panel of fig.[vispb ] . in this case , the visibilities again show a similar profile along all directions , though for all the baselines we consider here the valley is missing . however , these profiles significantly deviate from gaussian profiles . if the results in group ii are obtained , the nearly face - on cases are excluded , and the position angle is estimated ranging from @xmath91 to @xmath92 , or from @xmath93 to @xmath94 . the structure of valley point may be detected on baselines ch or hs at @xmath95 , except the highly - inclined cases , i.e. , @xmath96 . in general , for a given inclination angle , the differences in correlated flux density between ac / al / as and ch / hl / hs are predicted to reach maxima at @xmath97 . for a given position angle , such differences are predicted to be larger with a higher inclination angle . vlbi observations reported in @xcite are insufficient to produce a unique estimate of the disk orientation in this case . however , earth - aperture synthesis with additional vlbi stations may be able to do so . _ sub - coverage iii _ is complementary to _ sub - coverage ii _ because of the different directions it covers . we sample the predicted visibilities with the new sub - coverage for the two cases in left panels of fig.[vispb1 ] . we then make the same sampling for these two cases by changing their position angles into @xmath98 . predicted visibilities of all these four cases are shown in fig.[vispb4 ] . two baseline groups are considered , am / ch / hl / hs which provides coverage on near east - west directions , and ac / ah / al / as / cl / ls which provides coverage from northeast to northwest . these two baseline groups can perform simultaneous observations with comparable baseline lengths but significantly different directions . in particular , baselines ls and cl track @xmath1 from baseline ch @xmath99 . baseline al tracks nearly perpendicular to baseline hl at @xmath87 . and am and as track nearly perpendicular to baseline am at @xmath100 . with a position angle close to @xmath92 ( or @xmath93 ) , e.g. , the two cases with @xmath72 , the visibilities yielded by the two baseline groups are predicted to be comparable at @xmath101 , while in larger difference with higher inclination angle at @xmath102 , i.e. between ac , as , hm , and am . if the position angle decreases from @xmath92 to @xmath103 ( or from @xmath93 to @xmath104 ) , e.g. , the two cases with @xmath105 , the visibilities yielded by group am / ch / hl / hs are predicted to become much lower than those yielded by group ac / ah / al / as / cl / ls . with higher inclination angle , these differences become larger , in particular , between al and hl at @xmath106 , and between hm ( or as ) and am at @xmath107 . in addition , the valley structure , or deviation from gaussian may appear at baselines hl and hs at @xmath108 . with the two sub - coverages discussed above , we may make an estimation with an error bar as small as @xmath109 of the position angle for an accretion flow of group ii . furthermore , an edge - on case may be easily recognized by visibilities at @xmath110 . cases in group iii are not preferred by the recent 1.3 mm measurements . those measurements have strictly constrained these cases with @xmath111 , @xmath112 , and @xmath113 , so that we wo nt discuss more in this paper . of course , we can further use _ sub - coverage iii _ for confirmation . small values of black hole spin parameter are preferred by recent measurements . however , this is not conclusive because it is based on our fiducial model , with assumptions of specific electrons heating mechanism and magnetic field structure . in @xcite , we adopted the general radiatively inefficient accretion flow based on the work of @xcite and @xcite , and found the recent measurements prefer extremely high values of black hole spin , in stark contrast to the result in this paper . therefore , the spin parameter inferred from observational visibilities depends significantly upon the adopted model . given a specific model , the emission region becomes more compact with the increase of the black hole spin , so that the structure of the visibility valley , if present , moves towards longer baselines . what can be concluded from the visibility measurements is the basic geometry , e.g. , compactness , or asymmetry , of the emission region . however , determining the black hole spin from this is generally model - dependent . we show the images and predicted visibilities of two cases for the fiducial model adopted in this paper ( model a ) in the left column in fig.[vispbspin ] . for comparison , two cases for the model adopted in @xcite ( model b ) with the same parameters of spin and orientation are shown in the right column . it is found that the emission region of model a with a non - rotating black hole is even more compact than that of model b with a fast - rotating black hole , e.g. , @xmath114 . notice the predicted visibilies on baseline ch and hs , in circle and triangle at @xmath115 . and the data of recent measurements . fast - rotating black hole is excluded by model a , but preferred by model b. such a difference is mainly caused by the different electron heating mechanism used . in model a , the heating mechanism is dominated by turbulence in plasma wave . this heating mechanism is so effective that the electron temperature exceeds @xmath116k in the innermost region . the angular velocity of the accretion flow is close to the keplerian velocity , @xmath117 , outside the sonic point . the strong doppler boosting in the blue - shifted side makes the emission region more compact . in model b , the electrons are heated by a small fraction of the viscous heating . the electron temperature keeps lower than @xmath116k even near to event horizon , except in cases with extremely fast - rotating black hole . compared to model a , it shows a flat profile in the innermost region , which extends the emission region with the same spin . moreover , the angular velocity of the accretion flow is much lower than the keplerian velocity , @xmath118 . the much weaker doppler effect makes the emission region more extended . in order to make a good estimation of the black hole spin parameter , we should take advantage of the properties of the space - time geometry that are independent of configuration of the emission region . for example , the shape , size of the black hole shadow , and its horizontal shift from the rotation axis of the black hole @xcite are the same for different models with the same spin parameter , as if the emission region is assumed to extend to the event horizon . therefore , high - fidelity imaging via vlbi at mm / sub - mm wavelengths is critical for this . detecting the innermost stable circular orbit of hot - spot structure , if exits , during flaring activity is another way to constrain the spin parameter directly @xcite . we consider eight potential stations to simulate the future 1.3 mm vlbi observations for sgr a to test the specific accretion flow model with plasma wave heating mechanism . these stations provide a highest angular resolution of @xmath119as , which is comparable to the apparent angular size of the predicted black hole shadow structure . for our fiducial model with @xmath120 and @xmath121 , the recent 1.3 mm vlbi measurements constrain its black hole spin parameter as @xmath122 . more detailed visibility analysis suggests that future vlbi observations can improve these estimates substantially . as mentioned in the previous section , the black hole parameter estimation is dependent on the specific model adopted to fit the observations . this is born out by other efforts to fit the 1.3mm - vlbi results . @xcite followed the general riaf model but adopted a keplerian velocity distribution . they preferred small values of black hole spin . this is different from the prediction of extremely large spin from our model b , which may be caused by different treatments of velocity field and gravitational effect , i.e. @xcite used pseudo - newtonian potential instead of kerr metric . interestingly , their likeliest values of black hole parameter , @xmath123 , are consistent with those preferred in our model a. the keplerian , or near keplerian , velocity field generically makes the emission region more compact , preferring small spin parameter . their most probable combination is with values @xmath124 . this orientation is consistent with the preference of a high inclination angle and a small position angle for the same riaf model , constrained by size measurements in 7 mm and 3 mm @xcite . however , our model a with the same combination is excluded by observational data because it predicts too compact an image for the emission region . this is due to our use of the plasma wave heating mechanism . there are other efforts of estimations on the black hole parameters , especially orientation , based on different observations and theoretical models . e.g. , @xcite derived a high inclination angle and a favorite value of @xmath125 for position angle , by comparing the visibilities observed in 7 mm and predicted by a jet - dominated model ; @xcite derived a high inclination angle and a range of @xmath126 for position angle , by assuming that an orbiting blob contributes to near - infrared flares and a jet component contributes to mm - to - submm emission . these models describe morphologies at 1.3 mm wavelength totally different from an accretion flow , which can only be distinguished by the future direct imaging . in order to obtain a good @xmath7 coverage , we need simultanesous observation on baselines with various lengths and directions . according to theoretical predictions , baseline lengths of @xmath127 are very important because the valley point structure of the visibility profile may be detected . fortunately , there are five potential stations , hawaii ( including jcmt ) , carma , smt , lmt , and the chilean ( including alma ) , can provide such suitable baseline lengths . although the coverage is still poor , they can perform simultaneous observations at similar lengths but at almost perpendicular directions . the basic geometry of the emission region of sgr a * may be estimated by them . furthermore , alma provides excellent sensitivity on baselines as long as @xmath128 ( on baseline ah ) . it is expected to make a critical contribution to future sub - mm vlbi observations of sgr a. the proposed station in new zealand , mt - cook , also makes great contribution . it provides long baselines at a large range of direction . with its participation , these eight stations cover almost all the directions at lengths of @xmath129 . the longest baseline length of @xmath130 , which can provide the highest resolution , is also achieved by mt - cook ( on baselines cm and ms ) . with the earth s rotation , these eight stations produce sufficient @xmath7 coverage to produce images of sgr a. however , it takes almost 11 hours to complete . in this paper , we treat observations within 2 hours as the simultaneous and those within 4 hours as the quasi - simultaneous . the estimate of the basic geometry of the emission region is made from the hypothetic observation of four hours duration , e.g. , with _ sub - coverage i _ mentioned in sec.[cov ] adopted . thus , we have to assume that sgr a * is in its stable state during the entire 4-hour period . if we want to make more detailed estimates using coverage in a larger range of directions , we must assume that this stable state exists even longer . however , a number of flares have been observed in sgr a * at 1.3 mm , on timescales as short as 30 minutes . therefore , reliable estimates can only be made when either sgr a * is quiescent , the flaring region is subdominant at 1.3 mm , or the observation lasts for @xmath131 hour ( resulting in poorer coverage and less opportunity to detect the visibility valley associated with the black hole shadow ) . there is one more short - coming in the @xmath7 coverage . region with baseline lengths of @xmath129 is covered in almost all the directions . however , the most important region with baseline lengths of @xmath132 is lack of coverage in northeast - southwest direction . during any simultaneous observation within 2 hours , coverage in two perpendicular directions can be obtained . however , at least one more direction between those two is necessary for a good estimation of the emission region geometry , also for a higher possibility of detection on the valley point structure . this can only be improved with an additional new station included , e.g. , one in new zealand . this work was supported in part by the national natural science foundation of china ( grants 10573029 , 10625314 , 10633010 , and 10821302 ) and the knowledge innovation program of the chinese academy of sciences ( grant no . kjcx2-yw - t03 ) , and sponsored by the program of shanghai subject chief scientist ( 06xd14024 ) and the national key basic research development program of china ( no.2007cb815405 ) . is supported by china postdoctoral science foundation ( grant 20090450822 ) . we are grateful to dr . s. liu , dr . m. j. cai , and prof . r. e. taam for comments on the modelling , to prof . p. t. p. ho for suggestions in visibility analysis , and to dr . k. asada for providing the uv - coverage . 99 abramowicz , m. a. , chen , x. -m . , granath , m. , & lasota , j. -p . , 1996 , apj , 471 , 762 aitken , d. k. , et al . 2000 , apj , 534 , l173 balbus , s. a. , & hawley , j. f. , 1991 , apj , 376 , 214 balbus , s. a. , & hawley , j. f. , 1998 , rev . phys . , 70 , 1 blandford , e. , & eichler , d. , 1987 , phys . report , 154 , 1 bower , g. c. , et al . 2004 , science , 304 , 704 bower , g. c. , falcke , h. , wright , m. c. h. , & backer d. c. 2005 , apj , 618 , l29 broderick , a. e. , & loeb , a. , 2005 , apj , 363 , 353 broderick , a. e. , & loeb , a. , 2006 , apj , 367 , 905 broderick , a. e. , fish , v. l. , doeleman , s. s. , & loeb , a. , 2009 , apj , 697 , 45 doeleman , s. s. , et al . , 2008 , nat , 455 , 78 doeleman , s. s. , fish , v. l. , broderick , a. e. , loeb , a. , & rogers , a. e. e. , 2009 , apj , 698 , 59 falcke , h. , melia , f. , agol , e. , 2000 , apj , 528 , l13 fish , v. l. , broderick , a. e. , doeleman , s. s. , & loeb , a. , 2009 , apj , 692 , l14 gammie , c. f. & popham , r. , 1998 , apj , 498 , 313 ghez , a. m. , et al . 2005 , apj , 620 , 744 ghez , a. m. , et al . 2008 , apj , 689 , 1044 gillessen , s. , eisenhauer , f. , trippe , s. , genzel , t. , martins , f. , & ott , t. , 2009 , apj , 692 , 1075 huang , l. , cai , m. , shen , z. -q . , & yuan , f. , 2007 , mnras , 379 , 833 huang , l. , liu , s. , shen , z. -q . , cai , m. j. , li , h. , & fryer c. l. , 2008 , apj , 676 , l119 huang , l. , liu , s. , shen , z. -q . , yuan , y .- f . , cai , m. j. , li , h. , & fryer c. l. , 2009 , apj , 703 , 557 lasota , j .- p . , 1994 , in theory of accretion disks-2 , ed . w. j. duschl , j. frank , f. meyer , e. meyer - hofmeister , & w. m. tscharnuter ( dordrecht : kluwer ) , 341 liu , s. , petrosian , v. , melia , f. , & fryer , c. l. , 2006 , apj , 648 , 1020 liu , s. , qian , l. , wu , x .- b . , fryer , c. l. , & li , h. , 2007 , apj , 668 , l127 macquart , j. -p . , bower , g. c. , wright , m. c. h. , backer , d. c. , falcke , h. , 2006 , apj , 646 , 111 manmoto , t , 2000 , apj , 534 , 734 markoff , s. , bower , g. c. , falcke , h. , 2007 , mnras , 379 , 1519 marrone , d. p. , moran , j. m. , zhao , j. -h . , & rao , r. 2006 , apj , 640 , 308 marrone , d. p. , moran , j. m. , zhao , j. -h . , & rao , r. 2007 , apj , 654 , 57 melia , f. , liu , s. , & coker , r. 2001 , apj , 553 , 146 meyer , l. , et al . 2007 , a&a , 473 , 707 miyoshi , m. , kameno , s. , ishitsuka , j. k. , shen , z .- q . , takahashi r. & horiuchi , s. , 2007 , publ . japan , 10 , 15 pessah , m. e. , chan , c. k. , & psaltis , d. 2006 , phys . lett . , 97 , 221103 schdel , r. , et al . 2002 , nature , 419 , 694 shen , z .- q . , lo , k. y. , liang , m .- c . , ho , p.t.p . , & zhao , j. -h . 2005 , nat , 438 , 62 takahashi , r. , 2004 , apj , 611 , 996 takahashi , r. , 2007 , mnras , 382 , 567 yuan , f. , quataert , e. , & narayan , r. 2003 , apj , 598 , 301 yuan , y .- f . , cao , x .- w . , huang , l. , & shen , z .- q . , apj , 699 , 722 . the dash - three - dotted lines are for @xmath133 . the dotted lines are for @xmath134 . _ top - left _ : radial velocity and sound speed ; _ top - middle _ : angular momentum , with the keplerian one in solid grey line ; _ top - right _ : ratio of scale height to radius ; _ bottom - left _ : temperatures of protons and electrons ; _ bottom - middle _ : number density of electrons ; _ bottom - right _ : amplitude of magnetic field . [ dynam],width=529 ] , @xmath71 , and @xmath72 . _ top _ : spectrum of total flux density of synchrotron emission ; _ middle _ : linear polarization degree of synchrotron emission ; _ bottom _ : position angle of electric vector ( evpa ) with an external rotation measure of @xmath135 considered . [ spec],width=529 ] coverage produced by eight potential stations at 1.3 mm wavelength . h : hawaii , including jcmt and sma , s : smto , c : carma , l : lmt , a : chilean , including aste and alma , p : pdbi , v : pv , m : mt - cook . tracks in different parts of observational time are shown in different symbols ( square , cross , triangle , circle , rhombus , and plus ) , with the interval of @xmath136hr . _ top - right _ : _ sub - coverage i _ during @xmath64hr ( ut ) , produced by h , s , and c. _ bottom - left _ : _ sub - coverage ii _ during @xmath137hr ( ut ) , produced by all the eight stations . _ bottom - right _ : _ sub - coverage iii _ during @xmath138hr ( ut ) , produced by all the eight stations . [ uv],title="fig:",width=283 ] coverage produced by eight potential stations at 1.3 mm wavelength . h : hawaii , including jcmt and sma , s : smto , c : carma , l : lmt , a : chilean , including aste and alma , p : pdbi , v : pv , m : mt - cook . tracks in different parts of observational time are shown in different symbols ( square , cross , triangle , circle , rhombus , and plus ) , with the interval of @xmath136hr . _ top - right _ : _ sub - coverage i _ during @xmath64hr ( ut ) , produced by h , s , and c. _ bottom - left _ : _ sub - coverage ii _ during @xmath137hr ( ut ) , produced by all the eight stations . _ bottom - right _ : _ sub - coverage iii _ during @xmath138hr ( ut ) , produced by all the eight stations . [ uv],title="fig:",width=283 ] + coverage produced by eight potential stations at 1.3 mm wavelength . h : hawaii , including jcmt and sma , s : smto , c : carma , l : lmt , a : chilean , including aste and alma , p : pdbi , v : pv , m : mt - cook . tracks in different parts of observational time are shown in different symbols ( square , cross , triangle , circle , rhombus , and plus ) , with the interval of @xmath136hr . _ top - right _ : _ sub - coverage i _ during @xmath64hr ( ut ) , produced by h , s , and c. _ bottom - left _ : _ sub - coverage ii _ during @xmath137hr ( ut ) , produced by all the eight stations . _ bottom - right _ : _ sub - coverage iii _ during @xmath138hr ( ut ) , produced by all the eight stations . [ uv],title="fig:",width=283 ] coverage produced by eight potential stations at 1.3 mm wavelength . h : hawaii , including jcmt and sma , s : smto , c : carma , l : lmt , a : chilean , including aste and alma , p : pdbi , v : pv , m : mt - cook . tracks in different parts of observational time are shown in different symbols ( square , cross , triangle , circle , rhombus , and plus ) , with the interval of @xmath136hr . _ top - right _ : _ sub - coverage i _ during @xmath64hr ( ut ) , produced by h , s , and c. _ bottom - left _ : _ sub - coverage ii _ during @xmath137hr ( ut ) , produced by all the eight stations . _ bottom - right _ : _ sub - coverage iii _ during @xmath138hr ( ut ) , produced by all the eight stations . [ uv],title="fig:",width=283 ] , @xmath74 , and @xmath75 , with the sampling of _ sub - coverage i_. the observational data are from @xcite . [ vispa],title="fig:",width=283 ] + , @xmath74 , and @xmath75 , with the sampling of _ sub - coverage i_. the observational data are from @xcite . [ vispa],title="fig:",width=283 ] + , @xmath74 , and @xmath75 , with the sampling of _ sub - coverage i_. the observational data are from @xcite . [ vispa],title="fig:",width=283 ] but different orientation , with the sampling of _ sub - coverage ii_. in these four cases , the visibilities yielded by baseline group ac / al / as can be higher than , lower than , or comparable to the visibilities yielded by baseline group ch / hl / hs , see in context for details . [ vispb1],title="fig:",width=283 ] but different orientation , with the sampling of _ sub - coverage ii_. in these four cases , the visibilities yielded by baseline group ac / al / as can be higher than , lower than , or comparable to the visibilities yielded by baseline group ch / hl / hs , see in context for details . [ vispb1],title="fig:",width=283 ] + but different orientation , with the sampling of _ sub - coverage ii_. in these four cases , the visibilities yielded by baseline group ac / al / as can be higher than , lower than , or comparable to the visibilities yielded by baseline group ch / hl / hs , see in context for details . [ vispb1],title="fig:",width=283 ] but different orientation , with the sampling of _ sub - coverage ii_. in these four cases , the visibilities yielded by baseline group ac / al / as can be higher than , lower than , or comparable to the visibilities yielded by baseline group ch / hl / hs , see in context for details . [ vispb1],title="fig:",width=283 ] but different orientation combinations , with the sampling of the total @xmath7 coverage . in both the cases , the visibilities yielded by baseline group ac / al / as are comparable to the visibilities yielded by baseline group ch / hl / hs . [ vispb],title="fig:",width=283 ] but different orientation combinations , with the sampling of the total @xmath7 coverage . in both the cases , the visibilities yielded by baseline group ac / al / as are comparable to the visibilities yielded by baseline group ch / hl / hs . [ vispb],title="fig:",width=283 ] but different orientation combinations , with the sampling of _ sub - coverage iii_. in these four cases , the visibilities yielded by baseline group am / ch / hl / hs can be higher than , lower than , or comparable to the visibilities yielded by baseline group ac / ah / al / as / cl / ls , see in context for details . [ vispb4],title="fig:",width=283 ] but different orientation combinations , with the sampling of _ sub - coverage iii_. in these four cases , the visibilities yielded by baseline group am / ch / hl / hs can be higher than , lower than , or comparable to the visibilities yielded by baseline group ac / ah / al / as / cl / ls , see in context for details . [ vispb4],title="fig:",width=283 ] + but different orientation combinations , with the sampling of _ sub - coverage iii_. in these four cases , the visibilities yielded by baseline group am / ch / hl / hs can be higher than , lower than , or comparable to the visibilities yielded by baseline group ac / ah / al / as / cl / ls , see in context for details . [ vispb4],title="fig:",width=283 ] but different orientation combinations , with the sampling of _ sub - coverage iii_. in these four cases , the visibilities yielded by baseline group am / ch / hl / hs can be higher than , lower than , or comparable to the visibilities yielded by baseline group ac / ah / al / as / cl / ls , see in context for details . [ vispb4],title="fig:",width=283 ] and @xmath75 , with the sampling of the total @xmath7 coverage . _ right _ : images and visibilities predicted by radiatively inefficient accretion flow model adopted in @xcite ( model b ) , with the same spin , orientation , and @xmath7 sampling as the left ones . [ vispbspin],title="fig:",width=283 ] and @xmath75 , with the sampling of the total @xmath7 coverage . _ right _ : images and visibilities predicted by radiatively inefficient accretion flow model adopted in @xcite ( model b ) , with the same spin , orientation , and @xmath7 sampling as the left ones . [ vispbspin],title="fig:",width=283 ] + and @xmath75 , with the sampling of the total @xmath7 coverage . _ right _ : images and visibilities predicted by radiatively inefficient accretion flow model adopted in @xcite ( model b ) , with the same spin , orientation , and @xmath7 sampling as the left ones . [ vispbspin],title="fig:",width=283 ] and @xmath75 , with the sampling of the total @xmath7 coverage . _ right _ : images and visibilities predicted by radiatively inefficient accretion flow model adopted in @xcite ( model b ) , with the same spin , orientation , and @xmath7 sampling as the left ones . [ vispbspin],title="fig:",width=283 ]	the vicinity of the supermassive black hole associated with the compact radio source sagittarius ( sgr ) a * is believed to dominate the observed emission at wavelengths near and shorter than @xmath0 1 millimeter . we show that a general relativistic accretion flow , heated via the plasma wave heating mechanism , is consistent with the polarization and recent mm - vlbi observations of sgr a * for an inclination angle of @xmath1 , position angle of @xmath2 , and spin @xmath3 . structure in visibilities produced by the black hole shadow can potentially be observed by 1.3 mm - vlbi on the existing hawaii - carma and hawaii - smt baselines . we also consider eight additional potential mm - vlbi stations , including sites in chile and new zealand , finding that with these the basic geometry of the emission region can be reliably estimated .
recently , novel spatially inhomogeneous and/or anisotropic phases have been found to occur in various condensed matter systems . similar to classical liquid crystals,@xcite these quantum phases can be classified according to their broken symmetries . one such phase is the smectic phase , often referred to as a stripe phase . this phase can be viewed as an unidirectional charge density wave phase which spontaneously breaks translational symmetry along one direction . experimental evidence for smectic phases has been found in several transition - metal oxide compounds . in particular , a stripe crystal phase has been observed in manganese oxide compounds by electron diffraction.@xcite moreover , neutron scattering experiments have provided indirect evidence for stripe phases in the high-@xmath0 cuprate superconductors.@xcite another such phase is the nematic phase an anisotropic phase with broken rotational symmetry , but with translational symmetry in tact . this phase can be thought of as a smectic phase `` melted '' via the proliferation of topological defects . nematic phases have been argued to be the source of the anisotropic transport observed in quantum hall systems in high landau levels@xcite , and to be relevant to the high-@xmath0 cuprate superconductors as well.@xcite however , an understanding of the transition from the smectic to nematic phase due to the proliferation of the topological defects still remains to be achieved . recently , a complementary approach to studying the nematic phase was suggested.@xcite in this approach , the nematic phase is obtained from an isotropic liquid by breaking rotational symmetry . in the continuum limit , it was shown that the goldstone mode resulting from breaking rotational symmetry leads to non - fermi liquid behavior in the single - particle lifetime.@xcite exotic superconducting instabilities mediated by the collective mode in the nematic phase were also studied.@xcite while the consequences of the nematic order in the recent complementary approach were studied in ref . , we still lack an understanding of the phase transition between the isotropic and nematic phases . in this work , we study a model for the nematic phase on a square lattice , with emphasis on the phase transitions between the isotopic and nematic phases . furthermore , we study the conductivity tensor and the hall constant as probes of nematic phase and its transition . the hamiltonian we consider is @xmath1 . \label{model}\ ] ] in eq . [ model ] , @xmath2 is the single - particle dispersion ; @xmath3 is the inter - electron interaction strength , which can be written as @xmath4 @xmath5 is the quadrupole density tensor @xmath6 where @xmath7 is the number of lattice sites . in what follows , we take @xmath2 to arise from a tight - binding model on a square lattice with nearest - neighbor ( @xmath8 ) and next - nearest neighbor ( @xmath9 ) hopping integrals @xmath10 where @xmath11 is the chemical potential , and we have set the lattice spacing to unity , @xmath12 . within a mean field approximation , we have the self - consistency equations @xmath13 where @xmath14 is the fermi function , and @xmath15 is the electron density . the expectation values @xmath16 and @xmath17 lead to a modification of the single - particle dispersion . defining @xmath18 the single particle dispersion becomes @xmath19 \nonumber\\ & - & 4 t^{\prime } ( \cos{k_x } \cos{k_y } + \lambda^{\prime } \sin{k_x } \sin{k_y } ) - \mu \ , . \label{band}\end{aligned}\ ] ] for given values of @xmath11 , @xmath20 , and @xmath9 , we solved for @xmath21 , @xmath22 and the density @xmath23 ( @xmath24 is the hole concentration ) using eq . [ nematic ] . for all values of the parameters we considered , we found @xmath25 . in fig . [ fig : lambdaplot ] , we show the behavior of the nematic order @xmath21 as a function of chemical potential , @xmath11 for @xmath26 , @xmath27 , and @xmath28 at @xmath29 . an important feature of fig . [ fig : lambdaplot ] is the sharp onset and offset of nematic order at critical values of @xmath11 : @xmath30 and @xmath31 . the free energy near the transition @xmath32 and @xmath33 shows three minima at finite @xmath34 and @xmath35 . this indicates a region of coexistence of the isotropic and nematic phases : it reflects the negative quartic term in the ginzburg - landau free energy.@xcite hence , the transition between the isotropic and nematic phases is first order . the electron density as a function of @xmath11 shows a discontinuity at these critical values of @xmath36 , which is a signature of the first order transition . to better understand the behavior of the nematic order , as well as the robustness of our result , we considered the dependence of @xmath21 on the parameters @xmath20 and @xmath9 . upon increasing @xmath37 , we found the maximum value of @xmath21 to increase . also , increasing @xmath37 shifts the critical values of @xmath11 , making the region over which nematic order exists wider . furthermore , we found an absence of nematic order for all doping concentrations if @xmath38 ( within the numerical precision ) for @xmath27 . by varying @xmath9 , we found the structure shown in fig . [ fig : lambdaplot ] to shift . in particular , when @xmath39 , @xmath21 is maximal at @xmath40 ( _ i.e. _ half - filling ) and is a symmetric function of @xmath11 . this is a consequence of the particle - hole symmetry present when @xmath39 . though the critical values of @xmath11 and the maximal value of @xmath21 depend on the value of @xmath20 and @xmath9 , we would like to stress that the qualitative features of the phase transition between the isotropic and nematic liquids are not sensitive to the choice of parameters . in what follows , we will present results for @xmath41 , and @xmath27 ( @xmath42 ) . for small hole concentrations , this gives a hole - like fermi surface centered about @xmath43 ( see inset ( a ) of fig . [ fig : nhallplot ] ) . such a fermi surface is consistent with angle resolved photoemission spectroscopy@xcite and hall constant measurements @xcite in the high-@xmath0 cuprates . above , we saw that the nematic phase occurs in a finite regime dopings . furthermore , we found a dramatic change in the fermi surface topology upon developing nematic order . ( see the inset of fig . [ fig : nhallplot ] . ) in this section , we study the conductivity tensor to probe the nematic order , as the conductivity tensor , especially the hall conductivity is sensitive to the topology of the fermi surface . within a boltzmann equation approach , the longitudinal conductivities are ( in units where @xmath44 ) @xmath45 where @xmath46 = @xmath24 or @xmath47 , @xmath48 is the electron s charge , @xmath49 is the component of the fermi velocity along the @xmath46-axis , and @xmath50 is the transport lifetime . in the presence of a magnetic field @xmath51 parallel to the @xmath52-axis , the hall conductivity is@xcite @xmath53 in what follows , we will report results for the case @xmath54 . some discussion regarding the form of @xmath50 is given in the following section . using the single - particle dispersion in eq . [ band ] with @xmath21 determined from eq . [ nematic ] , we plot the conductivity tensor vs. @xmath11 in fig . [ fig : sigmaplot ] . for comparison , the conductivity tensor without nematic order ( @xmath55 ) is shown as a dashed line . when nematic order develops , the fermi surface topology changes dramatically . more specifically , the hole - like fermi surface with nematic order becomes an open fermi surface upon developing nematic order . in the nematic state , the fermi surface is rather flat along the @xmath56-direction . ( see inset ( b ) of fig . [ fig : nhallplot ] . ) hence , the hole s motion along the @xmath57-direction is enhanced , while it is suppressed along the @xmath56-direction . these changes in the fermi surface are reflected in the longitudinal conductivities , figs . [ fig : sigmaplot](a ) and [ fig : sigmaplot](b ) @xmath58 increases suddenly with the onset of nematic order ( at @xmath59 ) , while @xmath60 shows a sudden decrease . the hall conductivity , fig . [ fig : sigmaplot](c ) , shows a dramatic drop at @xmath59 as well . however , its change is small compared with @xmath58 and @xmath60 . this occurs because the hall conductivity is not only sensitive to the fermi surface s topology , but also its curvature.@xcite as shown by the dashed line in fig . [ fig : sigmaplot](c ) , the hall conductivity already decreases fairly rapidly without nematic order . the dotted lines at the transition indicate that boltzmann equation , eq . [ boltz ] is not applicable near the transition , since there are two phases with different densities coexist . another quantity of interest is the hall constant @xmath61 , which measures the number of charge carriers , and is related to the hall resistivity . for weak magnetic fields , the hall resistivity is @xmath62 which defines @xmath61 in terms of @xmath63 . for the case of a single - band fermi liquid model with a circular ( or elliptical ) fermi surface , @xmath61 is the density of charge carriers per unit cell _ i.e. _ the area of the fermi surface . in fig . [ fig : nhallplot ] , we plot the hall constant , @xmath61 as a function of chemical potential , @xmath11 . for comparison , the hall number without nematic order ( @xmath55 ) is shown as a dashed line . notice the sharp drop of @xmath61 at @xmath59 , where nematic order becomes finite . it is expected that @xmath61 is almost constant in the region where nematic order exists , @xmath64 . this can be understood from the conductivity tensor in fig . [ fig : sigmaplot ] the decrease of hole motion along the @xmath47 direction is compensated by the increase of hole motion along @xmath24 direction , giving a product @xmath65 which is nearly constant . to begin with , our results for the phase transition in fig . [ fig : lambdaplot ] were obtained within a mean - field approximation . once fluctuations about the mean - field state are included , we expect the range over which the nematic phase exists to be reduced . furthermore , fluctuations will reduce the value of @xmath21 . the effect of fluctuations on the nature of phase transition will be addressed in near future . another approximation in our treatment was the neglect of collective modes . in the continuum limit of eq . [ model ] , the breaking of rotational symmetry leads to a goldstone mode ; scattering from this goldstone mode gives rise to non - fermi liquid behavior.@xcite on a lattice , however , this collective mode is gapped ; we expect scattering from the collective mode to have little effect on the @xmath66 conductivities . the subject of disorder deserves some discussion . though disorder was introduced to understand the behavior of the conductivity through the transitions , we did not address its effects on the transition between the isotropic and nematic phases , nor did we address its consequences on the nematic phase itself.@xcite in the absence of disorder , the mean - field treatment of the hamiltonian eq . [ model ] gave the transition between the isotropic and nematic phases to be first order . for systems that undergo a first order transition without disorder , it was shown that the system with quenched disorder could be mapped onto a random field ising model.@xcite based on this mapping , it was shown that in @xmath67 dimensions , the coexistence at the phase transition disappears for arbitrarily small amount of disorder ; in @xmath68 dimensions this occurs by the application of a finite amount of disorder . for the transition in this work , however , there are gapless fermions presents on both sides of the transition . therefore it is unclear to us how much the results obtained from the mapping may apply . we hope to address these issues in the future . it is also important to note that the conductivity tensor depends of the form of the transport scattering rate , and hence on the disorder . realistic models for disorder usually give transport scattering rates which are momentum dependent . in particular , in most high-@xmath0 materials , disorder is mainly due to the dopant ions , which lie outside of the @xmath69 planes . as a result , these dopant ions cause primarily forward scattering . in the above discussion , we described our results for the case @xmath70 . however , we have also considered the case where @xmath71 . @xcite such a form arises for out - of - plane disorder , causing primarily forward scattering.@xcite when @xmath55 , we found the overall features of @xmath61 to be very different from @xmath72 . however , the effect of the nematic order on the hall constant gives the same result as when @xmath70 there is a sharp drop in the hall constant in the nematic phase compared with the isotropic phase . our findings for the hall constant have potential relevance to the experiments of the normal state hall resistivity in the high-@xmath0 superconductor , bi@xmath73sr@xmath74la@xmath75cuo@xmath76.@xcite in these experiments , the hall constant was measured in the normal state at low temperatures , by suppressing superconductivity with a large pulsed magnetic field of 60 t. they observed saturation of the hall constant at low temperatures , suggesting conventional transport properties . however , the hall constant as a function of doping concentration shows a maximum at @xmath77 , and decreased sharply in underdoped regime . in overdoped regions , there is a regime near optimal doping @xmath78 , where the hall constant shows the dip the decrease is about @xmath79 ( @xmath80 ) of its maximum value ( @xmath81 ) at @xmath77 . the dip of the hall constant around @xmath82 becomes more pronounced at low temperatures ( @xmath83 ) . this suggests that the nematic phase may be a candidate to explain the anomaly observed in the hall constant in high @xmath0 cuprates . in conclusion , to study the nematic phase we considered the hamiltonian in eq . [ model ] with quadrupolar density - density interactions on a square lattice . within a mean - field analysis , we found the nematic phase to occur in a finite region of dopings . moreover , the transition between the isotropic and nematic phases was found to be first order ; a dramatic change in the fermi surface topology accompanied this transition . as the conductivity is sensitive to the topology of the fermi surface , we investigated the effects of nematic order on transport . we found the conductivity tensor and hall constant to show significant changes at the transitions between the isotropic and nematic phases . if observed , such signatures would give evidence for nematic order ; it would give evidence for the approach pursued in this paper as an appropriate description of nematic phases in physical systems . finally , we discussed the effect of disorder and the scattering from collective modes on the phase transitions and on the transport properties at the transitions . _ acknowledgments _ we would like to thank s. a. kivelson , v. oganesyan p. coleman , y - b kim , and e. fradkin for insightful discussions , and yoichi ando for useful comments . we also thank p. fournier and w. metzner for pointing out ref . and ref . , respectively . this work was supported by the nserc of canada(hyk , ehk , chc ) , canada research chair(hyk ) , canadian institute of advanced research(hyk ) , alfred p. sloan research fellowship ( hyk ) , and emerging material knowledge program funded by materials and manufacturing ontario ( hyk ) . , p. de gennes ( oxford university press , 1987 ) . s. mori , c. h. chen , and s. w. cheong , nature , * 392 * , 473 ( 1998 ) . j. tranquada , b. j. sternlieb , j. d. axe , y. nakamura , and s. uchida , nature , * 375 * , 561 ( 1995 ) . lilly , k. b. cooper , j. p. eisenstein , l. n. pfeiffer , and k. w. west , phys . * 82 * , 394 ( 1999 ) ; r. r. du , d. c. tsui , h. l. stormer , l. n. pfeiffer , k. w. baldwin , k. w. west , solid state comm . * 109 * , 389 ( 1999 ) ; k. b. cooper , m.p . lilly , j. p. eisenstein , l. n. pfeiffer , k. w. west , phys . rev . b * 65 * , 241313 ( 2002 ) . e. fradkin and s. a. kivelson , phys . rev . b * 59 * , 8065 ( 1999 ) . s. a. kivelson , e. fradkin , and v. j. emery , nature * 393 * , 550 ( 1998 ) . v. oganesyan . s. kivelson , and e. fradkin , phys . rev . b * 64 * , 195109 ( 2001 ) . c. j. halboth , and w. metzner , phys . rev . lett . * 85 * , 5162 ( 2000 ) ; a. neumayr , and w. metzner , phys . b * 67 * , 035112 ( 2003 ) ; v. hankevych , and f. wegner , eur . j. b * 31 * , 333 ( 2003 ) . y. b. kim and h .- y . kee , cond - mat/0204037 . kee , phys . b , * 67 * , 073105 ( 2003 ) . s. a. kivelson , e. fradkin , v. oganesyan , i. p. bindloss , j. m. tranquada , a. kapitulnik , and c. howald , cond - mat/0210683 , to be published in rev . mod . phys . c. h. chung , h .- y . kee , and e. h. kim , unpublished . see the review article : j. c. campuzano , m. r. norman , and m. randeria , in in _ physics of conventional and unconventional superconductors _ , edited by k. h. bennemann and j. b. ketterson ( springer - verlag ) . h. y. hwang , b. batlogg , h. takagi , h. l. kao , j. kwo , r. j. cava , j. j. krajewski , and w. f. peck , jr . lett . * 72 * , 2636 ( 1994 ) . n. p. ong , phys . b * 43 * , 193 ( 1991 ) . j. ziman , _ electrons and phonons _ ( clarendon , oxford 1960 ) . the effect of the disorder on electronic liquid crystal is one of the interesting issues . for the classical liquid crystal in the continuum , see l. radzihovsky and j. toner , phys . b * 60 * , 206 ( 1999 ) . j. cardy , cond - mat/9806355 ; see the references therein . kee , phys . b * 64 * , 012506 ( 2001 ) . f. f. balakire , j. b. betts , s. ono , t. murayama , y. ando , and g. s. boebinger , in _ physical phenomena at high magnetic fields - iv _ edited by g. boebinger , z. fisk , l. p. gorkov , a. lacerda , and j. r. schrieffer ( world scientific , singapore , 2002 ) . a theoretical interpretation was given in s. chakravarty , c. nayak , s. tewari , and x. yang , phys . lett . * 89 * , 277003 ( 2002 ) .	the electronic nematic phase occurs when the point - group symmetry of the lattice structure is broken , due to electron - electron interactions . we study a model for the nematic phase on a square lattice with emphasis on the phase transition between isotropic and nematic phases within mean field theory . we find the transition to be first order , with dramatic changes in the fermi surface topology accompanying the transition . furthermore , we study the conductivity tensor and hall constant as probes of the nematic phase and its transition . the relevance of our findings to hall resistivity experiments in the high-@xmath0 cuprates is discussed .
the exploitation of a high power coherent thz source as a part of the lcls - ii user facility has been proposed in the lcls - ii cdr @xcite . thz radiation pulses can be generated by the spent electron beam downstream of the x - ray undulator . in this way , intrinsic synchronization with the x - ray pulses can be achieved . a first , natural application of this kind of photon beams is for the pump - probe experiments . through the combination of thz pump and x - ray probe , xfels would offer unique opportunities for studies of ultrafast surface chemistry and catalysis @xcite . also , the lcls team started an @xmath9 project on thz radiation production from the spent electron beam downstream of the lcls baseline undulator @xcite-@xcite . in that case , thz pulses are generated by inserting a thin beryllium foil into the electron beam . in this paper we describe a scheme for integrating such kind of radiation source at the european xfel facility @xcite . we begin our considerations with the generation of thz radiation from the spent electron beam downstream of the sase2 undulator in the electron beam - dump area . we then move to consider the transport of thz radiation pulses from the xfel beam dump - area to the experimental hall . this constitutes a challenge , because the thz output must propagate at least @xmath0 meters in the photon beam tunnel and in the experimental hall to reach the sase2 x - ray hutches . since thz beams are prone to significant diffraction , a suitable beam transport system must be provided to guide the beam along large distances maintaining it , at the same time , within a reasonable size . moreover , the thz beamline should be designed to obtain a large transmission efficiency for the radiation over a wide wavelength range . transmission of the thz beam can only be accomplished with quasi - optical techniques . in this paper , similarly as in @xcite , which focused on the lcls baseline , we propose to use an open beam waveguide such as an iris guide , that is made of periodically spaced metallic screens with holes , for transporting the thz beam at the european xfel . the eigenmodes of the iris guide have been calculated numerically for the first time by fox and li @xcite and later obtained analytically by vainstein @xcite . in @xcite we already presented a complete iris guide theory . in particular , the requirements on the accuracy of the iris alignment were studied . in order to efficiently couple radiation into the transmission line , it is desirable to match the spatial pattern of the source radiation to the mode of the transmission line . to this end , it is advisable to generate radiation from the spent electron beam directly in the iris line with the same parameters used in transmission line . in this way , the source generates thz radiation pulses with a transverse mode that automatically matches the mode of the transmission line . the theory described in @xcite supports this choice of thz source . in the present work we present a conceptual design for a thz edge radiation source at the european xfel . it includes an @xmath10 m - long electron beam vacuum chamber equipped with an iris line , and a @xmath0 m - long transmission line with the same parameters . the transmission line , which develops through the xtd7 distribution tunnel and field @xmath11 of the experimental hall , includes at least ten @xmath12-degrees turns with plane mirrors at @xmath13 degrees as functional components . it is possible to match incident and outgoing radiation without extra losses in these irregularities . the transport line introduces a path delay of about @xmath1 m between thz and x - ray pulses generated from the same electron bunch . since thz pump / x - ray probe experiments should be enabled , in order to cope with this delay we propose to exploit the unique bunch structure foreseen as baseline mode of operation at the european xfel , with @xmath2 ns electron bunch separation , together with an additional delay line in experimental hall fs ) rather than by the jitter between the electron bunches ( about @xmath14 fs ) . ] . the availability of any thz source at the european xfel facility should be complemented with the availability of a suitable thz beam transport system , which must guide the beam for distances in the @xmath15 meters range . the thz beam can only be transmitted with quasi - optical techniques . in particular , the idea of providing a periodic phase correction for the free- space beam , in order to compensate for its divergence , is very natural . in the 1960s numerous attempts of designing various quasi - optical transmission lines were reported . in particular , it was proposed to use open beam waveguides such as lens guides , mirror guides , and iris guides @xcite,@xcite-@xcite . the competition among different proposals ended with the victory of mirror guides , still in use today for example for plasma heating @xcite . focusing of the thz beam can only be provided with reflective optics since lenses made of any material would reflect and absorb all radiation at long distances . [ t1 ] ( a ) shows the optical arrangement of the transport system based on the use of a mirror guide . each focusing unit is composed by two matched copper mirrors . the mirrors are separated in such a way that the incident angle is sufficiently small to minimize astigmatism . existing mirror guides are characterized by a maximal length of about 40 - 60 m @xcite . as discussed above , at european xfel facility scientists need a beam transport system working for significantly longer distances , and the advantages of a mirror - guide based transport system are not evident . in this paper we propose , as an alternative to the mirror line solution , to use an iris line , fig . [ t1 ] ( b ) . iris lines are characterized by a low attenuation of the fundamental mode , self - filtering of high order modes , wide operational wavelength range and mechanical integrity of the structure . the first investigation of the influence of diffraction effects on the formation of the field eigenmodes in an iris line was carried out by fox and li using physical optics techniques @xcite . a very different and mathematically solid approach to the same problem was introduced by vainstein @xcite . his studies are based on direct solution of maxwell equations . an analysis of diffraction and reflection from the iris edges allowed vainstein to derive for the first time analytic expressions for field distribution and mode losses in iris guides . an infinite , discrete set of eigenfunctions with corresponding complex eigenvalues can be found . this set of modes comprises leaky modes which are similar albeit not identical to modes of microwave waveguide with resistive walls . for instance , one can find that iris line modes , in contrast with microwave waveguide leaky modes , are independent of the polarization of the radiation . consider an axially symmetric iris guide with inner radius @xmath16 and distance between the two screens @xmath17 , and call @xmath18 the temporal fourier transform of the transverse electric field of the radiation in the guide . since the transverse size of the radiation is much larger than the reduced wavelength @xmath19 , the field envelope @xmath20 $ ] turns out to be a slowly varying function of the longitudinal coordinate @xmath21 with respect to the wavelength , and it obeys the paraxial maxwell equation . given the symmetry of the problem we can introduce polar coordinates @xmath22 , and seek a solution for the slowly varying field amplitude @xmath23 of given polarization component in the form @xmath24~ , \label{solu}\end{aligned}\ ] ] with n = 0 , 1 , 2 , ... . in the first order of the small parameter @xmath25 , where @xmath26 is the fresnel number , the functions @xmath27 assume the form @xmath28 where @xmath29~ , \label{knj}\end{aligned}\ ] ] with turns out to be related with one of the most famous mathematical functions , the riemann zeta function @xmath30 . in fact , it is given by @xmath31 , see @xcite . ] @xmath32 , and @xmath33 the @xmath34-th root of the @xmath35-th order bessel function of the first kind ( i.e. @xmath36 ) . substituting the expression for @xmath37 into the dispersion relation @xmath38 we obtain the following expression for @xmath39 : @xmath40 for the eigenmode with transverse wavenumber @xmath37 , the fraction of the radiation power losses per transit of one iris is given by @xmath41 the relative loss of the @xmath34-th mode of order @xmath35 after traveling for a distance @xmath21 is therefore given by @xmath42 due to the exponential dependence on @xmath43 , only the lower order modes tend to survive . note that the exponent in eq . ( [ loss ] ) depends on the distance between two irises , @xmath17 , only weakly as @xmath44 , while there is a much stronger dependence on @xmath45 and @xmath16 . as already mentioned , an iris line constitutes a suitable thz beam transport system for the european xfel . for example , calculations indicate that the losses of the principal mode at @xmath46 m , @xmath47 cm and @xmath48 cm should be in the order of @xmath49 after traveling for a distance @xmath50 m. the transfer line pipe diameter is estimated to be @xmath1 cm . it is therefore technically feasible to install such transfer line inside the distribution tunnel xtd7 . however , as discussed before , in order to efficiently couple radiation into the transmission line , one needs to match the spatial pattern of the source radiation to the mode of the transmission line . let us consider a coherent edge radiation source in the thz range , which is relatively simple to implement at the european xfel . a setup where edge radiation is formed with the help of upstream and downstream metallic screens is shown in fig . the edge radiation from the upstream screen is extracted by the downstream screen , which acts as a mirror , and is sent to the iris transmission line . the length of straight section between upstream screen and mirror plays the role of the length of the insertion device for edge radiation ( see for example @xcite for a review on these matters ) . the matching problem is easily solved if the thz source is equipped with an iris line as well . when dealing with a setup where thz radiation from an ultra relativistic electron beam is extracted by a mirror , one usually talks about backward transition radiation ( btr ) . the main problem to solve is the specification of the electric field distribution at some position where the mirror is present . it should be stressed that the specification of the field at the mirror position must be considered as the first step to the specification of the field at the sample position . such first step is considered separately , because the field at the mirror position is independent of the type of mirror and outcoupling optics . once the field at the mirror position is known , the problem of specification of the field at the sample position can be solved for example with the help of physical optics techniques . let us discuss the problem of field characterization at the mirror position in more detail . when electrons are in unbounded space and come from an infinitely long straight line , the field distribution in the mirror plane can be calculated analytically following ginzburg and frank @xcite . in most practical cases , however , the ginzburg - frank equation is not applicable because two basic assumptions of the analytical derivation are not fulfilled : the electron beam is moving inside the metallic vacuum chamber , which effectively acts like an overmoded waveguide , and the straight line has a finite length with a bending magnet at the upstream end . it is known that ginzburg - frank theory is the limiting case of the more general theory of edge radiation in unbounded space . for a review on these matters see , among others , @xcite . for a very recent review on edge radiation and further references see , for example , @xcite . emission of edge radiation in the presence of metallic boundaries has been a much less - treated subject in literature , compared to the unbounded space case . to the best of our knowledge , there is only one article reporting on edge radiation from electrons in a homogeneous metallic overmode waveguide , in particular with circular cross - section @xcite . the method described in @xcite is therefore capable of treating realistic experimental setups . in @xcite we applied the method described in @xcite to the case of an iris guide . to fix ideas we focus our attention on the setup in fig . electrons travel through the usual edge radiation setup . the difference is that now we account for the presence of the iris guide along the straight section . since electrons pass through an upstream edge screen , one may assume that the iris guide starts at the upstream screen position . one should calculate the field distribution at the mirror position , which should be subsequently propagated to the experimental hall . the presence of an upstream edge screen seems at first glance not necessary , because electrons come in any case from the straight section . however , due to variations of the vacuum chamber cross - section upstream of the edge radiation setup , characterization of the field distribution at the upstream open - end of the iris guide is problematic . in spite of this , the electric field of the electron beam in the plane immediately behind the upstream screen can be well defined as zero , leading to extra simplifications . in this case we deal with a well - defined problem , and this allow us to characterize the field distribution at the end of the iris guide in the mirror plane . in the case of the thz source at the european xfel , the electron beam transverse size is much smaller than the diffraction size . this means that , as pertains the characterization of the thz pulses , the electron beam can be modeled as a filament beam . in this case the electron beam current is made up of moving electrons randomly arriving at the entrance of the iris guide , and obeying @xmath51 where @xmath52 is the dirac delta function , @xmath53 is the electron charge , @xmath54 is number of electrons in a bunch , and @xmath55 is the random arrival time of the @xmath56-th electron at the undulator entrance . the electron bunch profile is described by the profile function @xmath57 . @xmath58 represents the probability of finding an electron between time @xmath59 and time @xmath60 . the beam current averaged over an ensemble of bunches can then be written in the form : @xmath61 the radiation power at frequency @xmath62 , averaged over an ensemble , is given by the expression : @xmath63~ , \label{pow}\end{aligned}\ ] ] where @xmath64 is the radiation power from one electron and @xmath65 is the fourier transform of the bunch profile function . for wavelengths shorter than the bunch length the form factor reduces to zero . for wavelengths longer than the bunch length it approaches unity . , corresponding to the current profile on the left plot.,title="fig:",scaledwidth=50.0% ] , corresponding to the current profile on the left plot.,title="fig:",scaledwidth=50.0% ] let us consider the practically important case of an electron bunch with non - gaussian shape similar to that , which can be used to drive the european xfel . [ curform ] ( left ) shows the current distribution along the bunch . the nominal charge is @xmath3 nc . the electron bunch has a complicated shape , which is reflected in the square modulus of the form factor shown in fig . [ curform ] ( right ) . in the case of the european xfel , the squared of the bunch form factor modulus falls off rapidly for wavelengths shorter than @xmath66 mm . at the opposite extreme , the dependence of the form factor on the exact shape of the electron bunch is rather weak and can be ignored for wavelengths longer than @xmath67 mm . we assume an electron beam moving along z - axis inside the axisymmetric iris line through the screens , fig . [ t2 ] . as already remarked , in reference @xcite we developed a theory of such kind of thz source using veinstein impedance boundary conditions and green s function approach to solve the field equations . we obtained the following expressions for the field at the mirror position : @xmath68}\exp\left(-\frac{i c \nu_{1k}^2(1 - 2\delta)}{2\omega a^2}\frac{l}{2}\right)\label{etildex}\end{aligned}\ ] ] for the horizontal component and @xmath69}\exp\left(-\frac{i c \nu_{1k}^2(1 - 2\delta)}{2\omega a^2}\frac{l}{2}\right)\label{etildey}\end{aligned}\ ] ] for the vertical one . here @xmath70 , @xmath71 is the relativistic factor , @xmath72 is the distance between upstream edge screen and mirror . coherence is accounted through the form factor @xmath65 . only non - azimuthal symmetric modes turn out to be driven by the uniform motion of the space - charge distribution in the axisymmetric iris guide . from a physical view point this is a sound result . in fact , radiation is related with the energy change of particles , which takes place through the scalar product of the electric field and velocity of the particles . since the transverse velocity is equal to zero in the edge radiation case , symmetric modes can not lead to any energy change of the electrons moving along the axis of the axisymmetric iris guide . let us focus on the fundamental non symmetric mode only . neglecting losses , for the moment , the amplitude for the orthogonal polarization components of the field in a cartesian coordinate system is given by @xmath73 where @xmath74 is a constant for a certain longitudinal position , @xmath75 , and @xmath76 . it follows from these equations that the direction of the electric field is radial i.e. varies as a function of the transverse position . therefore , for the edge radiation case we have two separate amplitudes for two orthogonal polarization directions . they are not azimuthal symmetric because they depend , respectively , on @xmath77 and @xmath78 , that is on the cosine and on the sine of the azimuthal angle . only if one sums up the intensity patterns referring to the two polarization components one obtains the azimuthal symmetric intensity distribution @xmath79 edge radiation is characterized by broad spectrum and radial polarization . many practical applications require control of the bandwidth and of the polarization of the thz pulse . the formation of the thz edge radiation pulse usually involves monochromatization and polarization filtering by wire - grid polarizer for producing linearly polarized thz radiation , fig . [ t3 ] . ( [ etildex ] ) and eq . ( [ etildey ] ) can be used as a basis to calculate the energy per unit spectral interval per unit surface @xmath80 in our case of interest , the distance between sample and extracting mirror , @xmath81 , is much longer compared with the length of the edge radiation setup . integrating over transverse coordinates over the iris hole , and assuming @xmath82 the energy of the radiation pulse at the sample position is given by @xcite @xmath83 \frac{\delta \omega}{\omega}~.\label{energy}\end{aligned}\ ] ] here we assumed that the bandwidth of spectral filter @xmath84 is small enough to neglect the dependence on @xmath62 in eq . ( [ etildex ] ) and eq . ( [ etildey ] ) . the thz edge radiation source proposed in this paper is compatible with the layout of the european xfel and can be realized with little efforts and minimal modifications to the baseline setup . the vacuum chamber equipped with iris line and outcoupling system can be installed in the unoccupied straight vacuum line upstream of the first electron beam dump , fig . [ t9 ] . the transport of the thz beam to the experimental hall can be performed exploiting the use of the tunnel xtd7 , as reproduced schematically in fig . [ t9 ] . the thz transmission line transporting the thz radiation introduces a path - delay of about 20 m with respect to the path of x - ray pulse from the sase2 undulator . since thz pump / x - ray probe experiments should be enabled , we propose to exploit the natural bunch spacing within a train , fig . the thz and x - ray pulses should be synchronized by a thz delay line installed in experimental hall , fig . [ t11 ] . the thz transmission line includes at least ten @xmath12-degrees turns , and will exploit plane mirrors as functional components . if the pipe of the transmission line has a diameter smaller than the distance between the irises it is possible to match incident and outgoing radiation without extra losses in these irregularities , fig . [ t7 ] . meters away from the extracting mirror . the curves are calculated with analytical formulas in eq . ( [ etildex ] ) , eq . ( [ etildey ] ) and eq . ( [ eneee ] ) , including losses as in eq . ( [ loss ] ) . here @xmath85 ( @xmath3 nc ) , @xmath86 m , @xmath47 cm , @xmath48 cm . the bunch form factor used is shown on the right plot of fig . [ curform].,scaledwidth=100.0% ] it is possible to calculate the energy spectral density at the sample position . the bunch form factor considered here is given in fig . [ curform ] . we set a total length of the transport iris line of @xmath0 m. the energy losses in the line are accounted for . the energy spectral density at the exit of transport iris line as a function of the frequency @xmath87 is shown in fig . [ enepulse ] . the curve is calculated with analytical formulas in eq . ( [ etildex ] ) , eq . ( [ etildey ] ) and eq . ( [ eneee ] ) including losses as in eq . ( [ loss ] ) . the first four modes and the total are shown . here @xmath88 , corresponding to 1 nc charge , @xmath47 cm , @xmath89 cm , @xmath86 m , and @xmath90 m. the maximum value of the energy spectral density is achieved at @xmath91 m . when the bandpass filter is tuned to this value of @xmath45 , the expression for the total edge radiation pulse energy at the sample can be written in the form @xmath92 \simeq 0.35 \frac{\delta \omega}{\omega}~. \label{ww}\end{aligned}\ ] ] the energy spectral density as a function of frequency exhibits a low frequency cutoff due to losses in the transport line and a high frequency cutoff due to electron bunch form factor suppression . of the edge radiation pulse energy spectral density transported at the sample position ( @xmath93 m downstream of the mirror ) as a function of frequency . here @xmath88 ( @xmath3 nc ) , @xmath47 cm , @xmath48 cm . the bunch form factor used is shown on the right plot of fig . [ curform ] . the curves indicated with numbers from @xmath3 to @xmath1 refer to different values of @xmath72 starting from @xmath94 m and ending with @xmath95 m , with a step of @xmath96 m.,scaledwidth=100.0% ] of the edge radiation pulse energy spectral density transported at the sample position ( @xmath93 m downstream of the mirror ) as a function of frequency . here @xmath88 ( @xmath3 nc ) , @xmath47 cm , @xmath86 m. the bunch form factor used is shown on the right plot of fig . [ curform ] . the curves indicated with numbers from @xmath3 to @xmath1 refer to different values of @xmath16 starting from @xmath97 cm and ending with @xmath98 cm , with a step of @xmath3 cm.,scaledwidth=100.0% ] we studied the influence of the edge radiation setup length @xmath72 and of the iris hole radius @xmath16 on the operation of the thz source . [ scanl ] and fig . [ scana ] show the dependence of the thz pulse energy spectral density on the frequency for different values of these two parameters . the nontrivial behavior of the energy spectral density reflects the complicated parametric dependence in eq . ( [ energy ] ) . finally , we note that according to our calculations in fig . [ enepulse ] , one obtains about @xmath99 mj total thz radiation pulse energy at the sample position at a wavelength around @xmath100 mm with a filter bandwidth @xmath101 . for one cycle in a pulse , corresponding to @xmath102 spectral bandwidth ) shows that the phase velocity of the radiation field of the iris guide mode is larger than the velocity of light and depends on the frequency as @xmath103 for large values of the fresnel number @xmath104 . in other words , the thz radiation pulse propagates through the transport line with a certain group velocity dispersion , and the shape of the envelope broadens . in our case of interest , the broadening is about @xmath105 , and can be neglected . ] , one obtains about 0.5 gw peak power level . by highly focusing this thz beam , one will approach the high field limit of 1 v / atomic size . the electron beam formation system of the european xfel accelerator complex operates with nominal charge of @xmath3 nc . the initially 2 mm ( rms ) long bunch is compressed in three magnetic chicanes by a factor of @xmath106 , @xmath107 and @xmath106 respectively , to achieve a peak current of @xmath11 ka . the design of our thz source is also based on the exploitation of electron bunches with nominal charge of @xmath3 nc . however , in order to optimize the thz source performance , the electron bunch should be made about two times shorter than in the nominal mode of operation . in this case , the peak current increases by a factor two . operation beyond the design parameters leads to some degradation of the electron beam quality . the operation of the proposed thz source is insensitive to the emittance and energy spread of the electron bunch . however , one should examine the performance of the electron bunches in terms of x - ray sase pulses generation , since thz pump / x - ray probe experiments are based on such pulses . first , we ran start - to - end simulations for the beam formation system @xcite , yielding the beam characteristics summarized in fig . [ sases2e ] in terms of current , emittance , energy and energy spread distribution along the electron bunch at the entrance of the sase2 undulator . subsequently , we studied the performance of the electron bunch with the help of the fel code genesis 1.3 @xcite running on a parallel machine . here we will present results based on a statistical analysis consisting of @xmath108 runs . fig . [ sigperp ] shows the evolution of the rms horizontal beam size as a function of the distance inside the sase2 undulator . the figure shows the evolution for the position of maximal current in the bunch . fel undulator cells . right plot : sase spectrum at the exit of the first @xmath96 fel undulator cells . grey lines refer to single shot realizations , the black line refers to the average over a hundred realizations.,title="fig:",scaledwidth=50.0% ] fel undulator cells . right plot : sase spectrum at the exit of the first @xmath96 fel undulator cells . grey lines refer to single shot realizations , the black line refers to the average over a hundred realizations.,title="fig:",scaledwidth=50.0% ] the sase2 undulator is composed of @xmath109 cells . the sase output after the first @xmath96 undulator cells is shown in fig . [ saseout ] . on the left plot we show the output power , while on the right plot one can see the spectrum around the the fundamental at @xmath110 angstrom . grey lines refer to single shot realizations , the black line refers to the average over a hundred realizations . the output characteristics are in line with the expected baseline performance , demonstrating an average power of several tens of gw , and a relative bandwidth in the @xmath111 range . the left plot in fig . [ saseevo ] shows the evolution of the pulse energy as a function of the position inside the sase2 undulator , while the right plot shows the variance of the energy fluctuations . as usual , fluctuations increase in the linear regime , and drop at saturation . finally , in fig . [ sasespot ] we present the transverse profile of the radiation distribution , obtained by averaging over the statistical ensemble and the directivity diagram of the radiation , obtained by fourier transformation . [ sasespot ] demonstrates a spot size of several tens of microns , and a divergence of a few microradians . simulations presented in this section show that the operation of sase2 with a @xmath3 nc bunch and a peak - current increase of a factor two compared to the nominal mode of operation ( also with a @xmath3 nc bunch but with design parameters ) , leads practically to the same peak power of the x - ray pulse that is in the order of @xmath1 gw . in other words , the two times shorter bunch duration in our case study has the simple consequence of a decrease in the x - ray pulse energy of about a factor two , from around @xmath112 mj in the design case to around @xmath3 mj in our case . the accelerator complex at the european xfel will produce ultrashort electron bunches approaching sub - hundred fs duration . it is natural to take advantage of these ultra - short bunches to provide coherent thz radiation . thz radiation pulses can be generated by the spent electron beam downstream of x - ray undulators . any method relying on the spent electron beam at the european xfel should also provide , at the same time , a way of transporting the radiation up to experimental hall for a distance in the @xmath15-meters range . transmission of the thz beam can only be accomplished with quasi- optical techniques . for the european xfel case we propose to use an open beam waveguide such as an iris guide . in order to efficiently couple radiation into the iris transmission line , it is desirable to match the spatial pattern of the source radiation to the propagating mode of the transmission line . to solve the matching problem , we propose to generate the thz radiation directly within the iris guide . the resulting electric field is found as superposition of the iris waveguide modes , and is studied for the parameters case of the european xfel . we demonstrated that the maximally achievable field strength at the sample is in the high field region of 1 v / atomic size . the present design describes a basic option for a thz source at the european xfel facility . a technical solution allowing for significant performance enhancement of the output , of about an order of magnitude , without significant changes of the baseline hardware , seems also feasible . such solution consists in the extension of the proposed method to a higher charge mode of operation ( up to @xmath113 nc ) , which is within the possibilities of the european xfel accelerator complex . it is hoped that future research may extend the thz source performance to a few gw power level . we are grateful to massimo altarelli , reinhard brinkmann , serguei molodtsov , thomas tschentscher and edgar weckert for their support and their interest during the compilation of this work . the lcls - ii conceptual design report , https:@xmath114slacportal.slac.stanford.edu@xmath115sites@xmath115 lcls@xmath116public@xmath115lcls@xmath116ii@xmath115published@xmath116documents@[email protected] g. geloni , v. kocharyan and e. saldin , `` scheme for generating and transporting thz radiation to the x - ray experimental floor at the lcls baseline '' , desy 11 - 134 ( 2011 ) , http://arxiv.org/abs/1108.1085 .	the design of a thz edge radiation source for the european xfel is presented . we consider generation of thz radiation from the spent electron beam downstream of the sase2 undulator in the electron beam dump area . in this way , the thz output must propagate at least for @xmath0 meters through the photon beam tunnel to the experimental hall to reach the sase2 x - ray hutches . we propose to use an open beam waveguide such as an iris guide as transmission line . in order to efficiently couple radiation into the iris transmission line , generation of the thz radiation pulse can be performed directly within the iris guide . the line transporting the thz radiation to the sase2 x - ray hutches introduces a path delay of about @xmath1 m. since thz pump / x - ray probe experiments should be enabled , we propose to exploit the european xfel baseline multi - bunch mode of operation , with @xmath2 ns electron bunch separation , in order to cope with the delay between thz and x - ray pulses . we present start - to - end simulations for @xmath3 nc bunch operation - parameters , optimized for thz pump / x - ray probe experiments . detailed characterization of the thz and sase x - ray radiation pulses is performed . highly focused thz beams will approach the high field limit of 1 v / atomic size . * deutsches elektronen - synchrotron * * + * desy 11 - 244 december 2011 @xmath4 @xmath5 * scheme for generating and transporting thz radiation to the x - ray experimental hall at the european xfel * @xmath6 winfried decking@xmath7 , gianluca geloni@xmath8 , vitali kocharyan@xmath7 , evgeni saldin@xmath7 and igor zagorodnov@xmath7 _ + @xmath7deutsches elektronen - synchrotron desy , hamburg _ _ + @xmath8european xfel gmbh , hamburg _ @xmath5 issn 0418 - 9833 @xmath5 * notkestrasse 85 - 22607 hamburg *
the size of conjugacy classes ( in the topological sense : dense , meager , comeager , etc . ) in the groups of automorphisms of the frass limits has recently become an active area of research . this is partially due to the newly revealed connections between combinatorial properties of the frass classes and algebraic , topological , and dynamical properties of the groups of automorphisms of their limits . one of the most astonishing links was established by kechris , pestov and todorcevic in @xcite and displays close relationship between the ramsey theory ( a purely combinatorial area ) and the extreme amenability ( a classical dynamical notion ) . another reason for the interest in the size of conjugacy classes of polish groups in general , and groups of automorphisms of frass limits in particular , comes from the special importance of some concrete groups , e.g. , the group of automorphisms of the countable atomless boolean algebra ( which is isomorphic via the stone s theorem to the group of homeomorphisms of the cantor space ) , the group of isometries of the rational urysohn space , and the group of order preserving bijections of the rationals . the reader may consult @xcite for details , examples , and a deep structural theory for the groups with large conjugacy classes . in @xcite j. k. truss looked at different possible notions of genericity of conjugacy classes and discussed advantages of each . in general , conjugacy classes are objects that are difficult to understand , and the relation of conjugation may sometimes be very complicated ( complete analytic ) . motivated by the work of truss , in this paper we look at a coarser equivalence relation than conjugation , namely at classes of topological similarity . they are much easier to work with and can be used to prove the _ non_-genericity in some cases . let us recall the definition from @xcite . let @xmath0 be a topological group , an @xmath1-tuple @xmath2 is said to be _ topologically similar _ to an @xmath1-tuple @xmath3 if the map @xmath4 sending @xmath5 extends ( necessarily uniquely ) to a bi - continuous isomorphism between the groups generated by these tuples @xmath6we denote this relation by @xmath7 . there is another natural relation on the @xmath1-tuples in @xmath0 , namely the relation of diagonal conjugation , i.e , @xmath8 is conjugate to @xmath9 if there is some @xmath10 such that @xmath11 more generally , if @xmath0 is a topological subgroup of @xmath12 , then one can restrict the conjugacy relation in @xmath12 to @xmath0 , i.e , say that an @xmath1-tuple @xmath13 conjugates in @xmath12 to an @xmath1-tuple @xmath3 if there is some @xmath14 such that @xmath15 it is easy to see that this is an equivalence relation on @xmath1-tuples of @xmath0 and we denote it by @xmath16 ( this relation also depends on @xmath0 and on the embedding of @xmath0 into @xmath12 , but this information is usually clear from the context ) . the following proposition is obvious . let @xmath0 be a topological group and @xmath17 , then a. relation of topological similarity is an equivalence relation ; b. let @xmath12 be any topological group such that @xmath18 is a topological subgroup of it , then @xmath16 is finer ( not necessarily strictly ) than @xmath7 . in particular , @xmath19 is finer than @xmath7 . equivalence classes of topological similarity on @xmath1-tuples are called _ @xmath1-dimensional similarity classes _ , in particular two - dimensional similarity class is a set of pairs . since conjugacy classes refine classes of topological similarity , if one wants to prove meagerness of the former , it suffices to prove meagerness of the latter ( it suffices , but may be impossible , meagerness of conjugacy classes does not implies meagerness of classes of topological similarity ) . this sometimes turns out to be an easier task . for example , rosendal in @xcite developed this idea to find a simple proof of the del junco s result , that each conjugacy class of measure preserving automorphisms of the standard lebesgue space is meager . hodkinson showed ( see @xcite for the details ) that in the group of order preserving automorphisms of the rationals all conjugacy classes of pairs are meager ( though it is known that there is a comeager one - dimensional conjugacy class in @xmath20 ) . in this paper we strengthen this result and show that all two - dimensional classes of topological similarity are meager in this group . the paper is organized as follows . in the second section we give a brief introduction to the theory of frass classes , which is the right context for the technique , developed in this paper . the third section is devoted to the strengthening of the truss s result on classes of diagonal conjugation in the rationals , in the fourth and fifth sections analogous theorem is proved for the group of order preserving isometries of the randomly ordered rational urysohn space . * acknowledgment . * the author wants to thank christian rosendal for advising him during the process of writing this paper , for posing questions , that are covered here , and for helpful and very inspiring communication and comments . the author also thanks sawomir solecki and ward henson for helpful discussions . in this section we give a short introduction to the theory of frass classes . in the next two sections we deal with two examples of them , so it is useful to keep in mind this more general setting . the classical text on frass classes is a beautiful book by hodges @xcite . let @xmath21 be a relational first order language . we use the standard notation : solid arrows correspond to `` for all '' quantifiers , and dashed arrows represent maps that `` exist '' . let @xmath22 and @xmath23 be two @xmath21-structures . a map @xmath24 is called a _ strong homomorphism _ if for any @xmath25 of arity @xmath26 and for any @xmath27 @xmath28 the map is a _ strong embedding _ if it is an injective strong homomorphism . let @xmath29 be a class of finite @xmath21-structures . for @xmath21-structures @xmath22 and @xmath23 by @xmath30 we mean `` @xmath22 strongly embeds into @xmath23 '' . @xmath29 is called a frass class if the following properties hold : a. if @xmath30 and @xmath31 then @xmath32 ; b. for @xmath32 and @xmath31 there is some @xmath33 such that @xmath34 and @xmath35 ; c. for @xmath32 , @xmath31 , @xmath33 and embeddings @xmath36 , @xmath37 there is @xmath38 and embeddings @xmath39 , @xmath40 such that @xmath41 , i.e. , the following diagram commutes + @xmath42^{k } & \\ { \mathbf{a}}\ar[ur]^{i } \ar[dr]^{j } & & { \mathbf{d}}\\ & { \mathbf{c}}\ar@{-->}[ur]^{l } & } \ ] ] d. @xmath29 contains structures of arbitrarily high finite cardinality and has up to isomorphism only countably many structures . basic examples of frass classes are : finite sets , finite linear orders , finite graphs and finite metric spaces with rational distances ( to satisfy ( inf ) condition ) . with a frass class @xmath29 one can associate its frass limit ( which is unique up to an isomorphism ) . countably infinite structure @xmath43 is called a _ frass limit _ of the class @xmath29 if the following holds : a. finite substructures of @xmath43 up to isomorphism are exactly the elements of @xmath29 ; b. @xmath43 is ultrahomogeneous ( that is any isomorphism between finite substructures of @xmath43 extends to a full automorphism of @xmath43 ) . frass limits of the above frass classes are : @xmath44 countably infinite set , @xmath45 dense linear ordering without endpoints , @xmath46 random graph , @xmath47 rational urysohn space . in the next section we deal with the simplest linearly ordered frass class : with the rationals @xmath45 . the fourth and fifth sections are devoted to the case of the linearly ordered rational urysohn space . let @xmath45 denote rational numbers viewed as a linearly ordered set . by an open interval @xmath49 we mean the set of rational numbers @xmath50 . a closed interval @xmath51 $ ] also includes endpoints @xmath52 and @xmath53 . if @xmath54 is a bounded interval ( open or closed ) @xmath55 will denote its left endpoint and @xmath56 will be its right endpoint . if @xmath57 is a finite subset , @xmath58 and @xmath59 will denote its minimal and maximal elements respectively . let @xmath0 denote the group @xmath20 of order preserving bijections of the rationals . a partial isomorphism of @xmath45 is an order preserving bijection @xmath60 between _ finite _ subsets @xmath22 and @xmath23 of @xmath45 . it is a basic property of rationals ( and , as mentioned earlier , of a frass limit in general ) that each partial isomorphism can be extended ( certainly , not uniquely ) to a full automorphism . letters @xmath60 and @xmath61 ( with possible sub- or superscripts ) will denote partial isomorphisms ; let @xmath62 be the domain of @xmath60 , and @xmath63 be its range . if @xmath64 then @xmath65 denotes restriction of @xmath60 on @xmath66 ; @xmath67 will be the set of fixed points in the domain of @xmath60 , i.e , @xmath68 first we recall that @xmath0 is a polish group ( i.e. , a separable completely metrizable topological group ) in the topology given by the basic open sets @xmath69 where @xmath60 is a partial isomorphism of @xmath45 . note that if @xmath60 and @xmath61 are two partial isomorphisms and @xmath61 extends @xmath60 then @xmath70 ; we will use this observation frequently . we denote the identity element of @xmath0 by @xmath71 . we use words generic and comeager as synonyms . for example , a property is generic in the group @xmath0 if the set of elements with this property is comeager in @xmath0 . let @xmath72 denote the free group on two generators : @xmath73 and @xmath74 ; elements of @xmath72 are reduced words on the alphabet @xmath75 . every element @xmath76 has certain length associated to it , namely the length of the reduced word @xmath77 . this length is denoted by @xmath78 . if @xmath79 are words , we say that the word @xmath80 is reduced if @xmath81 , that is there is no cancellation between @xmath82 and @xmath83 . if @xmath84 is a reduced word , @xmath85 , and @xmath60 , @xmath61 are partial isomorphisms , then we can define a partial isomorphisms @xmath86 by @xmath87 , whenever the right - hand side is defined . the _ orbit _ of @xmath88 under @xmath86 is by definition @xmath89 we say that a word @xmath77 starts from the word @xmath83 if @xmath77 can be written as @xmath90 for some word @xmath82 , where @xmath91 is reduced . similarly , we say that @xmath77 ends in @xmath83 if there is a word @xmath82 such that @xmath92 , where @xmath93 is reduced . on the one hand this is consistent with intuitive understanding of this notions for , say , left - to - right languages . on the other hand , we consider left actions , and then _ the end of the word acts first _ , i.e. , if @xmath94 then @xmath95 . this may be a bit confusing , we apologize for that and emphasize this possible confusion . let @xmath60 be a partial isomorphism of @xmath45 . an interval @xmath96 is called @xmath60-_increasing _ if @xmath97 , @xmath98 , @xmath99 and @xmath100 for any @xmath101 . definition of @xmath60-_decreasing _ interval is analogous . note that if @xmath51 \cap { \mathop{\mathrm{dom}}(p ) } = \{a , b\}$ ] and @xmath98 , @xmath99 then the interval @xmath102 is both @xmath60-increasing and @xmath60-decreasing . an interval is @xmath60-_monotone _ if it is either @xmath60-increasing or @xmath60-decreasing . let @xmath60 be a partial isomorphism . let @xmath103 and assume that @xmath104 . we say that @xmath60 is _ informative _ if @xmath105 , @xmath106 and there is a list @xmath107 of indices such that a. @xmath108 , @xmath109 ; b. @xmath110 for @xmath111 ; c. for any @xmath112 the interval @xmath113 is @xmath60-monotone . if @xmath60 is an informative partial isomorphism and @xmath114 as above then we set @xmath115 and refer to it as to _ the set of essential points of @xmath60_. a pair @xmath116 of partial isomorphisms is called _ piecewise elementary _ if the following holds a. @xmath60 and @xmath61 are informative ; b. @xmath117 , c. @xmath118 . if additionally @xmath119 has cardinality at most @xmath120 ( i.e. , consists of the above minimum and maximum ) then the pair @xmath121 is called _ let @xmath116 be a piecewise elementary pair , and @xmath122 with @xmath123 for @xmath124 . set @xmath125 $ ] , then @xmath126 is elementary for any @xmath127 . thus every piecewise elementary pair @xmath121 can be decomposed into finitely many elementary pairs . the following obvious lemma partially explains the importance of piecewise elementary pairs . [ pelem - dense ] for any open @xmath128 there is a piecewise elementary pair @xmath121 such that @xmath129 . [ def : elem ] let @xmath121 be an elementary pair . we say that a triple @xmath130 _ liberates @xmath60 in @xmath121 _ , where @xmath131 and @xmath132 are partial isomorphisms that extend @xmath60 and @xmath61 respectively , and @xmath133 is a reduced word , if the following holds a. [ def : elem : inf ] @xmath131 and @xmath132 are informative ; b. [ def : elem : presmin ] @xmath134 , @xmath135 ; c. [ def : elem : presmax ] @xmath136 , @xmath136 ; d. [ def : elem : wform ] the word @xmath77 starts from a non - zero power of @xmath74 , @xmath137 for @xmath138 ; e. @xmath139 is defined for any @xmath140 and @xmath141 f. [ def : elem : monot ] there is an open interval @xmath142 such that @xmath143 , @xmath132 is monotone on @xmath142 and @xmath144 for any @xmath145 ; moreover , if @xmath146 in the item , then @xmath142 is @xmath132-increasing , and it is @xmath132-decreasing otherwise . similarly , we say that a triple @xmath147 _ liberates @xmath61 in @xmath121 _ if the above holds with roles of @xmath60 and @xmath61 , @xmath73 and @xmath74 interchanged . for a piecewise elementary pair @xmath121 , we say that a triple @xmath147 liberates @xmath60 [ liberates @xmath61 ] in @xmath121 if a. @xmath134 , @xmath149 ; b. @xmath136 , @xmath150 ; c. for any interval @xmath54 , such that @xmath151 is elementary , the triple @xmath152 liberates @xmath153 [ liberates @xmath154 in @xmath151 . [ liberating - elem - weak ] for any elementary pair @xmath121 there is a triple @xmath147 that liberates @xmath60 [ liberates @xmath61 ] in @xmath121 . we show the existence of a triple that liberates @xmath60 , the second clause then follows by symmetry . extending @xmath60 and @xmath61 if necessary , we may assume that a. @xmath155 , @xmath156 ; b. @xmath157 are all the ( open ) intervals of monotonicity for @xmath60 and @xmath158 are all the ( open ) intervals of monotonicity for @xmath61 ; we list them in the increasing order , i.e. , @xmath159 , @xmath160 ; c. [ smp : nonemp ] @xmath161 and @xmath162 ; d. @xmath163 and @xmath164 ; e. @xmath165 . . then @xmath167 by ( and in particular @xmath168 is not a fixed point of @xmath60 or @xmath61 ) . we first find an extension @xmath169 of @xmath60 and @xmath170 such that @xmath171 is defined and is `` close enough '' to the right endpoint of @xmath172 . `` close enough '' exactly means the following . since by assumptions @xmath173 is not fixed by @xmath61 ( because @xmath121 is elementary ) , there is some @xmath174 such that @xmath175 and we want @xmath176 . note that @xmath169 is informative and has the same intervals of monotonicity as @xmath60 . at the second step we find extension @xmath177 of @xmath61 and @xmath178 such that @xmath179 is defined and is `` close enough '' in the above sense to the right endpoint of @xmath180 . we proceed in this way and stop as soon as the image of @xmath168 reaches @xmath181 , i.e , we obtain extensions @xmath182 , @xmath183 of @xmath60 and @xmath61 and a word @xmath184 , where @xmath185 such that @xmath186 is defined , lies in @xmath181 and @xmath187 . note that since we added to the domain of @xmath61 only points of the orbit of @xmath168 under @xmath82 , this implies @xmath188 . also by induction @xmath182 and @xmath183 are informative with the same decomposition into intervals of monotonicity as for @xmath60 and @xmath61 . the following figure illustrates the construction ( horizontal arrows indicate monotonicity of partial isomorphisms , bars stand for fixed points , the black dot is the minimal element @xmath168 , and gray dots are its images under @xmath77 ) : we now take extension @xmath189 , @xmath190 of @xmath182 and @xmath183 such that a. @xmath191 is defined for every @xmath192 ; b. @xmath189 and @xmath190 are informative with the same decomposition into intervals of monotonicity as for @xmath182 and @xmath183 ; c. minimum and maximum of the domains of @xmath189 and @xmath190 are equal to the minimum and maximum of the domains of @xmath182 and @xmath183 ; d. @xmath190 is increasing on @xmath181 ( this is possible since @xmath193 ) . set @xmath194 . finally extending @xmath190 to @xmath132 we can find @xmath195 such that @xmath196 and so let @xmath197 , then @xmath147 liberates @xmath60 in @xmath121 . [ fg - word - sym ] in the lemma above we started our construction from applying a power of @xmath60 , but we likewise could start it from applying a power of @xmath61 . [ rem : metcl ] we view rationals as a dense linear ordering without endpoints . but note that if would have a usual metric on @xmath45 then the above construction gives us @xmath131 , @xmath132 , and @xmath77 such that @xmath198 is as close in this metric to the endpoint @xmath199 as one wants . we will use this observation later . [ liberating - elem - strong ] for any elementary pair @xmath121 and any word @xmath82 there is a word @xmath83 , and partial isomorphisms @xmath131 and @xmath132 such that the triple @xmath200 liberates @xmath60 [ liberates @xmath61 ] in @xmath121 and @xmath201 ( i.e. , no cancellation between @xmath83 and @xmath82 happens ) . first we take extensions @xmath169 and @xmath177 of @xmath60 and @xmath61 respectively such that @xmath202 is defined for any @xmath203 , @xmath204 is elementary and @xmath205 @xmath206 by lemma [ liberating - elem - weak ] one can find a word @xmath83 and extensions @xmath131 , @xmath132 of @xmath169 , @xmath177 such that @xmath207 liberates @xmath169 in @xmath204 . by remark [ fg - word - sym ] we may also assume that there is no cancellation in @xmath91 . then @xmath208 liberates @xmath60 in @xmath121 . [ liberating - preserve ] let @xmath121 be a piecewise elementary pair and assume a triple @xmath147 liberates @xmath60 [ liberates @xmath61 ] in @xmath121 . let @xmath209 [ @xmath210 be a reduced word such that @xmath211 is irreducible . then there is a triple @xmath212 that liberates @xmath60 [ liberates @xmath61 ] in @xmath121 . moreover , one can take @xmath213 to be an extension of @xmath131 and @xmath214 to be an extension of @xmath132 . by the definition of liberation for piecewise elementary pairs it is enough to prove the statement for elementary triples only . so assume @xmath121 is elementary . since @xmath77 liberates @xmath60 in @xmath121 it has to start with a non - zero power @xmath215 of @xmath74 , i.e , @xmath216 . we prove the statement by induction on @xmath217 . if @xmath82 is empty the statement is trivial . now consider the inductive step . either @xmath218 and sign of @xmath219 matches the sign of @xmath215 ( because @xmath211 has to be reduced by assumptions ) or @xmath220 with @xmath221 . in the former case extend @xmath132 to @xmath222 in such a way that @xmath223 is defined for @xmath145 , then @xmath224 will be a @xmath60 liberating tuple by the item of the definition of liberation . in the second case we can find @xmath225 such that @xmath226 liberates @xmath132 in @xmath227 by taking @xmath225 such that @xmath228 for any @xmath229 . this proves the step of induction and the lemma . [ liberating - general ] let @xmath121 be a piecewise elementary pair and @xmath230 . then there is a triple @xmath147 that liberates @xmath60 [ liberates @xmath61 ] in @xmath121 and such that @xmath231 is reduced . we prove the statement by induction on the number of elementary components of @xmath121 . lemma [ liberating - elem - strong ] covers the base of induction . assume we have proved the lemma for @xmath232-many elementary components and inductively constructed a triple @xmath233 that liberates @xmath234 in @xmath235 , where @xmath234 and @xmath236 are restrictions of @xmath60 and @xmath61 onto the first @xmath232 many elementary components . consider the @xmath237 elementary component of @xmath121 . by lemma [ liberating - preserve ] and by the base of induction there is a word @xmath238 and extensions @xmath239 , @xmath240 of @xmath241 and @xmath242 such that @xmath243 liberates @xmath244 in @xmath245 . by remark [ fg - word - sym ] we may assume that @xmath246 is reduced . this proves the step of induction and hence the lemma . [ lem : density - words ] for any pair @xmath121 of partial isomorphisms and any word @xmath247 there are extensions @xmath131 and @xmath132 of @xmath60 and @xmath61 respectively and a reduced word @xmath231 such that @xmath248 for any @xmath249 . by lemma [ pelem - dense ] it is enough to prove the statement for a piecewise elementary pair @xmath121 . by lemma [ liberating - general ] we can find extensions @xmath182 , @xmath183 and a word @xmath83 such that @xmath250 liberates @xmath60 in @xmath121 . by the definition of liberation we can now extend @xmath182 to @xmath131 by declaring @xmath251 now set @xmath252 and @xmath253 . then @xmath248 for any @xmath249 . [ main - lemma ] fix a sequence @xmath254 of reduced words . for a generic @xmath255 there is a sequence of reduced words @xmath256 such that @xmath257 . take an enumeration @xmath258 of the rationals . let @xmath259 we claim that each @xmath260 is dense and open . indeed , assume for a certain @xmath1 one has @xmath261 . this is witnessed by a word @xmath77 . set @xmath262 and let @xmath263 , @xmath264 . then @xmath265 and so @xmath260 is open . density follows from lemma [ lem : density - words ] . now by baire theorem @xmath266 is a dense @xmath267 . the lemma follows . [ thm : rationals ] each two - dimensional topological similarity class in @xmath0 is meager . assume towards the contradiction that there is a pair @xmath268 that has a non - meager class of topological similarity . then by lemma [ main - lemma ] there must be a sequence @xmath269 of reduced words such that @xmath270 converges to the identity along this sequence ( we apply lemma [ main - lemma ] for the sequence @xmath271 ) . take and fix @xmath272 . set @xmath273 let @xmath274 then @xmath275 is open and dense in @xmath276 . to see density take a basic open set @xmath277 and assume @xmath278 ( the case when @xmath279 , but @xmath280 is similar ) . for some @xmath281 @xmath282 is not in the domain of @xmath60 . thus there are infinitely many values that @xmath283 may attain at @xmath284 , and so ( by induction ) there are infinitely many values that @xmath285 may attain for a pair @xmath286 . hence for some @xmath287 one has @xmath288 . and so @xmath275 is dense in @xmath289 . an application of baire theorem shows that @xmath290 is a dense @xmath267 and so for a generic @xmath291 one has @xmath292 . since @xmath293 we get a contradiction with the assumption that @xmath294 and that the class of topological similarity of @xmath270 is non - meager . we now turn to the group of homeomorphisms of the unit interval . this is a polish group in the natural topology , given by the basic open sets : @xmath295 ) } : \|g(a_i ) - f(a_i)\| < { \varepsilon}\}.\ ] ] we may write this neighborhood as @xmath296 , where @xmath297 is a partial isomorphism . since @xmath45 is dense in @xmath298 $ ] , we may assume that @xmath60 is a partial isomorphism of the rationals : this will give us a base of open sets . this group @xmath299)}$ ] has a normal subgroup of index @xmath120 , namely the subgroup @xmath48)}$ ] of order preserving homeomorphisms . let @xmath300)}$ ] , then @xmath301 naturally embeds into @xmath12 ( this embedding is a continuous injective homomorphism , its inverse , though , is not continuous ) , and the image of @xmath0 under this embedding is dense in @xmath12 . every two - dimensional class of topological similarity in @xmath12 is meager . we imitate the proof of theorem [ thm : rationals ] . let @xmath302 be an enumeration of the rationals @xmath303 $ ] , then @xmath302 is dense in @xmath298 $ ] . set @xmath304 @xmath305 note that @xmath306 is open for every @xmath26 and @xmath1 . then @xmath307 and so it is enough to prove that each two - dimensional class of topological similarity is meager in each of @xmath306 . let @xmath308 be a sequence of words such that for every piecewise elementary pair @xmath121 ( here @xmath60 and @xmath61 are partial isomorphisms of the rationals , like before ) there are infinitely many @xmath219 such that for some @xmath309 , @xmath310 , @xmath311 liberates @xmath60 in @xmath121 . then by lemma [ main - lemma ] for a generic pair @xmath312 there is a sequence of reduced words @xmath256 such that @xmath257 . this implies that for a generic pair @xmath313 there is a sequence @xmath314 as above ( because the topology in @xmath12 is coarser than in @xmath0 ) . if there is a non - meager two - dimensional class of topological similarity then there is a sequence of reduced words @xmath315 ( for some @xmath316 ) such that the set of pairs @xmath317 that converges to the identity along @xmath314 is non - meager . fix now @xmath318 and a sequence of reduced words @xmath256 . set @xmath319 each @xmath320 is open , we claim that it is also dense in @xmath306 . let @xmath321 be an open set . without loss of generality we may assume that @xmath322 , where @xmath60 and @xmath61 are partial isomorphisms of the rationals . let @xmath323 then there is @xmath324 and @xmath131 , @xmath132 such that @xmath325 liberates @xmath60 in @xmath121 . now repeat the proof of lemma [ liberating - preserve ] and use remark [ rem : metcl ] to get @xmath213 , @xmath214 that extend @xmath131 and @xmath132 and such that @xmath326 . hence each @xmath320 is dense in @xmath306 . now by baire theorem the intersection @xmath327 is a dense @xmath267 in @xmath306 and thus for any specific sequence @xmath314 the set of elements @xmath328 that converges to the identity along this sequence is meager in @xmath306 . finally we showed that each two - dimensional topological similarity class is meager in @xmath306 for any @xmath26 , @xmath1 and so is in @xmath329 . in this section we prove several results , that will be used later , when dealing with the randomly ordered urysohn space . but we believe that some of the theorems below are of independent interest for understanding the group of isometries of the urysohn space . we remind that the urysohn space @xmath330 is a complete separable metric space , that is uniquely characterized by the following properties : * every finite metric space can be isometrically embedded into @xmath330 ; * @xmath330 is ultrahomogeneous , that is each partial isometry between finite subsets of @xmath330 extends to a full isometry of @xmath330 . there is a rational counterpart @xmath331 of the urysohn space . it is called rational urysohn space . this is a countable metric space with rational distances , characterized by similar properties : * every finite metric space with rational distances can be isometrically embedded into @xmath331 ; * @xmath331 is ultrahomogeneous . groups of isometries @xmath332 and @xmath333 of these spaces are polish groups when endowed with the topology of pointwise convergence ( for this @xmath331 is viewed as a discrete topological space ) . let @xmath334 be a finite metric space with at least two elements . the _ density _ of @xmath22 is denoted by @xmath335 and is the minimal distance between two distinct points in @xmath22 : @xmath336 [ sec : extens - part - isom - ordmetric ] an ordered metric space is a triple @xmath337 , where @xmath338 is a metric on @xmath22 and @xmath339 is a linear ordering on @xmath22 . a _ partial isometry _ or _ partial isomorphism _ of a metric space @xmath340 is an isometry @xmath341 between _ finite _ subspaces @xmath22 , @xmath342 . a partial isomorphism of and ordered metric space is a partial isometry of the metric space that also preserves the ordering on its domain . [ sec : extens - part - isom - perpoint ] let @xmath60 be a partial isometry of a metric space . then we let @xmath62 to denote the domain of @xmath60 and @xmath63 to denote its range . a point @xmath343 is called _ periodic _ if there is a natural number @xmath146 such that @xmath344 the set of periodic points is denoted by @xmath345 . a point @xmath343 is called _ fixed _ if @xmath346 and the set of fixed points is denoted by @xmath67 . in this section we deal mostly with the classical urysohn space , but some of the results will be later applied to the randomly ordered rational urysohn space . the following proposition will let us do that . [ sec : isom - to - ord ] let @xmath22 be a finite ordered metric space , let @xmath60 be a partial isomorphism of @xmath22 . let @xmath23 be a finite metric space ( with no ordering ) and let @xmath61 be a partial isometry of @xmath23 with @xmath347 . suppose that @xmath348 as metric spaces and @xmath61 extends @xmath60 . if @xmath349 then there is a linear ordering on @xmath23 that extends an ordering on @xmath22 and such that @xmath61 becomes a partial isomorphism of an ordered metric space @xmath23 . we prove the statement by induction on @xmath350 . if @xmath351 the statement is obvious . for inductive step we consider two cases . * there is some @xmath352 such that @xmath353 but @xmath354 . then by assumption @xmath355 . extend now a linear ordering on @xmath22 to a partial ordering on @xmath356 by declaring for @xmath357 @xmath358 @xmath359 it is straightforward to check that this relation is indeed a partial ordering on @xmath356 . extend this partial ordering to a linear ordering on @xmath356 in any way . then @xmath61 is a partial isomorphism of @xmath356 and we apply the induction . * case 2 . * assume the opposite to the first case happens . then @xmath360 . take any @xmath361 ( if there is no such @xmath362 then @xmath363 and the statement is obvious ) . assume first that @xmath362 is not a fixed point of @xmath61 . then define a linear ordering on @xmath364 by declaring @xmath365 then @xmath61 is a partial isomorphism of @xmath364 and we can apply the induction hypothesis . if @xmath362 was a fixed point then we declare @xmath366 and , again , induction does the rest . let @xmath367 , @xmath368 , and @xmath369 be finite metric spaces and @xmath370 , @xmath371 be isometries . we define _ free amalgamation _ @xmath372 of metric spaces as follows : substituting @xmath23 and @xmath340 by their isomorphic copies we may assume that @xmath373 , set @xmath374 and define the metric @xmath375 by : @xmath376 note that the first and the second clauses agree for @xmath362 , @xmath377 . if @xmath22 is empty then we set @xmath378 , @xmath379 and @xmath380 the core of our arguments will be the following seminal result due to sawomir solecki established in 2005 , see @xcite . the second item is slightly modified compared to the original statement , but the modification follows from the proof in @xcite without any additional work . [ thm : solecki ] let a finite metric space @xmath22 and a partial isometry @xmath60 of @xmath22 be given . there exist a finite metric space @xmath23 with @xmath348 as metric spaces , an isometry @xmath182 of @xmath23 extending @xmath60 , and a natural number @xmath381 such that a. @xmath382 ; b. if @xmath383 is aperiodic then @xmath384 for @xmath385 , and moreover for any @xmath386 @xmath387 iff @xmath388 ; c. @xmath389 is the free amalgam of @xmath22 and @xmath390 over @xmath391 . moreover , the distances in @xmath23 may be taken from the additive semigroup generated by the distances in @xmath22 . [ sec : isom - rand - order - disjstruct ] let @xmath22 , @xmath23 , @xmath340 be metric spaces and let @xmath340 be embedded into @xmath22 and @xmath23 . we say that @xmath23 _ extends @xmath22 over _ @xmath340 if there exist and embedding @xmath370 such that the following diagram commutes : @xmath392^{i } & \\ { \mathbf{c}}\ar@{^{(}->}[u ] \ar@{^{(}->}[r ] & { \mathbf{b}}}\ ] ] we say that @xmath22 and @xmath23 are _ disjoint over _ @xmath340 if neither @xmath23 extends @xmath22 over @xmath340 nor @xmath22 extends @xmath23 over @xmath340 . [ sec : isom - rand - order - diffext ] let @xmath22 be a finite metric space , let @xmath60 be a partial isometry of @xmath22 , let @xmath343 be such that @xmath393 ( i.e. , @xmath394 ) . then there are metric spaces @xmath395 and @xmath396 that both extend @xmath22 : @xmath397 and @xmath398 , and partial isometries @xmath169 of @xmath395 and @xmath399 of @xmath396 that both extend @xmath60 and such that @xmath400 and @xmath401 are disjoint over @xmath402 . moreover , one can assume that @xmath403 @xmath404 @xmath405 apply theorem [ thm : solecki ] to get a full isometry @xmath182 of a finite metric space @xmath23 that extends @xmath60 and a natural number @xmath381 . set @xmath406 where @xmath407 is a new point , i.e. , a point not in @xmath23 . let @xmath408 denote the density of @xmath22 and fix an @xmath409 such that @xmath410 . we turn @xmath411 into a metric space by defining the distance between @xmath412 , @xmath413 as follows . @xmath414 we claim that @xmath415 is a metric space . we have to check the triangle inequality ( other conditions are obviously fulfilled ) . for this note that both @xmath416 and @xmath417 are isometrically embeddable into @xmath23 , where the triangle inequality is known to be satisfied . so one needs to prove two claims . * claim 1 . * for any @xmath418 @xmath419 if @xmath420 then the statement is obvious . if @xmath421 then @xmath422 and @xmath423 and claim 1 follows . * claim 2 . * for any @xmath418 @xmath424 @xmath425 note that for @xmath426 one has @xmath427 . from this both inequalities follow immediately . so @xmath411 is a metric space . we denote it by @xmath428 to signify the dependence on epsilon . define a partial isometry @xmath429 on @xmath428 by @xmath430 whenever @xmath431 and @xmath432 ; and @xmath433 . it is straightforward to check that @xmath429 is indeed a partial isometry . now the construction of two extensions that are disjoint over @xmath402 is easy . take , for example , two different @xmath434 , @xmath435 , @xmath436 such that @xmath437 let @xmath438 . then @xmath400 and @xmath401 are disjoint over @xmath402 . the main power of theorem [ thm : solecki ] is the explicit construction of an extension of a partial isometry to a full isometry of a finite metric space . moreover , this extension is as independent as possible . for our purposes we need only an extension to a partial isomorphism , but we want to keep the independence . let us state explicitly a corollary of the theorem that keeps all we need . [ cor : slawek ] for any finite metric space @xmath22 and a partial isometry @xmath60 there is finite metric space @xmath340 , a partial isometry @xmath169 of @xmath340 , which is an extension of @xmath60 , and a natural number @xmath381 such that a. @xmath439 ; b. @xmath440 is the amalgam of @xmath22 and @xmath441 over @xmath442 . c. for any @xmath443 @xmath444 moreover , the distances in @xmath340 are taken from the additive semigroup generated by the distances in @xmath22 , and hence the density is preserved @xmath445 . apply theorem [ thm : solecki ] to @xmath22 and @xmath60 to get a metric space @xmath23 , a full isometry @xmath182 of @xmath23 and a natural number @xmath381 . now set @xmath446 and @xmath447 . it is trivial to check that such a @xmath340 and @xmath169 satisfy the conditions . let @xmath448 be a metric space , let @xmath449 . we say that the distance @xmath450 _ passes through a point _ @xmath451 if @xmath452 we are going to apply corollary [ cor : slawek ] to partial isometries that also preserve an ordering . that is why we impose additional assumption : all periodic points are fixed points , i.e. , @xmath453 . [ thm : isomext ] let @xmath22 be a finite metric space . let @xmath60 and @xmath61 be two partial isometries of @xmath22 such that @xmath453 and @xmath347 . suppose @xmath454 . then there are finite metric space @xmath23 , extensions @xmath182 , @xmath183 of @xmath60 and @xmath61 respectively ( these extensions are partial isometries of @xmath23 ) and an element @xmath455 , @xmath456 such that a. [ thm : prop : fix ] @xmath457 , @xmath458 ; b. [ thm : prop : amal ] @xmath459 is the free amalgam of @xmath460 and @xmath461 over @xmath462 . moreover , the distances in @xmath23 are taken from the additive semigroup generated by the distances in @xmath22 , and hence @xmath463 , @xmath464 . let @xmath465 define inductively the sequence of elements @xmath466 , extensions @xmath467 , @xmath468 and metric spaces @xmath469 as follows : * step 0 : * let @xmath470 , @xmath471 , @xmath472 empty word , @xmath473 ; * step k : * if @xmath219 is odd then apply corollary [ cor : slawek ] to @xmath474 and @xmath475 to get @xmath467 and @xmath476 ; set @xmath477 , @xmath478 , @xmath479 . if @xmath219 is even do the same thing with roles of @xmath60 and @xmath61 interchanged . we claim that @xmath480 , @xmath481 , @xmath482 , and @xmath483 fulfill the requirements of the statement . let @xmath338 denote the metric on @xmath23 . it is obvious that @xmath484 and @xmath485 ( this is given by corollary [ cor : slawek ] at each stage ) . the moreover part is also obvious , since it is fulfilled at every step of the construction . it remains to show that for any @xmath486 and any @xmath487 one has @xmath488 note that by the last step of the construction for any @xmath486 and @xmath489 we have @xmath490 we first prove several claims . * claim 1 . * it is enough to show that @xmath491 holds for all @xmath492 and @xmath493 . _ proof of claim 1 . _ assume @xmath491 holds for all @xmath492 and @xmath493 . let @xmath494 , then for some @xmath495 @xmath496 by assumptions of the claim we get @xmath497 for some @xmath498 ; and so , by @xmath499 , @xmath500 this proves the claim . let @xmath501 denote @xmath502 . * claim 2 . * let @xmath503 , @xmath504 and suppose that for some @xmath498 and for some @xmath505 the distance between @xmath501 and @xmath362 passes through @xmath506 . then for any @xmath507 the distance between @xmath508 and @xmath362 passes through the same point @xmath506 . _ proof of claim 2 . _ this follows by induction . here is an inductive step . assume for definiteness that @xmath509 is odd ( the case , when @xmath509 is even , is similar ) . the distance between @xmath362 and @xmath510 passes through a point @xmath511 ( @xmath512 if @xmath513 is even ) . then @xmath514 @xmath515 but @xmath516 ( this is because @xmath517 and @xmath518 is fixed by @xmath60 ) . hence @xmath519 but also @xmath520 and so @xmath521 . this proves the claim . * claim 3 . * let @xmath522 ( here @xmath523 is a symmetric difference of sets ) , @xmath524 . suppose that the distance between @xmath525 and @xmath362 does nt pass through a point in @xmath119 . then @xmath526 . _ proof of claim 3 . _ suppose first that @xmath527 . we prove the statement by induction on @xmath505 . the base of induction is trivial , so we show the inductive step : the statement is true for @xmath505 and we need to show it for @xmath528 . if @xmath505 is even then , since @xmath529 and because @xmath530 ( this is since @xmath505 is even and @xmath531 ) the statement follows immediately . so , assume @xmath505 is odd . then the distance between @xmath532 and @xmath362 passes through a point @xmath533 . now two things can happen . suppose first for some @xmath534 the distance between @xmath508 and @xmath506 passes through a point @xmath535 . then by claim 2 , the distance between @xmath506 and @xmath532 must pass through @xmath518 . now @xmath536 and so the distance between @xmath532 and @xmath362 passes through a point @xmath535 . contradiction with the assumptions of the claim . so , for no @xmath534 the distance between @xmath508 and @xmath362 passes through a point in @xmath119 . then , applying induction to @xmath501 and @xmath506 , we get @xmath537 . but since @xmath538 and since @xmath539 we get @xmath540 in the case , when @xmath541 , the distance increases by @xmath335 at even stages of the construction , the rest of the argument for this case is similar . the claim is proved . fix now @xmath504 and @xmath493 . it remains to show that @xmath542 we now have two cases ( we ll show , though , that case @xmath120 is impossible ) . * case 1 . * for some @xmath543 the distance between @xmath525 and @xmath407 passes through a point @xmath544 . then @xmath545 applying claim 2 for @xmath546 , we get @xmath547 and the theorem is proved for this case . * case 2 . * for no @xmath543 the distance between @xmath501 and @xmath407 passes through a point in @xmath548 . then by claim 3 @xmath549 let , on the other hand , @xmath498 be any common fixed point . then @xmath550 . so this case never happens . [ sec : main - remark ] note that the same result is also true for ordered metric spaces . for this one just has to apply proposition [ sec : isom - to - ord ] at each step of the construction of @xmath182 and @xmath183 . before we apply this result to the classes of topological similarity let us mention another application . for a subset @xmath551 ( @xmath552 let @xmath553 ( @xmath554 , respectively ) denote the subgroup of isometries that pointwise fix @xmath22 . recall a theorem of julien melleray from @xcite . [ sec : isom - rand - order - gm ] let @xmath330 be the urysohn space , let @xmath555 be two finite subsets . then @xmath556 let us give an equivalent reformulation of the above result . [ sec : isom - rand - order - gm2 ] let @xmath330 be the urysohn space , let @xmath555 be two finite subsets . then for any @xmath409 , for any @xmath557 , for any @xmath558 there is @xmath559 such that @xmath560 we show that one can actually eliminate the epsilon in the above reformulation . [ sec : isom - rand - order - noeps ] let @xmath330 be the urysohn space , let @xmath555 be two finite subsets . then for any @xmath557 for any @xmath558 there is @xmath559 such that @xmath561 without loss of generality we may assume that @xmath562 and @xmath342 . let @xmath563 then @xmath564 is a partial isometry of @xmath565 . define two partial isometries @xmath169 and @xmath399 of @xmath565 by @xmath566 @xmath567 apply now theorem [ thm : isomext ] to @xmath169 , @xmath399 and @xmath565 to get a metric space @xmath568 and extension @xmath177 of @xmath169 and @xmath569 of @xmath399 , and a word @xmath570 . extend now @xmath177 by @xmath571 such a @xmath177 is then a partial isometry of @xmath568 . now extend @xmath177 and @xmath569 to full isometries ( we still denote them by the same symbols ) and set @xmath572 then for any @xmath573 , @xmath574 , and @xmath575 , @xmath576 . note that if we start from metric spaces with rational distances , then the space @xmath568 , constructed in the proof , would also have rational distances . and we arrive at the let @xmath331 be the rational urysohn space , let @xmath577 be two finite subsets . then @xmath578 there is a rich variety of linearly ordered frass limits , of which countable dense linear ordering without endpoints is the simplest example . in fact , as proved in @xcite , if the group of automorphisms of a particular frass class @xmath29 is extremely amenable , then there is a linear ordering on the frass limit of @xmath29 that is preserved by all automorphisms . let us briefly remind the definition of this structure . formally speaking , one has to consider the frass class @xmath580 of finite ordered metric spaces with rational distances . then @xmath579 is , by definition , the frass limit of @xmath580 . intuitively one can think of this structure as a classical rational urysohn space with a linear ordering on top ( such that ordering is isomorphic to the ordering of the rationals ) and such that this ordering is independent of the metric structure . our goal is to prove that every two - dimensional class of topological similarity in the group of automorphisms of @xmath579 is meager . we would like to emphasize that the structure of conjugacy classes in @xmath20 and @xmath581 is substantially different . as was mentioned earlier there is a generic conjugacy class in @xmath20 , while it is not hard to derive from results in @xcite , that each conjugacy class in @xmath581 is meager . class @xmath29 of finite structures satisfies _ weak amalgamation property _ ( wap for short ) if for every @xmath32 there is @xmath31 and an embedding @xmath582 such that for all @xmath33 , @xmath583 and all embeddings @xmath584 , @xmath585 there are @xmath586 and embeddings @xmath587 , @xmath588 such that @xmath589 , i.e , in the following diagram paths from @xmath22 to @xmath590 commute ( but not necessarily paths from @xmath23 to @xmath590 ) . let @xmath29 be a frass class . we associate with it a class of structures @xmath592 . elements of @xmath592 are partial isomorphisms of @xmath29 , more precisely tuples @xmath593 where @xmath22 , @xmath594 and @xmath595 , @xmath594 , @xmath596 and @xmath60 is an isomorphisms . by theorem [ thm : localwap ] it is enough to show that the class @xmath597 has no local wap . let @xmath598 , and assume without loss of generality that @xmath599 has at least one non - fixed point ( otherwise take an extension of @xmath599 ) . we claim that the class of structures that extend @xmath22 does not have wap . fix @xmath600 that extends @xmath22 and assume for notational simplicity that @xmath348 . let @xmath601 be such that @xmath602 and let @xmath603 be the orbit of @xmath362 under @xmath599 . then @xmath604 . now take ( by lemma [ sec : isom - rand - order - diffext ] and proposition [ sec : isom - to - ord ] ) two structures @xmath605 , @xmath606 with embeddings @xmath607 , @xmath608 and such that @xmath609 and @xmath610 are disjoint over @xmath603 . then there is no weak amalgamation of @xmath611 and @xmath612 over @xmath613 and @xmath411 . with classes of topological similarity the situation is rather different . all non - trivial elements in @xmath581 fall into a single class of topological similarity . and more generally , if @xmath614 is any countable linearly ordered structure and @xmath615 is endowed with a topology of pointwise convergence ( @xmath614 is discrete here ) , then @xmath615 has exactly two classes of topological similarity ( unless @xmath616 , then , of course , there is only one ) : all non - trivial automorphisms generate a discrete copy of @xmath617 and hence fall into a single class . thus , in spite of the previous proposition , it makes sense to ask if there is a non - meager two - dimensional similarity class in @xmath581 . it turns out that the analog of theorem [ thm : rationals ] for the randomly ordered urysohn space takes place . let us first briefly sketch the idea of the proof before diving into the details . we will prove that , again , for a generic pair there is a sequence of reduced words , such that this pair converges along it . one can repeat all the arguments up to lemma [ main - lemma ] ( only obvious changes are necessary ) . so one gets for a piecewise elementary pair @xmath121 a triple @xmath147 that liberates @xmath60 in @xmath121 . but now , opposite to the case of rationals , one can not in general declare that @xmath618 for @xmath145 , since such a @xmath131 may be not an isometry . at this moment we have to take further extensions of @xmath131 and @xmath132 . but once an analog of lemma [ main - lemma ] is proved for the urysohn case , the rest of theorem [ thm : rationals ] goes unchanged . if @xmath60 is a partial isometry , we can use amalgamation of its domain with a one point metric space over an empty set to add a fixed point for @xmath60 . using this observation the following two lemmata , which are analogs of lemma [ liberating - general ] and lemma [ liberating - preserve ] , are proved like for the rationals , and we omit the details . [ lem : ur - libpres ] let @xmath121 be a piecewise elementary pair of partial isomorphisms of @xmath579 and assume a triple @xmath147 liberates @xmath60 [ liberates @xmath61 ] in @xmath121 . let @xmath209 [ @xmath210 be a reduced word such that @xmath211 is irreducible . then there is a triple @xmath212 that liberates @xmath60 [ liberates @xmath61 ] in @xmath121 . moreover , one can take @xmath213 to be an extension of @xmath131 and @xmath214 to be an extension of @xmath132 . [ lem : ur - libgen ] let @xmath121 be a piecewise elementary pair of partial isomorphisms of @xmath579 and @xmath230 be a reduced word . then there is a triple @xmath200 that liberates @xmath60 in @xmath121 [ liberates @xmath61 ] and such that @xmath619 . [ lem : ur - lib - words ] for any pair @xmath121 of partial isomorphisms of the @xmath579 and any word @xmath230 there are extensions @xmath131 and @xmath132 of @xmath60 and @xmath61 respectively and a reduced word @xmath620 such that @xmath621 for any @xmath249 . we can assume that @xmath121 is piecewise elementary . by lemma [ lem : ur - libgen ] there are extensions @xmath622 , @xmath623 of @xmath60 and @xmath61 and a reduced word @xmath91 such that @xmath624 liberates @xmath60 in @xmath121 . now apply theorem [ thm : isomext ] ( with remark [ sec : main - remark ] and lemma [ liberating - preserve ] ) to @xmath622 , @xmath623 and @xmath625 to get extension @xmath182 and @xmath183 and a reduced word @xmath626 . note that @xmath627 is reduced , because @xmath83 starts from a power of @xmath74 and @xmath626 by construction ends in a power of @xmath73 . by the item of theorem [ thm : isomext ] we can extend @xmath182 to @xmath131 by declaring @xmath628 set @xmath252 and @xmath629 . it is easy to see that @xmath248 holds for any @xmath249 . all the results in this section can be proved for the randomly ordered random graph in the same way , as they were proved for the randomly ordered rational urysohn space . one can also formally deduce this case from the above results viewing graphs as metric spaces with all the distances in @xmath630 . 99 e. akin , e. glasner , b. weiss , _ generically there is but one self homeomorphism of the cantor set _ , trans . ( 2008 ) , no . 7 , 36133630 . frass r. , _ theory of relations , north - holland _ , 1986 . hodges w. , _ model theory _ , cambridge univ . press . kechris , a. s. , pestov , v. g. , todorcevic , s. , _ frass limits , ramsey theory , and topological dynamics of automorphism groups_. geom . anal . 15 ( 2005 ) , 1 , 106189 . a. s. kechris , c. rosendal , _ turbulence , amalgamation and generic automorphisms of homogeneous structures _ , london math . soc . , 94 ( 2007 ) , 349371 . j. melleray , _ topology of the isometry group of the urysohn space _ 207 ( 2010 ) , no . 3 , 273287 . c. rosendal , _ the generic isometry and measure preserving homeomorphism are conjugate to their powers _ , fund . , 205 ( 2009 ) , no . solecki , s. , _ extending partial isometries _ , israel j. math . 150 ( 2005 ) , 315332 . j.k . truss , _ on notions of genericity and mutual genericity _ , journal of symbolic logic , 72 ( 2007 ) , 755766 . urysohn , _ sur un espace mtrique universel _ , c. r. acad . paris 180 ( 1925 ) 803806 . urysohn , _ un espace mtrique universel _ , bull . 51 ( 1927 ) , 4364 and 7490 .	we show that every two - dimensional class of topological similarity , and hence every diagonal conjugacy class of pairs , is meager in the group of order preserving bijections of the rationals and in the group of automorphisms of the randomly ordered rational urysohn space .
in spite of its apparently limited scope the study of the electroweak corrections ( ewc ) has played a crucial role in three different directions : * provide evidence for the consistency at the quantum level of the gauge and the higgs sectors of the standard model ( sm ) ; * give significant information , although indirect , on the existence and the mass of the top quark and the higgs boson before their direct discovery ; * constrain and orient the search for possible new physics beyond the sm . although not making justice to important earlier works on the subject , in order to recall the progressive role of the ewc it makes sense , as we do in the following , to set the @xmath0 time in 1971 , when t hooft proved the renormalizability of the electroweak sector of the sm@xcite . since then , this means more than forty years of progress until the discovery of the higgs boson , which offers new opportunities for precision physics . a few papers of veltman in 1977 played a key role both in establishing the program of computing the ewc to physical observables in the sm and in pointing out their possible sensitivity to physics in the 100 gev range or above , even by measurements of lower energy processes@xcite . the relevant physical observables are determined at tree level by three parameters : the two gauge couplings in the electroweak sector , @xmath1 and @xmath2 , and the vacuum expectation value of the higgs boson doublet @xmath3 . to fix these parameters , three basic observables may be chosen , which is equivalent to a renormalization procedure . since the beginning of the electroweak correction program , two such observables have emerged : the electromagnetic fine structure constant , @xmath4 , as measured by the josephson effect or the electron @xmath5 , and the fermi constant , @xmath6 , as determined from the muon lifetime . in the seventies the third observable was more often coming from the @xmath7-@xmath8 or the @xmath7-hadron cross sections , in particular the ratio @xmath9 between the neutral and the charged current processes , with a determination , at that time , of the weak mixing angle @xmath10 at about @xmath11 level . the emergence from @xmath12 and @xmath13 processes of a @xmath14 parameter close to unity , as predicted at tree level by the sm , was recognized by veltman as evidence for the isodoublet nature of the higgs boson and later , more generally , as a consequence of an approximate _ custodial symmetry_@xcite . at the same time the finding of @xmath14 close to unity motivated to establish the ewc program . the absence of any other precisely measured observable , as the same relatively large error on @xmath10 itself from neutrino scattering , could not allow any significant quantitative comparison with the ewc . yet , as already mentioned , the sensitivity to heavy non - decoupled particles was identified and relevant one loop effects calculated both due to a split fermion doublet or to the higgs boson . a first calculated one loop effect was the correction to the @xmath14-parameter , the ratio @xmath15 between the neutral - current fermi constant @xmath16 , defined in analogy with @xmath6 from the elastic electron - neutrino scattering amplitude at @xmath17 , and the charged - current fermi constant itself . the deviation from unity of @xmath14 was calculated from a charged lepton of mass @xmath18 , split from a massless neutrino in the same isodoublet@xcite @xmath19 this is the same correction as for a heavy top , of mass @xmath20 , and a quasi massless bottom up to the extra colour factor of @xmath21 that has to be introduced . the neutrino data mentioned above allowed to set an upper bound on @xmath22 of 800 gev , or - knowing now that the top is also heavy - of about 450 gev on the top mass . somewhat later came the calculation in the sm of the two independent one loop corrections growing like the logarithm of the higgs mass , @xmath23@xcite . as recalled later on , these corrections affect in a precisely defined way every observable . for example one has @xmath24 @xmath25 veltman called `` screening theorem '' the observation of the mild logarithmic dependence on the higgs mass , taken much larger than @xmath26 , of all the one loop corrections . the root of this theorem is in the absence in the unitary gauge of any coupling of the w and the z to the higgs boson proportional to the quartic higgs coupling @xmath27 . although clear evidence for the ewc had to wait for the @xmath29 data at the z - pole , new more precise measurements came in the eighties from various sources : the direct measurements of the w and z masses at hadron colliders , @xmath7-@xmath30 and @xmath7-hadron scattering , atomic parity violation and @xmath29 collisions at various energies below the z - boson mass . this quickly led the focus on the constraints that could be inferred from these data on the top quark mass and , at least indirectly , on the higgs boson mass , which was rapidly recognized as relevant to the constraint on the top mass itself . at the summer conferences in 1989 , with the lep operation just about to start , it was possible to quote the following allowed ranges for the top mass@xcite : @xmath31 @xmath32 the fit leading to these constraints had already incorporated as well the fresh first measurement of the z mass by slc , @xmath33 gev , close to _ per mille _ precision@xcite . at the same summer conferences ua1 , ua2@xcite and cdf@xcite were quoting lower bounds on the top mass in the @xmath34 gev range . with the same motivations in mind several calculations were also made , among which : i ) the one loop heavy - top correction to the @xmath35 coupling@xcite @xmath36 ii ) the two loop heavy - top corrections to the @xmath14-parameter@xcite and iii ) the corrections , again to the @xmath14-parameter , growing linearly with @xmath37 and occurring at two loops@xcite . a comprehensive and efficient calculation of all these effects was made possible by the observation that they come from the so called _ gaugeless limit _ of the sm , through a loop expansion controlled only by the top yukawa coupling , @xmath38 , and by the quartic higgs boson coupling , but they also have nothing to do with the gauging of the electroweak interactions and , as such , are relatively simple to calculate.]@xcite . as an example the corrections to the previously defined quantities , the only ones that grow like powers of @xmath39 , take a simple form up to two loop order @xmath40 @xmath41 these equations hold for @xmath42 , but simple closed forms can be given for any value of @xmath23 and @xmath28 . to get these corrections the w and the z bosons only enter in loops through their goldstone boson components with couplings controlled by @xmath38 and @xmath27 . similarly observables with w and z bosons as external legs are immediately related , via simple ward identities , to amplitudes for the corresponding goldstone bosons . these and other two loop calculations@xcite , quickly incorporated in programs like zfitter , topaz0 or similar , made possible a comparison at the _ per mille _ level of the sm with a number of precision observables measured at the z pole in the early nineties . consequently the constraint on the top mass became increasingly precise . in the winter conferences of 1994 an indirect determination of the top mass was quoted as@xcite @xmath43 with the second error due to the higgs mass scanning the @xmath44 gev range . alternatively one had @xmath45 these numbers can be compared with the value of @xmath46 in the cdf paper , published a few months later , that provided the first evidence for direct @xmath47 production@xcite . a pretty analogous story for the higgs boson mass had to wait twenty years more to be accomplished . in july 2012 the higgs boson discovery was announced with a mass @xmath48 gev ( atlas@xcite ) and @xmath49 gev ( cms @xcite ) , against an indirect determination from the electroweak fit of @xmath50 gev . the indirect sensitivity of the electroweak precision tests ( ewpt ) to physics at higher energies has not only played a role in constraining @xmath28 and @xmath23 , but has also been seen from the beginning as a possible search for signals of new physics . as a matter of fact , until the discovery of at least the top , the separation between the two issues was to some extent not easy to achieve . there are three different ways that can be used to try to identify the sensitivity of the ewpt to new physics , with relations to each other : i ) the examination of effects in specific models ; ii ) the identification of effective parameters or of effective observables relevant in specific contexts ; iii ) the use of effective field theory ( eft ) . for clarity , to describe them it is better not to follow a strict chronological order . inspired by the early work on heavy higgs and top in the sm , as already described , and in preparation for the @xmath29 colliders working at the @xmath51 pole , attention was drawn to the contributions to precision observables from the electroweak vacuum polarizations amplitudes@xcite @xmath52 \eta^{\mu\nu } + ( q^\mu q^\nu\mathrm{\small -terms})\ ] ] with @xmath53 or possibly @xmath54 for the @xmath55 or the @xmath56 bosons respectively . of particular relevance in this context are two quantities @xmath57@xcite @xmath58 whenever an expansion in @xmath59 is meaningful , @xmath60 is the leading parameter describing `` custodial symmetry '' breaking , since @xmath61 , whereas @xmath62 is the leading term that respects it . we shall see in a moment how @xmath62 and @xmath60 relate to physical observables . although @xmath62 and @xmath60 are normally defined as due to new physics only , in the sm they are the only terms receiving the one loop logarithms in a large @xmath23 expansion that we have already encountered , @xmath63 after electroweak symmetry breaking there are four non vanishing vacuum polarization amplitudes , @xmath64 and @xmath65 , which , expanded up to first order in @xmath59 , are determined by eight coefficients . three of them are absorbed in the definition of the parameters @xmath66 and two are related by gauge invariance of unbroken electromagnetism : @xmath67 . as a consequence , other than @xmath62 and @xmath60 , it is useful to introduce a third quantity@xcite @xmath68 which , however , while also breaking `` custodial symmetry '' , is subleading with respect to @xmath60 in a @xmath59 expansion and , as such , does not contain any `` large '' @xmath69 term . at second order in @xmath59 four more parameters appear@xcite . unlike @xmath70 , two of these four extra parameters respect the electroweak gauge group , but , as @xmath70 , they are also calculable in the sm , since they are related to operators of dimension 6 . as already mentioned , observables at the @xmath51 pole have played and are still playing a major role in the comparison between theory and experiment . this rapidly led to realize that every new physics respecting quark - lepton and flavour universality could affect the @xmath51 pole observables and the w - mass only via three effective parameters @xmath71 , defined in terms of @xmath72 in eq . ( [ deltar ] ) and of the effective couplings @xmath73 to the @xmath51 boson of a fermion @xmath74 of non zero charge @xmath75 and third component of the weak isospin @xmath76 as@xcite @xmath77 @xmath78 where @xmath79 a few comments on these parameters are useful to make . eq.s ( [ ga ] ) and ( [ gvga ] ) refer to any light charged fermion except the @xmath80-quark , where already in the sm there is the flavour non - universal correction ( [ z->bb ] ) due to top loops . to incorporate it together with any other possible new physics effect of the same kind , thus breaking flavor universality , one has to include another parameter @xmath81 in the effective couplings of the @xmath80 quark@xcite : @xmath82 in the parameters @xmath83 there is no reference to a @xmath59 expansion nor any restriction to vacuum polarization corrections only . an explicit expression for the vacuum polarization contributions to @xmath84 is @xcite @xmath85 where @xmath86 @xmath87 with new physics in oblique corrections only and a characteristic scale @xmath88 such that an expansion in @xmath89 is meaningful , a comparison of these equations with eq.s ( [ st]-[u ] ) shows that @xmath84 may be approximated by @xmath90 respectively and @xmath60 contain the leading @xmath91 terms , the full dependence of the ewc on @xmath92 is only included in the @xmath83@xcite ] . two - dimensional probability distributions for @xmath93 and @xmath94 at @xmath95 ( the dark region ) and @xmath96 ( the light region ) with all four @xmath97 floating ( upper left ) or fixing @xmath98 ( upper right),title="fig:",scaledwidth=44.0% ] two - dimensional probability distributions for @xmath93 and @xmath94 at @xmath95 ( the dark region ) and @xmath96 ( the light region ) with all four @xmath97 floating ( upper left ) or fixing @xmath98 ( upper right),title="fig:",scaledwidth=44.0% ] two - dimensional probability distributions for @xmath93 and @xmath94 at @xmath95 ( the dark region ) and @xmath96 ( the light region ) with all four @xmath97 floating ( upper left ) or fixing @xmath98 ( upper right),title="fig:",scaledwidth=44.0% ] note finally that with the definitions ( [ ga]-[eps_b ] ) the @xmath83 include every electroweak correction , both from the sm and from possible new physics , so that it is useful to define the separation @xmath100 . now that both @xmath28 and @xmath23 are known with significant accuracy , so are the @xmath101@xcite : @xmath102 this makes possible to plot the two - dimensional probability distributions from current data for @xmath103 in fig.s 1 , when all four @xmath97 are allowed to float or fixing some of them to zero@xcite . it depends on a specific model of new physics which one among fig.s 1 has to be used to constrain it . as one can see from these figures , the size of the error on the @xmath83 is currently at the @xmath104 level , which means , comparing with ( [ eps_i_sm ] ) , that the electroweak radiative corrections in the sm have been seen at about the @xmath105 level of their typical value . interestingly this is about the same level of precision at which the loops in flavour changing neutral current processes are currently tested . so far no deviation from the sm has emerged in either case . it is customary to view the sm as an effective low energy approximation of a more fundamental theory involving the same low energy degrees of freedom and already valid at shorter distances . as a consequence a more complete lagrangian describing physics at any scale below a cutoff @xmath106 is @xmath107 in terms of any gauge invariant operator @xmath108 of dimension @xmath109 , made of the sm fields . while the presence of these extra operators has in itself nothing to do with the ewc , the study of their effects has proven effective in describing the implications of the ewpt . in turn such effects typically arise from the interference of these operators with radiatively corrected sm amplitudes . this approach has indeed been pursued already in the nineties during the first decade of lep operation and is receiving further recent attention through the measurements of the higgs boson properties . the relative drawback of this approach is in the large number of independent operators that are involved and can affect many different observables . nevertheless it is interesting to recall the most significant bounds on the scale @xmath106 that were first obtained after a decade of lep operation@xcite and compare them with the corresponding current bounds@xcite . they are shown in table 1 at @xmath96 c.l .. the bounds are obtained by considering one operator at a time with the relative coefficient @xmath110 , since what counts is the interference with the radiatively corrected sm amplitudes . furthermore , in the columns from the data in the year 2000 , the higgs boson mass was taken at 115 gev , consistent with the direct bound and corresponding to the indirect indication of the ewpt at that time . the first operator in table 1 contributes as well to correct the higgs boson width into a pair of photons , which is loop suppressed in the sm . from current data on this width one obtains the bounds@xcite @xmath111 these are the strongest bounds obtainable to date from the higgs boson coupling measurements . the bounds on @xmath106 in the year 2000 , compared with the `` naturalness '' estimate of the radiative corrections to the higgs boson mass @xmath112 and the already mentioned indication for a light higgs boson in the same year 2000 , motivated the title `` lep paradox '' of ref . @xcite . needless to say , the identification of the two @xmath106 parameters entering in eq.s ( [ lbsm ] ) and ( [ naturalness ] ) has to be taken with a grain of salt , since in suitable extensions of the sm the squares of these parameters are related to each other by a loop factor . . @xmath96 lower bounds on @xmath113tev for the individual operators . the first two columns are from the year 2000 data @xcite , whereas the third and fourth columns are from the currently available data@xcite . [ cols="<,^,^,^,^",options="header " , ] all the content of sections 3 and 4 rests on the high precision that has been achieved so far in computing the ewc in the sm . some key quantities like the w mass ( or @xmath72 in eq . ( [ deltar ] ) ) and the effective leptonic mixing angle @xmath114 have been computed in the sm to full two loop order , which consists of many different contributions@xcite due to the several coupling constants and particles involved , and including as well some relevant 3 loops , going e.g. like @xmath115 or @xmath116@xcite . as a result the current theoretical uncertainty from the missing higher order terms , as inferred for example from comparing the calculations in different renormalization schemes , is estimated for @xmath117 at about @xmath118 mev and on @xmath119 at about @xmath120 . to these uncertainties one has to add the parametric errors of 9 mev for @xmath117 and of @xmath121 for @xmath119 , obtained with @xmath122 , @xmath123 and @xmath124@xcite . this has to be compared with the current experimental errors of 15 mev for @xmath117 and @xmath125 for @xmath119 . all these numbers are collected in table 2 . while the theoretical uncertainties look adequate to compare with current precision measurements , significant improvements will be necessary to make full use of the precision foreseen at future facilities like , e.g. , fcc - ee@xcite , also shown in table 2 . it is a pleasure to thank g. altarelli , m. beccaria , b. bellazzini , f. caravaglios , p. ciafaloni , g. curci , m. frigeni , s. jadach , l. maiani , a. pomarol , r. rattazzi , s. rychkov , a. strumia , a. varagnolo , a. vicere , for their collaborations on the subject treated in this work . i am indebted with s. mishima and l. silvestrini for providing me with fig , s 1 and to g. degrassi for help on the uncertainties in table 2 . this work is supported in part by the european programme unification in the lhc era " , contract pitn - ga-2009 - 237920 ( unilhc ) and by miur under the contract 2010yj2nyw-010 . g. t hooft , nucl . b * 33 * ( 1971 ) 173 . g. t hooft , nucl . b * 35 * ( 1971 ) 167 . m. j. g. veltman , nucl . b * 123 * ( 1977 ) 89 . m. j. g. veltman , phys . b * 70 * ( 1977 ) 253 . p. sikivie , l. susskind , m. b. voloshin and v. i. zakharov , nucl . b * 173 * ( 1980 ) 189 . g. passarino and m. j. g. veltman , nucl . b * 160 * ( 1979 ) 151 . f. antonelli , m. consoli and g. corbo , phys . b * 91 * ( 1980 ) 90 . m. j. g. veltman , phys . b * 91 * ( 1980 ) 95 . w. j. marciano and a. sirlin , phys . d * 22 * ( 1980 ) 2695 [ erratum - ibid . d * 31 * ( 1985 ) 213 ] . j. r. ellis and g. l. fogli , phys . b * 232 * ( 1989 ) 139 . p. langacker , phys . * 63 * ( 1989 ) 1920 . g. s. abrams , c. adolphsen , r. aleksan , j. p. alexander , m. a. allen , w. b. atwood , d. averill and j. ballam _ et al . _ , phys . lett . * 63 * ( 1989 ) 724 . j. m. gaillard , nucl . phys . proc . * 16 * ( 1990 ) 30 . l. j. nodulman , nucl . * 16 * ( 1990 ) 40 . j. bernabeu , a. pich and a. santamaria , phys . b * 200 * ( 1988 ) 569 . j. j. van der bij and f. hoogeveen , nucl . b * 283 * ( 1987 ) 477 . j. van der bij and m. j. g. veltman , nucl . b * 231 * ( 1984 ) 205 . j. fleischer , o. v. tarasov , f. jegerlehner and p. raczka , phys . b * 293 * ( 1992 ) 437 . j. fleischer , o. v. tarasov and f. jegerlehner , phys . b * 319 * ( 1993 ) 249 . m. koratzinos , in * la thuile 1994 , results and perspectives in particle physics * 319 - 332 f. abe _ et al . _ [ cdf collaboration ] , phys . * 73 * ( 1994 ) 225 [ hep - ex/9405005 ] . g. aad _ et al . _ [ atlas collaboration ] , phys . b * 716 * ( 2012 ) 1 [ arxiv:1207.7214 [ hep - ex ] ] . s. chatrchyan _ et al . _ [ cms collaboration ] , phys . b * 716 * ( 2012 ) 30 [ arxiv:1207.7235 [ hep - ex ] ] . m. consoli , s. lo presti and l. maiani , nucl . b * 223 * ( 1983 ) 474 . d. c. kennedy and b. w. lynn , nucl . b * 322 * ( 1989 ) 1 . m. e. peskin and t. takeuchi , phys . * 65 * ( 1990 ) 964 ; phys . d * 46 * ( 1992 ) 381 . r. barbieri , a. pomarol , r. rattazzi and a. strumia , nucl . b * 703 * ( 2004 ) 127 [ hep - ph/0405040 ] . g. altarelli and r. barbieri , phys . b * 253 * ( 1991 ) 161 . g. altarelli , r. barbieri and f. caravaglios , nucl . b * 405 * ( 1993 ) 3 . r. barbieri , m. frigeni and f. caravaglios , phys . b * 279 * ( 1992 ) 169 . v. a. novikov , l. b. okun and m. i. vysotsky , nucl . b * 397 * ( 1993 ) 35 . a. orgogozo and s. rychkov , jhep * 1306 * ( 2013 ) 014 [ arxiv:1211.5543 [ hep - ph ] ] . m. ciuchini , e. franco , s. mishima and l. silvestrini , jhep * 1308 * ( 2013 ) 106 [ arxiv:1306.4644 [ hep - ph ] ] . m. ciuchini , e. franco , s. mishima , m. pierini , l. reina and l. silvestrini , arxiv:1410.6940 [ hep - ph ] . a. falkowski , f. riva and a. urbano , jhep * 1311 * ( 2013 ) 111 [ arxiv:1303.1812 [ hep - ph ] ] . a. pomarol and f. riva , jhep * 1401 * ( 2014 ) 151 [ arxiv:1308.2803 [ hep - ph ] ] . j. de blas , m. ciuchini , e. franco , d. ghosh , s. mishima , m. pierini , l. reina and l. silvestrini , arxiv:1410.4204 [ hep - ph ] . r. barbieri and a. strumia , phys . b * 462 * ( 1999 ) 144 [ hep - ph/9905281 ] . r. barbieri and a. strumia , hep - ph/0007265 . r. barbieri and a. tesi , phys . d * 89 * ( 2014 ) 5 , 055019 [ arxiv:1311.7493 [ hep - ph ] ] . r. barbieri , t. gregoire and l. j. hall , hep - ph/0509242 . r. barbieri , b. bellazzini , v. s. rychkov and a. varagnolo , phys . d * 76 * ( 2007 ) 115008 [ arxiv:0706.0432 [ hep - ph ] ] . r. barbieri and l. maiani , nucl . b * 224 * ( 1983 ) 32 . l. alvarez - gaume , j. polchinski and m. b. wise , nucl . b * 221 * ( 1983 ) 495 . j. fan , m. reece and l. t. wang , arxiv:1412.3107 [ hep - ph ] . g. degrassi , p. gambino and a. vicini , phys . b * 383 * ( 1996 ) 219 [ hep - ph/9603374 ] . g. degrassi , p. gambino and a. sirlin , phys . b * 394 * ( 1997 ) 188 [ hep - ph/9611363 ] . a. freitas , w. hollik , w. walter and g. weiglein , phys . b * 495 * ( 2000 ) 338 [ erratum - ibid . b * 570 * ( 2003 ) 260 ] [ hep - ph/0007091 ] ; nucl . b * 632 * ( 2002 ) 189 [ erratum - ibid . b * 666 * ( 2003 ) 305 ] [ hep - ph/0202131 ] . m. awramik and m. czakon , phys . * 89 * ( 2002 ) 241801 [ hep - ph/0208113 ] ; phys . b * 568 * ( 2003 ) 48 [ hep - ph/0305248 ] . m. awramik , m. czakon , a. onishchenko and o. veretin , phys . d * 68 * ( 2003 ) 053004 [ hep - ph/0209084 ] . m. awramik , m. czakon , a. freitas and g. weiglein , phys . * 93 * ( 2004 ) 201805 [ hep - ph/0407317 ] . m. faisst , j. h. kuhn , t. seidensticker and o. veretin , nucl . b * 665 * ( 2003 ) 649 [ hep - ph/0302275 ] . g. degrassi , p. gambino and p. p. giardino , arxiv:1411.7040 [ hep - ph ] . m. bicer _ et al . _ [ tlep design study working group collaboration ] , jhep * 1401 * ( 2014 ) 164 [ arxiv:1308.6176 [ hep - ex ] ] .	* abstract * the test of the electroweak corrections has played a major role in providing evidence for the gauge and the higgs sectors of the standard model . at the same time the consideration of the electroweak corrections has given significant indirect information on the masses of the top and the higgs boson before their discoveries and important orientation / constraints on the searches for new physics , still highly valuable in the present situation.the progression of these contributions is reviewed . a contribution to : + the standard theory up to the higgs discovery + - 60 years of cern - + l. maiani and g. rolandi , eds .
it has been known for a long time since angell @xcite has proposed a famous classification in viscosities of glass - forming materials that there exist two types of glass - forming liquids , fragile liquids ( f ) and strong liquids ( s ) , near the glass transition @xcite . the systems with short - range interactions such as @xmath11-terphenyl and glycerol are typical examples of fragile liquids , while the covalently bonded network glass formers such as sio@xmath12 and geo@xmath12 are known as typical examples of strong liquids . thus , it has been understood commonsensically since then that the transport coefficients of both liquids , such as viscosity and self - diffusion coefficient , are well described by the vogel - fulcher - tammann ( vft ) law @xcite , although the fitting temperature range for strong liquids is shorter than that for fragile liquids . however , it is not clear yet how the dynamics of strong liquids is different from that of fragile liquids in a supercooled state . thus , it is still important to clarify it not only qualitatively but also quantitatively from a unified point of view . in order to classify the long - time self - diffusion coefficient @xmath13 into two types of glass forming liquids from a unified point of view consistently , tokuyama @xcite has recently shown that the @xmath14- and @xmath15-relaxation times , @xmath16 and @xmath17 , obey power laws @xmath18 and @xmath19 in a supercooled state , where the exponent @xmath20 is given by @xmath21 for ( f ) and 2/11 for ( s ) . then , the following master curve @xmath22 for @xmath13 has been proposed : @xmath23 , \label{1 - 2}\end{aligned}\ ] ] where @xmath24 is a fictive singular temperature to be determined and @xmath25 a positive constant to be determined . here the exponent @xmath26 is given by @xmath27 ; @xmath28 for ( f ) and 5/3 for ( s ) . thus , it has been shown by analyzing many different data that both types of liquids are well described by two types of master curves up to the deviation point @xmath29 , below which all the data start to deviate from them and obey the arrhenius law , where @xmath30 . here we note that @xmath29 coincides with the so - called thermodynamic glass transition temperature @xmath31 and the master curves can be also fitted by the vft law well for @xmath32 @xcite . thus , all the diffusion data in each type collapse onto each single master curve @xmath22 ( see fig . [ d ] ) . their material differences are just characterized by a set of parameters ( @xmath24 , @xmath25 , @xmath26 ) . from this viewpoint , therefore , those parameters may correspond to the so - called degree of fragility usually discussed among different systems @xcite . one can now distinguish the long - time self - diffusion coefficient of strong liquids from that of fragile liquids safely by using the master curve @xmath22 . by using such a master curve , we have recently succeeded in creating a new type of strong liquids which is different from usual network glass formers @xcite . in fact , the static structure factor @xmath33 of usual network glass formers has the so - called first sharp diffraction peak , which is related to the size of tetrahedron in sio@xmath12 ( see fig . [ sq ] ) . on the other hand , that of a new type does not have such a peak and its structural properties are the same as those of fragile liquids . thus , there exists another type of strong liquids , that is , non - network glass formers ( s@xmath34 ) , in addition to usual network glass formers ( s@xmath35 ) . although both types can be classified as strong liquids by using the same master curve @xmath22 with @xmath36 , it is not possible to clarify how a new type is different from the usual strong one . this situation is also true for other materials such as se which has a network structure and is usually believed to be a fragile system . it is interesting to know whether se is a fragile liquid ( f@xmath35 ) or not . however , those materials are not investigated here because there is no simulation data available . in the present paper , by using the mean-@xmath0th displacement @xmath37 ) , we only investigate the dynamics of glass - forming materials ( f@xmath34 ) , ( s@xmath35 ) , and ( s@xmath34 ) from a unified point of view based on the mean - field theory @xcite , where @xmath38 is a position vector of @xmath39th atom @xmath14 at time @xmath40 , the brackets an average over the equilibrium ensemble , and @xmath0 even numbers . analyses of many data then suggest an existence of a master curve @xmath2 for @xmath1 in each type as @xmath41 where @xmath5 is the characteristic length such as an interatomic distance , @xmath7 the average thermal velocity , and @xmath39=f@xmath34 , s@xmath35 , and s@xmath34 . any data in each type are thus shown to collapse onto a single master curve @xmath2 at the same value of @xmath4 . then , we also show that even at the same value of @xmath4 the master curve @xmath2 for type @xmath39 does not coincide at all with @xmath42 for other type @xmath43 in the cage region for @xmath44 , in which each particle behaves as if it is trapped in a cage mostly formed by neighboring particles , where @xmath45 is a mean - free time before which each particle undergoes a ballistic motion . on the other hand , @xmath2 and @xmath42 @xmath46 are easily shown to coincide with each other both for a short - time region ( @xmath47 ) and for a long - time region ( @xmath48 ) at the same value of @xmath4 . in fact , for both time regions we have @xmath49 with @xmath50 where @xmath51 . thus , we emphasize that an explicit disagreement in the dynamics of each type appears only in the cage region , although the analytic form of @xmath2 is not known there yet . finally , we note that although the even number @xmath0 is taken up to 6 here for simplicity , the same results as those discussed in the present paper also hold for @xmath52 . we begin in section ii by briefly reviewing the mean - field theory recently proposed . we first discuss the mean - field equation for the mean - square displacement and its related characteristic times . then , we show two types of master curves for the long - time self - diffusion coefficient . one is a master curve for fragile liquids and another is for strong liquids . in section iii , we introduce several potentials to perform extensive molecular - dynamics simulations . in section iv , we briefly review how physical quantities satisfy the universality near the glass transition . based on such a universality , we then show that there exist a master curve @xmath2 for the mean-@xmath0th displacement in each liquid , ( f@xmath34 ) , ( s@xmath35 ) , and ( s@xmath34 ) . in section v , we show that the master curves @xmath2 and @xmath42 in different types @xmath39 and @xmath43 do not coincide with each other in the cage region even at the same value of @xmath4 . we conclude in section vi with a summary . here we briefly summarize the mean - field theory of the glass transition ( mft ) for molecular systems recently proposed by tokuyama @xcite . the mean - field theory consists of the following two essential points : ( a ) mean - field equation for @xmath53 and ( b ) two different types of singular functions for @xmath13 , the mean - field curve @xmath54 for a liquid state and the master curve @xmath55 for a supercooled state , where @xmath56 and @xmath24 are fictive singular temperatures to be determined and @xmath57 . the mean - square displacement @xmath53 of @xmath39th particle @xmath14 in molecular systems is described by a nonlinear equation @xcite @xmath58e^{-m_2(t)/\ell^2},\label{2 - 1}\ ] ] where @xmath59 is a mean - free path of particle @xmath14 over which the particle can move freely by a ballistic motion and @xmath60 the average thermal velocity . equation ( [ 2 - 1 ] ) can be easily solved to give a formal solution @xmath61 , \label{2 - 2}\end{aligned}\ ] ] where @xmath62 denotes a time for a particle to diffuse over a distance of order @xmath59 with the diffusion coefficient @xmath6 and is identical to the so - called @xmath15-relaxation time . here @xmath63 is a mean - free time , within which each particle can move freely by a ballistic motion . the solution ( [ 2 - 2 ] ) satisfies the asymptotic forms given by eq . ( [ 1 - 4 ] ) . as shown in ref . @xcite , the mean - free path @xmath59 is uniquely determined by @xmath64 . hence the solution ( [ 2 - 2 ] ) suggests that the dynamics is described by only one parameter @xmath65 if the length and the time are scaled by @xmath5 and @xmath66 , respectively . this means that the dynamics in different systems coincides with each other if @xmath65 has the same value in them . hence this is called a universality in dynamics . since the single - particle dynamics is determined by only one parameter @xmath64 , it is convenient to introduce a new parameter @xmath67 by @xcite @xmath68 as @xmath67 increases ( or @xmath69 decreases ) , there exist three states ; a liquid state [ l ] for @xmath70 ( or @xmath71 ) , a supercooled state [ s ] for @xmath72 ( or @xmath73 ) , and a glass state [ g ] for @xmath74 ( or @xmath75 ) , where @xmath76 is a supercooled point and @xmath31 a glass transition point . the values of @xmath77 are listed in table [ table-01 ] . here @xmath78 ( or @xmath76 ) is determined by the intersection point of the mean - field curve @xmath54 with the master curve @xmath79 @xcite and coincides with a peak position of a specific heat , while @xmath80 ( or @xmath31 ) is determined by a deviation point @xmath29 at which the simulation results and the experimental data for the long - time self - diffusion coefficient start to deviate from the master curve @xmath79 since @xmath29 coincides with the thermodynamic glass transition point @xcite . thus , the mean - field fitting values for the mean - free path @xmath81 and the @xmath15-relaxation time @xmath82 are uniquely determined by @xmath67 . in general , however , the length @xmath5 is not known . as a well - known example in which @xmath5 is known , one can take the lennard - jones ( lj ) binary mixtures a@xmath83b@xmath84 , where the lj potential @xmath85 is given by @xmath86.\label{2 - 3 - 1}\ ] ] here @xmath87 , @xmath88 , @xmath89 , @xmath90 , @xmath91 , and @xmath92 , where @xmath93 is a length unit and @xmath94 an energy unit @xcite . then , one can choose @xmath93 as @xmath5 for a particle . thus , one can use the simulation results for the lj binary mixtures as reference to determine @xmath5 for fragile systems based on the universality . this will be discussed later . .universal value @xmath77 . [ cols="^,^,^,^",options="header " , ] [ table-8 ] versus @xmath95 for different value of @xmath67 in fragile liquids al@xmath12o@xmath96 and non - network strong liquids a@xmath83b@xmath84 ( sw @xmath97 ) , where @xmath981.842 , 1.969 , 2.302 , 2.638 , and 3.143 . the dotted lines indicate the simulation results for al and the solid lines for a ( @xmath97 ) . the symbols indicate the relaxation times @xmath45 and @xmath17 at @xmath99 ; @xmath100 for a ( @xmath97 ) and @xmath101 for al . temperature at each @xmath67 is listed in table [ table-8].,width=302 ] we next compare the simulation results for the non - network strong liquids with those for fragile liquids at the same value of @xmath67 . in fig . [ aqal ] , the simulation results for the mean - square displacement @xmath102 are plotted versus @xmath95 for different values of @xmath67 in fragile liquids al@xmath12o@xmath96 ( bm ) and non - network strong liquids a@xmath83b@xmath84 ( sw @xmath97 ) . the dynamics of a is compared with that of al . in the @xmath15 stage for @xmath44 , the dynamical behavior of a ( @xmath97 ) is shown to be quite different from that of al . this must be caused by the fact that a particle moves through the interactions with slowly - moving b particles . the results obtained here are also seen for other combinations between al@xmath12o@xmath96 and a@xmath83b@xmath84 in @xmath1 . versus @xmath95 at @xmath983.143 for network strong liquids sio@xmath12 ( nv ) , non - network strong liquids a@xmath83b@xmath84 ( sw @xmath97 ) , and non - network fragile liquids al@xmath12o@xmath96 ( bm ) . the solid lines indicate the simulation results for si at @xmath103 ( k ) , the dotted lines for a ( @xmath97 ) at @xmath104 , and the dashed line for al at @xmath105 ( k ) . the symbols indicate the relaxation times @xmath45 and @xmath17 ; ( @xmath106 ) for si , ( @xmath107 ) for a ( @xmath97 ) , and ( @xmath108 ) for al.,width=302 ] in the @xmath15 stage , the dynamical behavior of both network strong liquids and non - network strong liquids have been shown to be quite different from that of fragile liquids at the same value of @xmath67 . hence we now compare the dynamics of network strong liquids with that of non - network strong liquids . in fig . [ m2fss ] , the simulation results for the mean - square displacement @xmath102 are plotted versus @xmath95 for network glass formers sio@xmath12 ( nv ) and non - network glass formers a@xmath83b@xmath84 ( sw @xmath97 ) . as a typical example , the value of @xmath67 is chosen in a supercooled state as @xmath109 because the difference between both glass formers is seen clearly . in the @xmath15 stage for @xmath44 , the dynamics of si is shown to be different from that of a ( @xmath97 ) , although both glass formers obey the strong master curve given by @xmath110 . this difference results from the fact that sio@xmath12 has a network structure , while a@xmath83b@xmath84 does not . the results obtained here are also seen for other combinations between sio@xmath12 and a@xmath83b@xmath84 ( sw @xmath111 ) at any temperatures . for comparison , the simulation results for al are also plotted . we note here that a particle moves faster than al around @xmath45 . this means that a particle is weakly influenced by the correlation of b particle because b particle hardly moves on a time scale of @xmath45 . in the present paper , we have proposed the systematic method to investigate how the dynamics of strong liquids is different from that of fragile liquids . as examples of glass - forming materials , we have taken al@xmath12o@xmath96 , the lj binary mixture a@xmath83b@xmath84 , the sw binary mixture a@xmath83b@xmath84 , and sio@xmath12 ( bks and nv ) . we have first applied the mean - field theory for the simulation results in those different glass - forming materials . then , we have obtained the long - time self - diffusion coefficient @xmath13 and also the mean - field values of the characteristic time @xmath17 . by using the master curve @xmath22 , the diffusion coefficients in different systems have been safely classified into two types of liquids , fragile liquids and strong liquids . thus , al@xmath12o@xmath96 , the lj binary mixture a@xmath83b@xmath84 , and the sw binary mixture a@xmath83b@xmath84 with @xmath112 were classified as fragile liquids with @xmath113 , while sio@xmath12 ( bks and nv ) and the sw binary mixture a@xmath83b@xmath84 with @xmath114 were classified as strong liquids with @xmath36 ( see fig . [ d ] ) . in order to use the universality that all the dimensionless physical quantities must coincide with each other at the same value of the universal parameter @xmath115 , we have then adjusted the unknown characteristic length @xmath5 in each system so that the dimensionless time @xmath82 coincides with that obtained by the simulations on the sw binary mixture since @xmath5 is known to be @xmath93 there . the results are shown in fig . [ ti ] and table [ table-4 ] . we have next investigated the simulation results for @xmath1 in each type , ( f ) and ( s ) , at the same value of @xmath67 . in type ( f@xmath34 ) , all the simulation results have been shown to collapse onto a single master curve @xmath116 at the same value of @xmath67 ( see figs . [ mnf ] ) . on the other hand , in type ( s ) two different master curves , @xmath117 and @xmath118 , have been shown to exist , depending on whether the static structure factor has the so - called first sharp diffraction peak or not . the master curve @xmath117 stands for network glass formers ( s@xmath35 ) , such as sio@xmath12 , while @xmath118 for non - network glass formers ( s@xmath34 ) , such as a@xmath83b@xmath84 with @xmath114 , whose static structure factor has structural properties similar to those of ( f ) . thus , all the simulation results in each type have been shown to collapse onto each master curve at the same value of @xmath67 ( see figs . [ mns ] and [ nsn ] ) . those classifications have been done based on the fact that the simulation results of type @xmath39 do not coincide with those of different type @xmath119 in the cage region even at the same value of @xmath67 ( see figs . [ mnfsal ] , [ aqal ] , and [ m2fss ] ) . in fact , the disagreement between ( f@xmath34 ) and ( s@xmath35 ) is reasonable because their static structure factors have different structural properties from each other . on the other hand , the static structure factor of ( s@xmath34 ) has structural properties similar with that of ( f@xmath34 ) . in ( s@xmath34 ) , however , mass of particle b is much larger than that of a particle . hence slow dynamics occurs anti - symmetrically between a and b since the length scale @xmath5 of b increases as @xmath120 increases , while that of a is unchanged . this might cause a strong character , which leads to a difference between ( f@xmath34 ) and ( s@xmath34 ) in the cage region . this situation would be rather similar to that seen in polymer gels @xcite . finally , from a unified point of view proposed in this paper one should also investigate the other interesting glass - forming materials such as se which has a network structure but is usually believed to be a fragile liquid ( f@xmath35 ) . this will be discussed elsewhere . authors ( m.t . and j.k . ) wish to thank prof . takashi nakamura for his hospitality and encouragement . this work was partially supported by high efficiency rare elements extraction technology area , institute of multidisciplinary research for advanced materials ( imram ) , tohoku university , japan . the simulations were performed by using the sgi altix3700bx2 in advanced fluid information research center , institute of fluid science , tohoku university . c. a. angell , j. non - cryst . solids * 102 * , 205 ( 1988 ) . c. a. angell , j. non - cryst . solids * 131 - 133 * , 13 ( 1991 ) . a. p. sokolov , e. rssler , a. kisliuk , and d. quitmann , phys . * 71 * , 2062 ( 1993 ) . r. bohmer , k. l. ngai , c. a. angell , and d. j. plazek , j. chem . phys . * 99 * , 4201 ( 1993 ) . t. a. vilgis , phys . b * 47 * , 2882 ( 1993 ) . c. a. angell , science * 5206 * , 1924 ( 1995 ) . c. a. angell , k. l. ngai , g. b. mckenna , p. f. mcmillan , and s. w. martin , j. appl . phys . * 88 * , 3113 ( 2000 ) . p. g. debenedetti and f. h. stillinger , nature * 410 * , 259 ( 2001 ) . proceedings of the 4th international discussion meeting on relaxation in fragile systems , edited by k. l. ngai , j. non - cryst . solids * 307 - 310 * ( 2002 ) . proceedings of the 3rd international symposium on slow dynamics in complex systems , edited by m. tokuyama and i. oppenheim ( aip , new york , 2004 ) , cp * 708. k. l. ngai , _ relaxation and diffusion in complex systems _ ( springer , new york , 2011 ) . proceedings of the 4th international symposium on slow dynamics in complex systems , edited by m. tokuyama and i. oppenheim ( aip , new york , 2013 ) , cp 1518. h. vogel , phys . zeit . 22 * , 645 ( 1921 ) . g. s. fulcher , j. ann . * 8 * , 339 ( 1925 ) . g. tammann and w. hesse , z. anorg . allg . chem . * 156 * , 245 ( 1926 ) . m. tokuyama , phys . e * 82 * , 041501 ( 2010 ) . m. tokuyama , j. phys . b * 115 * , 14030 ( 2011 ) . m. tokuyama , aip conf . proc . * 1518 * , 47 ( 2013 ) . t. scopigno , g. ruocco , f. sette , and g. monaco , science * 302 * , 849 ( 2003 ) . v. n. novikov and a. p. sokolov , nature * 431 * , 961 ( 2004 ) . voigtmann , j. non - cryst . solids * 352 * , 4826 ( 2006 ) . m. tokuyama and s. enda , physica a * 392 * , 2999 ( 2013 ) . m. tokuyama , physica a * 364 * , 23 ( 2006 ) . m. tokuyama , phys . e * 62 * , r5915 ( 2000 ) . m. tokuyama , physica a * 378 * , 157 ( 2007 ) . m. tokuyama , phys . e * 80 * , 031503 ( 2009 ) . w. kob and h.c . andersen , phys . . lett . * 73 * , 1376 ( 1994 ) . m. tokuyama , physica a * 388 * , 3083 ( 2009 ) . t. a. weber and f. h. stillinger , phys . b * 31 * , 1954 ( 1985 ) . v. v. hoang , phys . rev . b * 70 * , 134204 ( 2004 ) . t. narumi and m. tokuyama , rep . inst . fluid science * 19 * , 73 ( 2007 ) . t. narumi and m. tokuyama , phys . e * 84 * , 022501 ( 2011 ) . b. w. h. van beest , g. j. kramer , and r. a. van santen , phys . lett . * 64 * , 1955 ( 1990 ) . a. nakano , l. s. bi , r. k. kalia and p. vashishta , phys . b * 49 * , 9441 ( 1994 ) . j. horbach , w. kob , k. binder and c. a. angell , phys . e * 54 * , r5897 ( 1996 ) . d. sidebottom , r. bergman , l. brjesson , and l. m. torell , phys . lett . * 71 * , 2260 ( 1993 ) . j. habasaki , i. okada , and y. hiwatari , phys . e * 52 * , 2681 ( 1995 ) . m. d. ediger , c. a. angell , and s. r. nagel , j. phys . chem . * 100 * , 13200 ( 1996 ) . a. pasquarello and r. car , phys . * 80 * , 5145 ( 1998 ) . p. g. debenedetti and f. h. stillinger , nature * 410 * , 259 ( 2001 ) . e. kohira , y. terada , and m. tokuyama , rep . inst . fluid science * 19 * , 91 ( 2007 ) . m. tokuyama and y. terada , physica a * 375 * , 18 ( 2007 ) . q. mei , c. j. benmore , s. sen , r. sharma , and j. l. yarger , phys . b * 78 * , 144204 ( 2008 ) . philip s. salmon and anita zeidler , phys . , * 15 * , 15286 ( 2013 ) . r. zallen , the physics of amorphous materials , wiley , new york , 1983 . l. v. woodcock , c. a. angell , and p. cheeseman , j. phys . * 65 * , 1565 ( 1976 ) . r. a. jackson and c. r. a. catlow , mol . simul . , * 1 * , 207 ( 1988 ) . s. tsuneyuki , m. tsukada , h. aoki , and y. matsui , phys . 61 * , 869 ( 1988 ) . p. vashishta , r. k. kalia , j. p. rino , and i. ebbsj , phys . b * 41 * , 12197 ( 1990 ) . m. wilson , p. a. madden , m. hemmati , and c. a. angell , phys . lett . * 77 * , 4023 ( 1996 ) . m. hemmati and c. a. angell , in _ physics meets geology _ , edited by h. aoki and r. hemley ( cambridge university press , 1998 ) . w. kob , j. phys . : condens . matter * 11 * , r85 ( 1999 ) . j. horbach and w. kob , phys . rev . b * 60 * , 3169 ( 1999 ) . a. carre , l. berthier , j. horbach , s. ispas , and w. kob , j. chem . 127 * , 114512 ( 2007 ) . voigtmann and j. horbach , j. phys . : condens . matter * 20 * , 244117 ( 2008 ) . s. seki , md . b. h. susan , t. kaneko , h. tokuda , a. noda , m. watanabe , j. phys . b * 109 * , 3886 ( 2005 ) .	the systematic method to explore how the dynamics of strong liquids ( s ) is different from that of fragile liquids ( f ) near the glass transition is proposed from a unified point of view based on the mean - field theory discussed recently by tokuyama . the extensive molecular - dynamics simulations are performed on different glass - forming materials . the simulation results for the mean-@xmath0th displacement @xmath1 are then analyzed from the unified point of view , where @xmath0 is an even number . thus , it is first shown that in each type of liquids there exists a master curve @xmath2 as @xmath3 onto which any simulation results collapse at the same value of @xmath4 , where @xmath5 is a characteristic length such as an interatomic distance , @xmath6 a long - time self - diffusion coefficient , @xmath7 a thermal velocity , and @xmath8f and s. the master curves @xmath9 and @xmath10 are then shown not to coincide with each other in the so - called cage region even at the same value of @xmath4 . thus , it is emphasized that the dynamics of strong liquids is quite different from that of fragile liquids . a new type of strong liquids recently proposed is also tested systematically from this unified point of view . the dynamics of a new type is then shown to be different from that of well - known network glass formers in the cage region , although both liquids are classified as a strong liquid . thus , it is suggested that a smaller grouping is further needed in strong liquids , depending on whether they have a network or not .
understanding transition metal oxides is of perpetual interest and importance in condensed matter physics and material science @xcite . recent advances in layer - by - layer growth techniques of heterostructures of transition metal compounds have created particular interest due to their great scientific and technological potential @xcite . exotic material phenomena that are clearly distinctive from the ` normal ' phases include interface superconductivity @xcite , magnetism @xcite , charge @xcite and orbital reconstruction @xcite . one of the most intriguing classes of materials may be the nickelate superlattices @xcite . a series of recent theoretical studies have created considerable interest , suggesting the heterostructuring - induced orbital polarization and the possible high - t@xmath3 superconductivity in lanio@xmath0 ( lao)/laalo@xmath0 ( lao ) superlattice @xcite . in this picture , the two degenerate ni @xmath4 orbital states are split by a combination of translational symmetry breaking and on - site coulomb interaction , leading to a cuprate - like band structure . although this kind of theoretical picture has been challenged by more recent dynamical mean - field theory ( dmft ) calculations based on a charge transfer model including oxygen states explicitly @xcite , several experimental papers have found other interesting phenomena in this system . boris and co - workers @xcite reported a metal to insulator transition as a function of lno layer thickness : even though the bulk lno is a paramagnetic ( pm ) metal , the heterostructure ( lno)@xmath5/(lao)@xmath6 becomes insulating and magnetic if @xmath7 is small , @xmath8 , while it remains pm and metallic when @xmath9 @xcite . the insulating behavior was also observed by freeland _ however the detailed magnetic and electronic structure changes , as well as the other important physical quantities such as orbital polarization have not yet been clearly understood as a function of layer thickness , @xmath7 . in this study , we performed a detailed first - principles analysis for the electronic structure , magnetism , and orbital polarization of lno / lao superlattice using the lda@xmath2 method @xcite . since previous calculations were performed at the lda @xcite or dmft level @xcite and assumed the bulk - like pm phase , lda@xmath2 calculation can provide meaningful information especially regarding the magnetism in this system . total energy calculations showed that the lda+@xmath1 stabilizes ferromagnetic ( fm ) spin order along in - plane and out - of - plane direction . this result may indicate the existence of another ground state configuration in between @xmath10 and @xmath11 superlattices . the orbital polarization is significantly reduced by @xmath1 which is in contrast to the simplified hubbard - type model prediction @xcite , but consistent with the extended charge - transfer model dmft calculation @xcite . the calculated ni-@xmath12 valence based on lda@xmath2 supports the recently suggested picture for the metal / insulator phase diagram based on the @xmath12-valency @xcite , and demonstrates the importance of the double counting issue . these results are discussed in comparison to the recent theoretical and experimental studies . /(lao)@xmath13 with the in - plane spin ordering of fm and afm as a function of @xmath1 . ( b ) calculated total energy of ( lno)@xmath14/(lao)@xmath13 with the out - of - plane spin ordering of fm and afm as a function of @xmath1 where the in - plane spins are set to be fm . ( c ) calculated ni magnetic moment for ( lno)@xmath13/(lao)@xmath13 and ( lno)@xmath14/(lao)@xmath13 as a function of @xmath1 ( fm order considered ) . ( d ) average number of ni-@xmath4 electrons in ( lno)@xmath13/(lao)@xmath13 and ( lno)@xmath14/(lao)@xmath13 as a function of @xmath1 . [ evsu],width=226 ] for the band - structure calculations , we employed troullier - martins type norm - conserving pseudopotential @xcite with a partial core correction and linear combination of the localized pseudo - atomic orbitals ( lcpao ) @xcite as a basis set . in this pseudo - potential generation , the semi - core 3@xmath15 electrons for transition metal atoms were included as valence electrons in order to take into account the contribution of the semi - core states to the electronic structure . we adopted the local density approximation ( lda ) for exchange - correlation energy functional as parametrized by perdew and zunger @xcite , and used energy cutoff of 400 ry and k - grid of @xmath16 per unit superlattice volume . the lda@xmath2 functional is adapted from the formalism of ref . and ref . the geometry relaxation has been performed with the force criterion of @xmath17 hartree / bohr . during the relaxation process , the in - plane lattice constant is fixed considering the substrate effect in the experimental situation . the orbital polarization @xmath18 , defined as @xmath19 can be calculated by integrating the projected density - of - states ( dos ) up to fermi level . all the calculations were performed using the density functional theory code openmx @xcite . /(lao)@xmath6 as a function of @xmath7 ( see the text ) . [ pdia],width=302 ] one natural and evidently important question raised from experimental studies on the magnetic moment formation in this heterostructure @xcite is about the ground state spin structure and its moment size . [ evsu ] summarizes our results . [ evsu](a ) shows the calculated total energy of ( lno)@xmath13/(lao)@xmath13 superlattice as a function of @xmath1 . while at @xmath1=0 ev both fm and antiferromagnetic ( afm ) spins eventually converge to a pm solution , the magnetic ground states are stabilized at the finite @xmath1 . as clearly seen in fig . [ evsu](a ) , fm spin ordering is energetically favored within the lno plane . the energy difference between fm and afm state is 3 , 246 , and 525 mev per ( lno)@xmath13/(lao)@xmath13 for @xmath1=3 , 5 , and 7 ev , respectively . fm spin order is also favored along the out - of - plane direction as shown in fig . [ evsu](b ) in which we present total energies of ( lno)@xmath14/(lao)@xmath13 ; the out - of - plane spin orderings are set to be fm or afm while the in - plane order is fm . the calculated fm - afm energy difference is 60 , 126 , and 201 mev per ( lno)@xmath14/(lao)@xmath13 for @xmath1=3 , 5 , 7 ev , respectively . the energy differences between fm and afm spin structure becomes larger as @xmath1 increases in both cases of in - plane and out - of - plane ordering . that is , the on - site correlations stabilize fm spin ordering , and as @xmath1 decreases , the two magnetic solutions becomes more close in their energies , and eventually converges to the pm phase at @xmath1=0 . the magnetic moment is also dependent on @xmath1 . [ evsu](c ) shows the calculated magnetic moment for fm case as a function of @xmath1 . the moment increases as @xmath1 increases as in the other typical correlated transition metal oxide materials @xcite . it is noted that , as soon as @xmath1 is turned on , the ni moment becomes @xmath20 1 @xmath21 ( @xmath1=3 ev ) , and further increases to be 1.2 and 1.4 @xmath21 at @xmath1=5 and 7 ev , respectively . dos of ( lno)@xmath5/(lao)@xmath13 superlattice geometry with pm spin ( @xmath1=0 ) : ( a ) @xmath22 , ( b ) @xmath23 , ( c ) @xmath24 , and ( d ) @xmath25 . solid ( red ) and dotted ( blue ) lines correspond to @xmath26 and @xmath27 states , respectively , and fermi level is set to be 0 . [ pdos_pm_u0],width=226 ] our calculation results may seem to suggest that the magnetic moment formed in the thin - lno superlattice @xcite is ordered ferromagnetically . since the spin polarized oxide heterostructure could be useful for the device applications , there has been active research for finding the structure that produces fm spin order @xcite . therefore our result of a fm ground state in lno / lao may have a positive implication for such an application . however it should be noted that the muon spin rotation ( @xmath28sr ) experiment by boris _ @xcite is not well interpreted in the long range fm ordering picture even though the @xmath28sr is basically a local probe and that the origin of metal - insulator phase transition in the nickelate series are not clearly understood yet . especially regarding the charge disproportion or ordering in nickelates , the conventional lda@xmath2 has a clear limitation to describe such phenomena @xcite . moreover fig . [ evsu](a ) and ( b ) indicate the possibility of afm ground state in the negative @xmath1 region which is more or less related to the reported charge disorders in the nickelate systems @xcite . therefore one needs to be careful in the interpretation of our lda@xmath2 results on the fm spin ground state as an indication of long range ordered ground state as one may see in the actinide systems for example @xcite . dos of ( lno)@xmath5/(lao)@xmath13 superlattice geometry with fm spin ( @xmath1=5 ev ) : ( a ) @xmath22 , ( b ) @xmath23 , ( c ) @xmath24 , and ( d ) @xmath25 . up / down panels represent the up / down spin states , respectively . solid ( red ) and dotted ( blue ) lines correspond to @xmath26 and @xmath27 states , respectively , and the fermi level is set to be 0 . [ pdos_fm_u5],width=226 ] our results have another interesting implication regarding the phase diagram . since the thin - lno superlattice , ( lno)@xmath29/(lao)@xmath6 , can have either fm insulating ( fm - i ) ground state as predicted by lda@xmath2 calculations , or , afm insulating ( afm - i ) one which is more consistent with the @xmath28sr experiment , the system would have fm metallic ( fm - m ) or afm metallic ( afm - m ) region in between the thin - lno limit ( @xmath30 ) , and the bulk - like thick - lno limit ( @xmath31 ; pm and metallic , pm - m ) @xcite . one may also expect the pm insulating ( pm - i ) phase stabilized in the same region of @xmath7 ; the intermediate regime , @xmath32 ( see , fig . [ pdia ] ) . as lda@xmath2 method is unable to describe correlated pm solutions properly @xcite , the further pursue along this line is beyond the scope of our study . for that purpose , one may resort the dynamical mean field theory ( dmft ) calculations with charge self - consistency @xcite , and compare the total energy for pm , fm and afm configurations as we did in this study within lda@xmath2 scheme . to understand metal - insulator phase transition as a function of lno thickness @xcite , we examined ni @xmath4 dos depending on the thickness , @xmath7 . the @xmath12-band width change as a function of @xmath7 may be an useful information as @xmath33 plays an important role in the metal - insulator phase transition of rare - earth nickelate @xcite . [ pdos_pm_u0 ] shows dos of the pm case with @xmath34 . even if the bulk lno locates at the vicinity of metal - insulator phase boundary and exhibits the correlated electron behaviors , the electronic structure of pm lno has been reasonably well described within lda ( or gga ) as shown in the previous studies @xcite . [ pdos_pm_u0](a)-(d ) presents the evolution of @xmath4 dos as a function of @xmath7 . it is noted that the @xmath27 band width notably changes , whereas the @xmath26 width remains almost same across @xmath7=24 . the @xmath23 case ( fig . [ pdos_pm_u0](b ) ) , which was reported to be magnetic and insulating @xcite , the right edge of the @xmath27 state is fairly similar to that of @xmath26 whereas the left - edge is reduced . it is not certain however that such a relatively small difference in terms of the effective band width is responsible for the metal - insulator phase transition between @xmath23 and 4 as reported in experiment @xcite . a complementary picture can be provided by lda@xmath2 calculations for the magnetic phase . [ pdos_fm_u5](a)-(d ) shows the evolution of @xmath4 dos as a function of @xmath7 for the fm ( lno)@xmath13/(lao)@xmath13 . once again the notable change is found in @xmath27 band ; for @xmath22 , the up - spin @xmath27 state ( upper panel ) forms a fairly localized dos around fermi level . this state becomes more and more delocalized as @xmath7 increases as seen in fig . [ pdos_fm_u5](b)-(d ) . once again , however , the amount of the effective band width change depending on @xmath7 does not seem to be enough to make metal - insulator phase transition across @xmath7=24 . other origins than the simple @xmath33 change may be more relevant to the transition @xcite . = 0 ( pm ) , @xmath1=3 ( fm ) , and @xmath1=5 ( fm ) , respectively . [ pvsm],width=188 ] an interesting point observed in fig . [ pdos_fm_u5 ] is that even in lda@xmath2 calculation , the systems do not become a perfect insulator , but have finite number of states around fermi level @xcite . it may partly be attributed to the strong covalency between ni-@xmath12 and o-@xmath15 states . as o-@xmath15 is much less affected by @xmath1 , the small amount of dos is not removed perfectly . another important factor is the double - counting energy correction for which several functional forms have been suggested @xcite , but there is no well defined solution still yet @xcite : as the double - counting energy is typically represented by @xmath35 , the lda@xmath2 charge self - consistency adjusts @xmath36 ( or , the effective @xmath12-level energy and therefore the charge transfer energy , @xmath37 ) , depending on @xmath1 . according to the recent dmft study , which tunes the double - counting term as a parameter , the system eventually becomes insulating at large @xmath1 @xcite . therefore the small states around the fermi level obtained by lda@xmath2 can be attributed to the double - counting error that is hardly handled within the current formalism of lda@xmath2 . an interesting recent finding on ( 3-dimensional ) nickelates is that the metal - insulator transition occurs at the very narrow region of @xmath36 and the same holds for the cuprates @xcite . from the single - site dmft calculations with a double - counting energy as another tuning parameter , wang _ et al . _ @xcite showed that there is a well defined @xmath38 , that is , the critical value of @xmath12-valency of transition metal ; the system is metallic if @xmath39 and insulating if @xmath40 . for nickelates , @xmath41 . that is , for the parameters ( implicitly including double - counting correction ) , which result in @xmath39 , the system remains metallic even for the very large @xmath1 . this conclusion suggests @xmath36 as the critical variable for understanding charge - transfer systems @xcite . now it might be instructive to analyze our lda@xmath2 results within this new picture of ref . . even if lda@xmath2 is a lower level approximation compared to dmft , a big merit of lda@xmath2 is that it can be performed with the whole charge self - consistency and take magnetism into account while the previous dmft calculations have dealt with the pm phase @xcite . [ evsu](d ) presents @xmath36 for pm ( @xmath1=0 ) and fm ( @xmath1=3 , 5 , 7 ) calculations of ( lno)@xmath42/(lao)@xmath13 and ( lno)@xmath43/(lao)@xmath13 . while @xmath36 decreases as @xmath1 increases , the difference is not significant ; @xmath44 . therefore the current implementation of lda@xmath2 and the charge self - consistency roughly follow the constant @xmath36 line . now we note that the value of @xmath45 is far from the @xmath46 predicted by dmft . therefore it is consistent with the metallic ground state and supports the conclusion of ref . . relative orbital occupation in the two @xmath4 orbitals is a central quantity in understanding lno / lao systems @xcite . the orbital polarization can be non - zero due to the translation symmetry breaking induced by heterostructuring while the bulk polarization is 0 . according to the previous calculations , highly polarized @xmath18 can possibly drives the system to be high - t@xmath3 superconductor @xcite . on the other hand , a more recent dmft calculation based on the realistic model hamiltonian including oxygen orbitals predicted that the polarization is actually reduced by the on - site correlation @xcite . since both of the previous dmft calculations assumed pm phase and did not consider the full charge self - consistency , lda@xmath2 calculation can give a complimentary information for the magnetic solution in spite of its limitation in describing correlation effect compared to dmft . the result is presented in fig . [ pvsm ] . it is noted that the inclusion of @xmath1 reduces the polarization significantly . there is a large separation between @xmath1=0 and @xmath1=3 results , while the differences between @xmath1=3 and 5 results are small . importantly the calculated polarizations for finite @xmath1 are order - of - magnitude same compared with the recent dmft results @xcite . therefore it supports the conclusion of the recent dmft calculation @xcite . the small polarization at the finite @xmath1 can also be seen in dos presented in fig . [ pdos_fm_u5 ] where no big difference can be found between the two @xmath4 orbital occupations . we note that the reduced orbital polarization by @xmath1 can be compatible with the ferromagnetic spin order . for cuprates , for example , the fully polarized @xmath47 orbital is directly related to the in - plane afm spin order . in the nickelates , we have one more orbital degree of freedom available and the spin can align ferromagnetically . and actually it is found that the enhanced ferromagnetic trend by increasing @xmath1 corresponds to the reduced orbital polarization as shown in fig . [ pvsm ] . using the band structure calculations based on lda@xmath2 , we examined the magnetic moment , electronic structure , @xmath12-valence , and orbital polarization as a function of @xmath1 and @xmath7 ( lno thickness ) . the calculated results clearly showed the formation of magnetic moment at the finite @xmath1 region being consistent with a recent @xmath28sr , but the long range ordering pattern is not so clear considering the experiment . while @xmath27 band width is reduced as @xmath7 approaches to 1 , it may not be enough to be responsible for the metal - insulator transition . the calculated orbital polarization is significantly reduced by @xmath1 , strongly supporting the conclusion of a recent dmft calculations and indicating the absence of high temperature superconductivity in this system . this work was supported by the u.s . department of energy ( doe ) , de - fg02 - 03er46097 , and niu s institute for nanoscience , engineering , and technology . work at argonne national laboratory was supported by the u.s . doe , office of science , office of basic energy sciences , under contract no . de - ac02 - 06ch11357 . n. reyren , s. thiel , a. d. caviglia , l. f. kourkoutis , g. hammerl , c. richter , c. w. schneider , t. kopp , a .- s . r " uetschi , d. jaccard , m. gabay , d. a. muller , j .- triscone , j. mannhart , science * 317 * , 1196 ( 2007 ) . e. benckiser , m. w. haverkort , s. brck , e. goering , s. macke , a. fra , x. yang , o. k. andersen , g. cristiani , h .- u . habermeier , a. v. boris , i. zegkinoglou , p. wochner , h .- j . kim , v. hinkov and b. keimer , nature mater . * 10 * , 493 ( 2011 ) . a. v. boris , y. matiks , e. benckiser , a. frano , p. popovich , v. hinkov , p. wochner , m. castro - colin , e. detemple , v. k. malik , c. bernhard , t. prokscha , a. suter , z. salman , e. morenzoni , g. cristiani , h .- u . habermeier , and b. keimer , science * 332 * , 937 ( 2011 ) .	density functional band calculations have been performed to study lanio@xmath0/laalo@xmath0 superlattices . motivated by recent experiments reporting the magnetic and metal - insulator phase transition as a function of lanio@xmath0 layer thickness , we examined the electronic structure , magnetic properties , and orbital occupation depending on the number of lanio@xmath0 layers . calculations show that the magnetic phase is stabler than the nonmagnetic for finite and positive @xmath1 values . the orbital polarization is significantly reduced by @xmath1 even in the magnetic regions . the implications of the results are discussed in comparison to recent experimental and theoretical studies within the limitations of the lda@xmath2 method .
in recent times , the need for a `` quintessence '' component has come out due to the several difficulties of the standard @xmath0 cold dark matter ( cdm ) model which proved unable to explain the observed features of large scale structure . in the context of inflationary cosmologies , we expect that the present spatial curvature of the universe is negligible and the total energy density equals the critical energy density ; on the other hand , there is growing observational evidence that the matter energy density is remarkably below the critical value , even taking into account exotic and so far undetectable particles known as cold dark matter . thus , we are faced with figuring out how to explain missing - energy values of as much as 70 or 80 percent of critical density . further , there is need to have the age of the universe , @xmath1 , exceeding the age of globular clusters in our galaxy ; the limits on @xmath1 are holding at about 13 gyr or more @xcite , and when combined with current estimates of the hubble expansion parameter , converging from different methods to @xmath2km / sec / mpc @xcite , give rise to an observed value of the `` expansion - age '' parameter @xmath3 , sensibly higher than 2/3 as predicted by the standard einstein - de sitter model . preserving the flatness of the universe , its age could be enhanced by lowering the matter content in models involving a component whose equation of state is different from matter and radiation , for example in models including a cosmological constant . one more problem with the cdm model arises from the mismatching of the galaxy clustering power spectrum shape , when only cobe normalized spectra are considered . all this motivations have ratified the demise of the standard cdm model , leaving cosmologists with the open question of what the missing - energy candidate could be . apart from the cosmological constant , introduced in some models in order to maintaining the spatial flatness , but still retaining serious unsolved theoretical issues , several `` quintessence '' models where proposed as candidates for the missing energy , often modeled as scalar fields rolling down their potentials @xcite , or more generally described in terms of an unspecified equation of state different from that of matter and radiation @xcite . we refer to any component whose properties are well described in terms of a scalar field evolving in a potential which couples to ordinary matter only through gravity . in some sense , it can behave like a cosmological constant when its kinetical energy is negligible with respect to the potential energy , so that the scalar field equation of state approaches -1 ; due to the strongly relativistic nature of such component , the characteristic scale of clustering processes for a scalar field is just the horizon , @xcite giving a similarity with a cosmological constant in the undetectability of quintessence energy concentrations on scales smaller than the horizon . the interesting feature of the `` quintessence '' component is just that , contrary to the cosmological constant , it is time - varying and spatially inhomogeneous , so that it can develop fluctuations which can be relevant in the perturbation growth and can leave a characteristic signature in the cmb and in the large scale structure . even though many of these imprints have been caught in previous works , the issue of initial conditions and scalar field perturbations has often been underestimated ; in particular , we found a gap regarding the opportunity to impose isocurvature initial conditions in a several - components system including a minimally coupled scalar field . our aim is to give a complete prescription for describing adiabatic and isocurvature initial conditions if an additional component is present in the form of such scalar field ; this can be acquired by giving the set of equations relating all the fluid components needed in the two cases . in order to do that , we need the background and perturbation equations which we briefly review on section ii and section iii . the original results of our work are presented in sections iv and v , where we work in the formalism of the synchronous gauge , by generalizing the work of ma and bertschinger @xcite and finding the super - horizon - scale behaviour of perturbations at early times , starting from initial zero entropy perturbations ( adiabatic case ) or initial zero curvature perturbations ( isocurvature case ) . we express the needed gauge - invariant quantities , namely entropy and curvature perturbations , in terms of synchronous perturbations of baryons , photons , massless neutrinos , cold dark matter and a minimally - coupled scalar field , as well as metric perturbations . in section vi the results are translated in conformal newtonian gauge , and in section vii we numerically investigate the growth of entropy and curvature perturbations starting from different initial conditions , and we compare them with the corresponding behaviours in the standard cdm model . in that computation , the adopted scalar field is associated with an ultralight pseudo nambu - goldstone boson @xcite , with global spontaneous symmetry breaking scale @xmath4 gev and explicit breaking scale @xmath5 ev ; such a field should be acting at present like an effective cosmological constant and dominating the energy density of the universe . also we plot , and discuss , the pure adiabatic and pure isocurvature cmb anisotropies spectra , again making a comparison with the standard cdm . we begin by a brief review of an homogeneous friedmann robertson walker ( frw ) cosmology in which there is the additional contribution coming from a minimally coupled real scalar field @xmath6 evolving in a potential @xmath7 . we consider only models with total @xmath8 in this paper , and we work in conformal coordinates , so that the line element is @xmath9 $ ] where @xmath10 is the cosmic scale factor and @xmath11 is the conformal time . + the scalar field energy density and pressure , associated to the lagrangian describing the classical behavior of @xmath6 @xmath12\ , \ ] ] follow from expression of the stress - energy tensor @xmath13 and are given by @xmath14 where the overdot denotes the derivative with respect to conformal time @xmath11 . the above quantities evolve according to the friedmann equation , in which we separate the contributes of matter , radiation and scalar field to the total energy density , @xmath15\ , \ ] ] together with the conservation equations @xmath16 @xmath17 where @xmath18 is the conformal expansion rate of the universe , @xmath19 is the energy density contributed by radiation @xmath20 or nonrelativistic matter @xmath21 and @xmath22 . note that , from ( [ rhopphi ] ) , the second order klein - gordon equation ( [ kg ] ) is equivalent to the conservation law @xmath23 including all the modifications due to the additional scalar field component , we shall carry out a fully relativistic treatment of the perturbations of this background , based on the notation of ma and bertschinger @xcite . we work in the fourier space and we perform the parametrization of the perturbed quantities in the formalism of the synchronous gauge , in which the perturbed line element is @xmath24 $ ] . since we are interested here only on scalar - type perturbations , the metric perturbations can be parametrized as @xmath25\ , \ ] ] with @xmath26 and @xmath27 denoting the trace of @xmath28 . + note that the synchronous potentials @xmath27 and @xmath29 in @xmath30-space are related to the gauge - invariant variables @xmath31 of bardeen @xmath32 and @xmath33 of kodama and sasaky @xmath34 by the relation @xmath35\ , \ ] ] which allows us to relate @xmath27 and @xmath29 to the gauge - invariant curvature perturbation @xmath36 ( see @xcite ) @xmath37 being @xmath38 ; the above expression will be useful in the following . now , focus on the equations describing the evolution of perturbations involving the various components . as it is well known , the scalar field can mimic a cosmological constant if its kinetic energy is negligible with respect to the potential one . however , a substantial difference is that it admits perturbations around the homogeneous solution of ( [ kg ] ) ; in linear theory , they are described by small fluctuations @xmath39 and @xmath40 around the background values , driven by the equation of motion @xmath41 the density , pressure and velocity perturbations for the scalar field are described as usual by the following quantities @xmath42 @xmath43 therefore , we define the differential ratio @xmath44 that differs from the scalar field sound velocity @xmath45 by the term @xmath46 describing the entropy contribution @xcite . it is useful to describe radiation in terms of the coefficients characterizing the legendre expansion of the temperature and polarization brightness functions , @xmath47 and @xmath48 : @xmath49 @xmath50 their evolution is completely determined by the boltzmann equations ; denoting by @xmath51 the thomson scattering cross section and by @xmath52 the electron density , we have for photons : @xmath53 - a n_e { \sigma}_{t}{\delta}_{t { l}(\gamma ) } \ \ \ ( { l } \ge{3})\ \ , \\ \label{deltapl } \dot{\delta}_{p { l}(\gamma ) } & = & { k \over 2 { l } + 1 } [ { l } { \delta}_{p ( { l } -1)(\gamma ) } - ( { l } + 1 ) { \delta}_{p ( { l } + 1)(\gamma ) } ] + a n_e { \sigma}_t \left[{1 \over 2 } ( { \delta}_{t 2(\gamma ) } + { \delta}_{p 0(\gamma ) } + { \delta}_{p 2 ( \gamma ) } ) \left ( { \delta}_{l 0 } + { { \delta}_{{l } 2 } \over 5 } \right ) -{\delta}_{p { l}(\gamma ) } \right]\ , \end{aligned}\ ] ] where @xmath54 the perturbed stress - energy tensor for radiation contributes with the following non - zero quantities : @xmath55 the expansion ( [ deltat ] ) also applies for massless neutrinos ; their evolution equations in the synchronous gauge are given by the following system : @xmath56 \ , \ \ \ ( l \ge{3})\ .\end{aligned}\ ] ] pressureless cold dark matter interacts only gravitationally with other particles and in the synchronous gauge its peculiar velocity is zero ; setting @xmath57 , the evolution of cdm density perturbations is given by @xmath58 and the non - zero component of its perturbed stress - energy tensor is @xmath59 taking into account the coupling between photons and baryons by thomson scattering , @xmath60 and the perturbed stress - energy tensor for baryons contributes by @xmath61 all these ingredients are to be implemented in perturbed einstein equations @xmath62 this system of differential equations can be integrated once fixed the appropriate initial conditions , which will be the content of the next sections . in order to start the numerical integration of the evolution equations wrote down in the previous section , one has to impose appropriate initial conditions to the fluid and metric perturbations . although a general perturbation need not to be either isothermal ( entropic ) or adiabatic ( isoentropic ) , it can always be expressed as a linear superposition of adiabatic and isothermal components @xcite . also , it is useful to recall that isocurvature perturbations may be present in this kind of models @xcite . we explore both these conditions in scalar field cosmology . in particular we search for the initial values of the field perturbations @xmath63 and @xmath64 ( initial real time derivative ) that realize initial adiabaticity and isocurvature , together with appropriate initial conditions on the other perturbed quantities . for this purpose , we will focus on the initial fluctuation of the real time derivative of the scalar field perturbation , since the conformal time derivative is always zero at @xmath65 by definition ( @xmath66 ) . the same happens of course to the initial conformal time derivative of the background scalar field @xmath6 ; if in general the latter has an initial non - vanishing kinetic energy , so that @xmath67 , its conformal time velocity @xmath68 is zero since @xmath69 . in the following , we will need to use the scale factor behaviour at early times , when @xmath70 . we will often use the expansion of the scale factor in powers of the dimensionless parameter @xmath71 : @xmath72 where @xmath73 and @xmath74 are the present critical density and hubble parameter respectively , and @xmath75 is the total radiation density contribution at the present ; indeed , as it can be easily verified , being the scale factor @xmath70 at early times , we can neglect the scalar field contribution in equation ( [ friedmann ] ) , that admits the simple solution @xmath76 besides , the expansion rate behaves as @xmath77\ .\ ] ] before to go on , it is worth to remind some general results concerning the synchronous gauge behaviour of metric and density perturbations on superhorizon scales ( we refer to the work of ma and bertschinger @xcite , although they did not include the scalar field component ) . we impose initial condition at an early time , deep in the radiation era , when photons and baryons are tightly coupled and can be considered as a single coupled fluid ; due to the large thomson scattering opacity , the @xmath78 moments of the photon temperature brightness function ( [ deltatl ] ) ( in particular , the shear @xmath79 ) and the polarization brightness function ( [ deltapl ] ) are driven to zero ; similarly , to the lowest order in @xmath80 , one can ignore the @xmath81 moments of the neutrino temperature brightness function . thus , the equations ( [ deltag]),([thetag ] ) , ( [ deltanu]),([thetanu ] ) become : @xmath82 when we impose initial conditions , at @xmath83 , to get starting values for numerical integration , all the @xmath84 modes are still outside the horizon , i.e. @xmath85 ( the last equality holds in a radiation dominated universe ) . our aim is to extract the analytical time - dependence of superhorizon - sized perturbations at early times , once the initial conditions are realized : thus we find the early time form of equations ( [ t00])-([tij ] ) , ( [ system1]-[system3 ] ) , ( [ kgp ] ) and we find their solutions in successive powers of @xmath86 . to set - up the growth of perturbations , we must assume that at least a single perturbation is nonzero at initial time , in order to generate all the others . the first necessary step to impose adiabatic conditions is setting to zero the initial entropy perturbation ; ultimately , the origin of this result is that there is initially a single curvature perturbation ( generated we suppose by inflation ) and all later perturbations are inherited from it . the entropy exchange between any two fluid species @xmath10 and @xmath87 is ruled by the gauge invariant quantity @xmath88 that must be set to zero initially @xcite . the second request comes from setting to zero the first time derivative of @xmath89 ; actually , @xmath89 obeys the following differential equation : @xmath90 where @xmath91 is defined as @xmath92 @xmath93 being the gauge - invariant amplitude of the entropy perturbation of the fluid species @xmath10 . the quantity @xmath94 is the gauge - invariant difference between the gauge - dependent velocity perturbations of the species @xmath10 and @xmath87 . in order to have adiabatic initial conditions , both these terms on the right hand side of equation ( [ entropy ] ) are initially set to zero . thus , for each pair of fluid components , we impose @xmath95 in particular , applying ( [ adiab ] ) to the scalar field and another component ( that we leave unspecified and label with @xmath96 ) , will relate the initial values of @xmath97 and @xmath98 to the other energy component and metric perturbations . the first condition in ( [ adiab ] ) gives : @xmath99 posing @xmath100 and using the klein gordon equation ( [ kg],[kgp ] ) , we obtain @xmath101\ .\ ] ] combining them together we find : @xmath102 @xmath103\ .\ ] ] the above expressions specify the general adiabatic conditions for the scalar field . now , let us make a link to previous works ; in @xcite the adiabatic initial values and early time behaviours of the matter and the radiation components were found in the synchronous gauge ; these results apply here , too . indeed , as it can be easily seen from the einstein equations , the contribution of the scalar field fluctuations is negligible at early times @xmath70 with respect to the matter and radiation ones , by a factor @xmath104 and @xmath105 respectively . thus the approximations and treatment developed in @xcite is valid also here for what concerns the ordinary fluid components , i.e. photons , massless neutrinos , baryons and dark matter ; the time - dependence of the resulting superhorizon - sized perturbations @xmath106 is found by expanding the einstein equations into powers of @xmath86 and resolving the system of coupled differential equations to obtain the leading - order terms : @xmath107 @xmath108 @xmath109 @xmath110 where @xmath111 and @xmath112 is a normalization constant . using these results , it is immediate to see from ( [ deltaphiad ] , [ deltaphidotad ] ) that , imposing adiabatic initial conditions , the initial values of @xmath97 and @xmath113 must be set to zero . adiabatic conditions can be strictly verified @xmath114 at this initial time , due to the effect of the mutual coupling between total density perturbations and entropy perturbations which appear in a generic multi - component fluid . starting from initial zero values and using ( [ mbheta ] ) , @xmath97 and @xmath115 will evolve according to the ( [ kgp ] ) which can be easily integrated once dropped terms of highest order in @xmath11 ; this gives the following behaviours at early times ( @xmath116 ) : @xmath117 @xmath118 having considered the lowest terms in @xmath11 , thereby approximating the time derivative of @xmath6 with its value at the initial time @xmath119 . we have inserted these inputs into the standard cmb code and in section vii we shall expose some numerical results . now , let us turn to the second class of initial conditions . the isocurvature initial conditions are obtained by setting to zero the gauge invariant curvature perturbation . its expression is given in terms of the gauge invariant perturbation potential @xmath120 @xcite : @xmath121 we point out that here , as in @xcite , @xmath120 indicates the gauge invariant @xmath122 of the original work of bardeen @xcite , while in @xcite the same quantity it is indicated as @xmath123 . its expression in terms of the metric perturbations @xmath27 and @xmath29 in the synchronous gauge is : @xmath124\ .\ ] ] therefore , the appropriate isocurvature initial conditions are realized by the time growing solutions of the system ( [ t00])-([tij ] ) in which @xmath120 and @xmath125 are zero initially . first , let us see that , if the variables describing all the perturbations are regular enough to be derivable at least four times in @xmath126 , then the isocurvature initial conditions are simply imposed by setting the metric and radiation perturbations to zero initially . @xmath127 this can be easily seen by using essentially the einstein equation ( [ tij ] ) ; multiplying both the members by @xmath105 , deriving once and factoring out the present critical density @xmath73 , it takes the following form : @xmath128 since by hypothesis @xmath27 and @xmath29 are derivable four times in @xmath126 , @xmath129 admits the following early time expansion @xmath130 since with the initial condition ( [ iso ] ) its initial value is zero . at @xmath126 the only term that survives in ( [ a4e4 ] ) is @xmath131 since @xmath132 . then , by using equations ( [ system1],[system2],[system3 ] ) one obtains : @xmath133 in the same way , by deriving again ( [ a4e4 ] ) one gains @xmath134 instead , it may be easily see that @xmath135 can be different from zero ; in fact , deriving ( [ a4e4 ] ) three times would bring @xmath136 that may be different from zero by hypothesis ( [ iso ] ) . this means that , for @xmath137 , @xmath138 ; since @xmath139 to the lowest order , this is evidently enough to make @xmath140 , showing that the initial conditions ( [ iso ] ) imply isocurvature . it is evident that the initial condition ( [ iso ] ) can be realized in several ways , depending on which matter component is initially perturbed , or , in other words , on which @xmath141 is initially different from zero . in the present case a further degree of freedom arises from the presence of the scalar field , and we will analyze separately two main situations : in the first case only one matter component ( cdm or baryons ) is initially perturbed ; in the second case the initially perturbed component is only the scalar field . let us consider first the case in which an initial density perturbation , with amplitude @xmath142 , resides only on the cdm component . by integrating ( [ deltacdot ] ) , one finds @xmath143 by hypothesis , this is the only initially perturbed quantity . all the others must to be set to zero at @xmath126 . let us search the early time behaviour of the perturbations . since all the modes are outside the horizon at early times , we first neglect all the terms proportional to @xmath30 in the einstein equations ; then we expand all the quantities in powers of @xmath71 defined in ( [ epsilon ] ) and we calculate the leading orders . in making this , we are assuming that all the perturbation quantities admit a taylor expansion in @xmath126 of course . by making use of the above criteria and of equations ( [ aepsilon],[hepsilon ] ) , the einstein equation ( [ t00 ] ) becomes @xmath144 @xmath145 and it is immediate to gain the early time behaviour of @xmath27 : @xmath146 from the arguments exposed at the beginning of this section , up to the order @xmath147 we have also @xmath148 let us come now to the fluid perturbation quantities . as it is evident from the fluid equations , the @xmath149 and @xmath150 quantities are of higher order in @xmath86 with respect to the purely metric perturbations @xmath27 and @xmath29 . therefore , their early time behaviour can be written as follows : @xmath151 @xmath152 the behaviour of @xmath129 is interesting even if of high order in @xmath11 since it is directly related to the gauge invariant curvature by ( [ curv],[psisync ] ) and it can be obtained by solving equation ( [ tij ] ) : @xmath153 where we have defined @xmath154 note that @xmath155 , according to the isocurvature nature of the present case , as we exposed in the beginning of this section . also , ( [ hp6eta ] ) can be used to find the behaviour of @xmath156 , by using again ( [ tij ] ) . it remains to find the early time behaviour of the scalar field perturbation @xmath39 . this can be done by expanding @xmath39 in powers of @xmath71 and looking at the perturbed klein gordon equation once the terms proportional to @xmath157 have been neglected . the inhomogeneous term is @xmath158 ; @xmath159 is of the order zero from ( [ hearly ] ) , and @xmath69 is at least of the order @xmath71 ; thereby , to the lowest order in @xmath71 , equation ( [ kgp ] ) is satisfied by @xmath160 this completes the early time behaviour of all the perturbation quantities in this case of isocurvature initial conditions . all these relations can be easily generalized to the case in which the initial perturbed matter component is the baryonic one . in the next subsection we study the other interesting case : the initial perturbed fluid component is the scalar field itself . let us suppose that at the initial time @xmath161 the only non - zero perturbed quantity is the scalar field , in a manner such that the total gauge - invariant energy density contrast is zero , all the others perturbations being zero ; this means that at least one of the two quantities @xmath162 must be different from zero initially ; from ( [ deltaphithetaphi ] ) , the corresponding expressions for @xmath163 and @xmath164 are @xmath165\ , \ \delta\phi_{t0}= { \phi_{t0}\theta_{\phi 0}\over k}\ .\ ] ] in order to have isocurvature , for the other quantities we impose again the initial conditions ( [ iso ] ) . the relevant difference with respect to the previous situation lies in the slower rise of the metric and fluid perturbations starting from their initial zero values : they will grow according to ( [ t00])-([tij ] ) and ( [ system1]-[system3 ] ) , the whole perturbation - growth - machinery being initially driven only by the @xmath166 contribution of the scalar field through the perturbed einstein equations , while the perturbed klein gordon equation starts its dynamics from the conditions @xmath167 and generates the inhomogeneous term driving the evolution of @xmath27 . from equation ( [ t00 ] ) , together with ( [ system1]-[system3 ] ) , it is easy to find the early - time behaviour of the metric perturbation @xmath27 : @xmath168 using the method applied in the previous sections , one finds the leading - order behaviours : @xmath169 @xmath170 and , from ( [ tij ] ) it can be seen that @xmath171 where @xmath172 from the above formulas we see that the perturbations regarding the metric and the ordinary fluid components rise very slowly ; indeed we found a substantial failure of this model in providing a significant amount of perturbations . for this reason we will not consider this case in the numerical integrations of section vii . it is interesting to find the behaviour of scalar field perturbation at early times , that moves it away from its initial value @xmath163 ; this contains corrections in @xmath173 together with a term proportional to @xmath174 , as can be verified by integration of ( [ kgp ] ) : @xmath175 where the expansion coefficients are given by @xmath176 @xmath177 @xmath178\delta\phi^{(3)}\ , \ ] ] @xmath179\delta\phi^{(4)}-{\cal c}^{2}v''\delta\phi^{(2)}- { 3\over 2}\left({\delta_{\phi 0}\rho_{\phi}\over \rho_{c}\omega_r } \right ) { \cal{c}}^{5}\phi_{t0}\ .\ ] ] we considered the expansion up to the sixth order in @xmath11 because , as we will see in the next section , going to the newtonian gauge changes the last coefficient . in the next section we extend the results of section iv and v to the conformal newtonian gauge . as it is well known , the synchronous gauge is a coordinate system corresponding to observers at rest with respect to the collisionless matter component . these `` lagrangian coordinates '' are defined by the rest frame of a set of preferred observers . more physical intuition can be achieved in the conformal newtonian gauge , where the metric tensor is diagonal . inside the horizon , the perturbation equations reduce to the standard non - relativistic newtonian equations . in this section we write the results of section iv and v in the newtonian gauge . the connection between two gauges is realized in general by performing a coordinate transformation relating the two frames . the link between the perturbations in the two gauges is expressed in the same coordinate point instead of the same spacetime point ; this is why in most cases it is interesting to know the difference of the fluctuations in the two gauges @xcite . first we write down the relations between the genuine metric perturbed quantities . in the newtonian gauge the perturbation to @xmath180 exists and it is represented by the potential @xmath120 ; the trace of @xmath181 is instead perturbed by @xmath123 . their relation with @xmath27 and @xmath29 are @xmath182\ , \ ] ] @xmath183 they can be easily expressed for @xmath184,@xmath185 by substituting directly the expressions for @xmath27 and @xmath29 contained in sections iv and v. now we concentrate on the transformations regarding fluids and scalar field . they are the contained in the stress energy tensor , that transforms as @xmath186 using this and taking care to compare the perturbations in the same coordinate point , the relations between the quantities in the two gauges ( synchronous and newtonian labeled as @xmath187 and @xmath188 respectively ) for each fluid are : @xmath189 where @xmath190 is the lapse between the synchronous and the newtonian time coordinate . regarding the scalar field , we compute the newtonian gauge expression of the amplitude fluctuation @xmath39 by using the transformation @xmath191 in the following subsections we write the behaviour of the fluid quantities in the @xmath184,@xmath185 regime and in the newtonian gauge , dropping the @xmath188 subscript . the leading orders for matter and radiation perturbations are @xmath192 @xmath193 @xmath194 the scalar field perturbation amplitude is @xmath195 note that in this case the scalar field perturbations grow in time faster ( @xmath196 ) than in the synchronous gauge ( @xmath197 ) ; as we point out below , this is not a feature of the isocurvature initial conditions . matter and radiation behave as @xmath198 @xmath199 @xmath200 @xmath201 the scalar field amplitude is given by @xmath202 and shows that the gauge change does not touch the order of the leading power in @xmath11 , although it modifies its numerical coefficient . matter and radiation behave as @xmath203 @xmath204 @xmath205 the scalar field amplitude is given by @xmath206 therefore , in this isocurvature case , the behaviour of @xmath39 is affected by the gauge change only at high orders in @xmath11 , leaving the leading terms unperturbed . in the next section we will numerically solve the linear cosmological perturbation equations with the initial conditions sets developed in the iv and v sections . we performed the numerical integration applying our considerations to a scalar field model based on ultra - light pseudo - nambu - goldstone bosons ; the potential associated to this field has the form @xcite : @xmath207\ ] ] our working point corresponds to the parameter choice @xmath208 gev and @xmath209 ev ; assuming an initial kinetic energy equal to the potential one , the starting values of the scalar field and its initial time derivative are obtained by requiring that the present contribution is @xmath210 , fixing @xmath211 km / sec / mpc and @xmath212 . furthermore , we have taken the primordial power spectrum to be exactly scale - invariant . even though the main cosmological consequences of this kind of scalar fields have been analyzed by many authors ( see @xcite ) , here we use the formulas developed in the previous sections to accurately compare the pure adiabatic and pure isocurvature regimes . also we give particular emphasis on the behaviour of entropy and curvature perturbations , again comparing their evolution starting from isocurvature ( from cdm ) and adiabatic initial conditions ; each case is compared with a pure cdm model with the same background parameters . first , let us consider the power spectra of the microwave background anisotropies , both temperature and polarization . they are expressed by the expansion coefficients of the two point correlation function into legendre polynomials ( see e.g. @xcite ) and admit the following expression in terms of the quantities defined in the previous sections : @xmath213 the adiabatic case is shown in figure 1 . the presence of the scalar field ( solid line ) produces an increase of the power of the acoustic oscillations with respect to the cdm model ( dashed line ) ; this is due to the fact that the universe is not completely matter dominated at decoupling : thus at this time the perturbations are growing faster than in the cdm models ( we recall that density perturbations in adiabatic models grow as @xmath214 and @xmath10 respectively in the radiation in the matter eras ) and this produces an early integrated sachs - wolfe effect ( isw ) , found first in @xcite . also , the position of the first peak is slightly shifted toward smaller angular scales due to the increase of the distance of the last scattering surface ( projection effect ) . note how these features regard both the polarization and temperature peaks . finally , the temperature spectra shows that the isw is active on the smallest multipoles due to the dynamics of the scalar field in the present case ; this is a distinctive feature with respect to the cosmological constant @xcite . figure 2 shows the spectrum from isocurvature perturbations . while the projection effect is the same as in the adiabatic case , now the situation regarding the amplitude of the acoustic oscillations is inverted : the peaks are lower than the ordinary models , both for polarization and temperature this is simply due to the reduction of the matter / radiation ratio as we include @xmath6 by keeping @xmath215 ; in fact , the scalar field has _ no _ intrinsic dynamical effect at last scattering since matter and radiation components were largely dominant : it is well known the opposite behavior of the anisotropies in adiabatic and isocurvature models as one varies @xmath216 ( see e.g. @xcite ) . to better see this point , we plot in figure 3 and 4 the power spectra for models having fixed @xmath217 and @xmath218 but varying @xmath219 and @xmath27 by mean of different @xmath220 . thus we expect the same amount of perturbations in the cmb except for the effects that are genuinely linked to the scalar field , as the projection effect and the isw on the smallest multipoles . this is precisely what happens for the spectra in figure 3 and 4 . the dashed lines represent again the curves for @xmath221 as in figures 1 and 2 ; the solid and thin lines represents @xmath222 and @xmath223 respectively . again , the spectra show remarkably the same features for polarization and temperature , even if it should be noted how the former , arising from acoustic oscillations occurring just _ at _ decoupling , is not influenced by the isw effect , since it comes from the line of sight integration . all this shows how the perturbation power at decoupling is not touched by the subdominant scalar field ; the opposite behaviours in the adiabatic and isocurvature cases is explained by the decrease of matter in favor of scalar field . more insight on the perturbation behaviours may be obtained by looking at the time evolution of some significant quantities ; we look at one single scale , or wavenumber , roughly entering in the horizon between matter and radiation equality and decoupling : @xmath224 let us begin with the gravitational potential @xmath120 , in figure 5 ; the oscillatory dynamics is associated , both in the isocurvature and adiabatic case , to the horizon crossing of the scale examined . the amplitude of the oscillations in the scalar field models are higher for the adiabatic case and lower for the isocurvature one when compared with the corresponding cdm models . these oscillations are the source of the cmb anisotropies on sub - horizon angular scales ( @xmath225 ) through the acoustic driving effect and the early isw effect @xcite , and therefore follow the different behaviour in the two cases . we concentrate now on two particularly significant quantities regarding both adiabatic and isocurvature regimes , the total entropy perturbation , defined below , and the curvature @xmath36 defined in ( [ curv ] ) ; we recall that these quantities are gauge invariant . the amplitude of the total entropy perturbation is given by @xmath226 where @xmath227 is the total pressure and the sound speed must take into account the scalar field contribution @xmath228 in the summation : @xmath229 where @xmath230 while the @xmath231 term comes from the intrinsic entropy perturbation of each component , and it is ultimately sourced only by the scalar field component due to the spatial and temporal variations of the scalar field equation of state , the @xmath232 term arises from the different dynamical behaviours of the components , and it is related to the @xmath89 quantities defined in eq . ( [ sab ] ) by the relation @xcite : @xmath233 in fig.6 and 7 we plot @xmath234 vs @xmath10 ( solid line and arbitrary units ) using isocurvature cdm and adiabatic initial conditions respectively . we note that in both cases @xmath235 as @xmath161 : in the last case , this is an obvious consequence of the definition of what adiabatic conditions are ; on the other hand , taking isocurvature cdm initial conditions , we started with non - zero values of the @xmath89 relative to the cdm component and the other non - compensating components , but again @xmath235 as @xmath161 . this is because at early times the cosmic fluid is radiation dominated , so that @xmath236 in ( [ gamma ] ) ; this kills @xmath234 for @xmath237 , since no radiation perturbations are present initially . instead , the initial value of the first time derivative of @xmath234 is different from zero only selecting isocurvature initial conditions , since it takes contributions directly from the @xmath89 terms @xcite . the behaviours of entropy perturbations in both cases have been compared with those in the standard einstein - de sitter model @xmath0 ( dotted line ) , with the same choice of @xmath238 and @xmath239 . the entropy perturbations remain nearly constant before horizon crossing ; at this time the perturbation starts its oscillations that are damped in amplitude when the scale is well below the effective horizon . as an expected feature , note that in the scalar - field model the peaks of the oscillations are shifted closer to the present when compared to the einstein - de sitter case , as the epoch of matter - radiation equality . in fig.8 and 9 we plot the evolution of the gauge - invariant curvature perturbation @xmath36 for isocurvature cdm and adiabatic initial conditions , respectively . at @xmath70 , this quantity is zero in the isocurvature case and non - zero in the adiabatic one , being an explicit indicator of the nature of the perturbations . again the significant dynamics occurs in correspondence of the horizon crossing , and the latter occurs slightly later then in the cdm model due to the presence of the scalar field . finally , note how in all the isocurvature cases ( figures 6 and 8) the amplitude of the oscillations are lower than in the corresponding cdm models ; as we mentioned above , this is due to the reduction of the matter component in favor of the scalar field energy density . most important , these graphs show that this is the only possible cause of this effect , since at the times of oscillations for the scale examined , roughly between equivalence and decoupling , the scalar field is very subdominant with respect to the other components . the hypothesis of a cosmic vacuum energy stored in the potential of a scalar field enlarges naturally the possibility to gain insight into high energy physics from the traces left in the cosmic radiation and in the matter distribution . due to the upcoming cmb experiments @xcite , it will be interesting to further study the cosmological imprints of these models , in the context of different theories attempting to describe the hidden sector of high energy physics . we are grateful to luigi danese and sabino matarrese for their hints and encouragement . we warmly thank marco bruni , scott dodelson , andrew liddle , ed copeland and elena pierpaoli for quick and accurate comments on our questions . also , we are grateful to the anonymous referee for the precious comments and suggestions . c.b . wishes to thank the sissa / isas institute for the kind hospitality . jimenez r. , 1998 , to appear in iop proc.series , astro - ph/9810311 paturel g. , lanoix p. , teerikorpi p. , theureau g. , bottinelli l. , gouguenheim l. , renaud n. & witasse o. 1998 to appear in astron . & astrophys . , astro - ph/9810155 viana p.t.p . & liddle a.r . 1998 , phys.rev.d 57 , 674 , astro - ph/9708247 ferreira p.g . & joyce m. 1998 , phys.rev.d 58 023503 , astro - ph/9711102 copeland e.j . , liddle a.r . & wands d. 1998 , phys.rev.d 57 , 4686 - 4690 , gr - qc/9711068 coble k. , dodelson s. & friemann j. 1997 , phys . d55 n.4 , 1851 , astro - ph/9608122 anderson g.w . & carroll s.m . , 1997 , preprint astro - ph/9711288 zlatev i. , wang l. , steinhardt p.j . , 1998 , preprint astro - ph/9807002 huey g. , wang l. , caldwell r.r . , steinhardt p.j . , preprint astro - ph/9804285 b. ratra , a. quillen , 1992 mnras 259 , 738 ratra b. & peebles 1988 , phys.rev.d 37 , 3406 hu w. 1998 , ap.j . in press , astro - ph/9801234 caldwell r.r . , dave r. & steinhardt p.j . 1998 , phys.rev.lett.80 1582 - 1585 , astro - ph/9708069 turner m. & white m. , 1997 phys . d56 , 4439 bardeen j.m . 1980 , phys.rev.d 22 , 1882 mollerach s. 1990 , phys.lett.242b , 158 microwave anisotropy probe home page : http://map.gsfc.nasa.gov ; planck surveyor home page : + http://astro.estec.esa.nl/sa-general/projects/planck hu w. , sugiyama n. & silk j. 1997 , nature 386 , 37 ; see also hu w. , springer - verlag 1996 p. 207 , astro - ph/9511130 kodama i. & sasaki m. 1984 progr . of theor.phys.supp . 78 , 1 hu w. , seljak u. , white m. & zaldarriaga m. 1997 phys.rev.d 56 , 596 , astro - ph/9709066 mukhanov v.f . , feldman h.a . & branderberger r.h . 1992 phys.rep . 215 , 203 zaldarriaga m. & seljak u. 1997 phys . d 55 , 1830 padmanabahn t. , `` structure formation in the universe '' , cambridge university press , 1993 ma c.p . & bertschinger e. 1995 ap.j . 455 , 7 c. hill and g. ross , nucl . b 311 , 253 ( 1988 ) ; phys . lett . b 203 , 125 ( 1988 ) ; j. friemann , c. t. hill , a. stebbins and i. waga , phys . rev . 75 , 2077 ( 1995 ) ; j. estrada vigil , l. masperi , preprint hep - ph/9710522	we consider the problem of the initial conditions and behaviour of the perturbations in scalar field cosmology with general potential . we use the general definition of adiabatic and isocurvature conditions to set the appropriate initial values for the perturbation in the scalar field and in the ordinary matter and radiation components . in both the cases of initial adiabaticity and isocurvature , we solve the einstein and fluid equation at early times and on superhorizon scales to find the initial behaviour of the relevant quantities . in particular , in the isocurvature case , we consider models in which the initial perturbation arises from the matter as well as from the scalar field itself , provided that the initial value of the gauge invariant curvature is zero . we extend the standard code to include all these cases , and we show some results concerning the power spectrum of the cosmic microwave background temperature and polarization anisotropies . in particular , it turns out that the acoustic peaks follow opposite behaviours in the adiabatic and isocurvature regimes : in the first case their amplitude is higher than in the corresponding pure cold dark matter model , while they make the opposite thing for pure isocurvature initial perturbations . 8.8 in 6.5 in -.25 in -.25 in
in the previous papers @xcite , here part 1 and part 2 , are developed the methods for describing of the electrostatic screening in single- and two - component systems ( gas of electrons on a positive charged background , some electron - ion and dusty plasmas etc . ) . these methods generate a group of the new screening parameters which characterize the considered systems . in this paper the physical sense of these parameters is discussed , and the reinterpretation of some known screening parameters ( landau radius , non - ideality parameters etc . ) on the base of results from part 1 and part 2 in this paper are obtained two new screening parameters which provide the possibility to form a new hierarchy system of the characteristic screening lengths and to compare the obtained results with the existing experimental data . the material presented in this paper is distributed in four sections and one appendix . in section [ sec : rem ] are given the some quantities and relations from part 1 and part 2 which should make easier reading of this paper ; in sections [ sec : nip1 ] and [ sec : nip2 ] are considered `` small '' characteristic lengths and connected with them non - ideality parameters for single- and two - component systems . besides , in the same sections are introduced `` medium '' and `` large '' characteristic lengths for the considered systems . the obtained results are compared with the existing experimental data in section [ sec : nip2 ] . the conclusions of this paper are given in section [ sec : conc ] . finally , one important example of the application of obtained results , related to the systems with more than two components , is considered in appendix [ sec : scon ] . the density , temperature and the charge of free particles in the initial ( homogeneous ) system : @xmath0 , @xmath1 and @xmath2 , where @xmath3 , and @xmath4 - the modulus of the electron charge ; the screening constant : @xmath5^{1 \over 2},\ ] ] where @xmath6 , @xmath7 is the chemical potential , and @xmath8- the corresponding characteristic length ; the charge density in the accessory ( non - homogeneous ) system with the probe particle : @xmath9 where @xmath10 is the background charge density , and @xmath11 - the free particle density at the distance @xmath12 from the origin of the chosen reference frame ; in the region @xmath13 we have that @xmath14 the radius @xmath15 and the parameters @xmath16 and @xmath17 : @xmath18 @xmath19/x , \qquad \gamma _ { \kappa } ( x ) = ( 1 + x^3)^{1 \over 3 } - 1,\ ] ] @xmath20 where @xmath21 is wigner - seitz s radius ( the radius of the sphere with the volume @xmath22 ) ; the condition of electro - neutrality of the accessory system : @xmath23 where @xmath2 is the charge of the fixed probe particle . the density , temperature and the charge of ions in the initial ( homogeneous ) system : @xmath24 , @xmath25 and @xmath26 , where @xmath27 ; the density , temperature and the charge of electrons in the same system : @xmath28 , @xmath29 and @xmath30 or @xmath31 , where @xmath32 ; the ion and electron screening constants and the characteristic lengths : @xmath33^{1 \over 2},\ ] ] where @xmath34 , @xmath35 and @xmath36 are the ion and electron chemical potentials , @xmath37 is the electron - ion correlation coefficient given by expression @xmath38 and @xmath39 ; the charge densities in the accessory ( non - homogeneous ) systems with the probe particles : @xmath40 where @xmath41 and @xmath42 are the ion and electron densities at the distance @xmath12 from the origin of the chosen reference frame ; upper indexes @xmath43 pointed the considered case : @xmath44 - the probe particle charge is equal to @xmath26 , @xmath45 - the probe particle charge is equal to @xmath30 ; in the region @xmath46 we have that @xmath47 where @xmath48 . the radii @xmath49 in the cases @xmath44 and @xmath45 and the parameters @xmath50 and @xmath51 : @xmath52 @xmath53 where @xmath54 are wigner - seitz s radii ( the radii of the spheres with the volumes @xmath55 and @xmath56 ) ; the electro - neutrality conditions of electro - neutrality of the accessory systems : @xmath57 where @xmath58 is the charge of the corresponding probe particle . * the connection of @xmath59 with landau radius @xmath61 . * as it was announced in part 1 , all characteristic lengths of the developed method appear already in the case of single - component systems . the first of them is `` small '' characteristic length @xmath59 , given by eqs . ( [ eq261])-([eq271 ] ) , which is interpreted as a radius of the sphere centered on the probe particle and classically forbidden for other free particles . because of that it was useful to compare the radius @xmath59 with a known characteristic length which has similar physical sense . here , we keep in mind landau s radius @xmath62 which is used in the case of the classical systems ( see e.g. @xcite ) . in accordance with eqs . ( [ eq261])-([eq271 ] ) we have that @xmath63 where in the classical case ( @xmath64 ) the relation @xmath65 is valid . so , in this case the radii @xmath59 and @xmath61 are connected by relation @xmath66 it means that @xmath67 represents an approximation od the characteristic length @xmath59 which is applicable in the region of small @xmath68 . consequently , _ the parameter @xmath59 can be treated as the generalization of landau s characteristic length _ @xmath61 _ which introduces into consideration from physical reasons_. let us emphasize that , contrary to @xmath61 which is principally unlimited , the radius @xmath69 for any @xmath70 , in accordance with the conditions in real physical systems require , and that wigner - seitz s radius @xmath71 . * the connection of @xmath60 with non - ideality parameters @xmath72 and @xmath73 . * from eqs . ( [ eq261])-([eq271 ] ) , ( [ eq33 ] ) and ( [ eq34a ] ) follow the expressions @xmath74 for the coefficients @xmath75 and @xmath76 . in connection with these expressions in the classical case one should keep in mind the relation ( [ eq34a ] ) and the fact that @xmath77 , where @xmath78 is the corresponding debye s radius . with respect to this it is useful to compare eqs . ( [ eq36 ] ) with the expressions for known quantities @xmath79 which usually use as the non - ideality parameters for the classical systems . usually the parameters @xmath72 and @xmath73 , similarly to landau s radius @xmath61 , introduce from some physical reasons ( see for example @xcite ) . from eqs . ( [ eq261])-([eq271 ] ) , ( [ eq36 ] ) and ( [ eq37 ] ) it follows that in the classical case @xmath80 where @xmath81 . it means that the parameters @xmath72 and @xmath73 appear here as approximative values of the coefficients @xmath82 and @xmath83 in the classical case in the region @xmath84 . consequently , _ the parameters _ @xmath82 _ and _ @xmath83 _ can be treated as the generalized non - ideality parameters of single - component systems with any _ @xmath0 , @xmath1 _ and _ @xmath85 _ which make possible their non - relativistic treatment_. from eqs . ( [ eq251 ] ) and ( [ eq271 ] ) it follows that the coefficients @xmath82 and @xmath86 represent the functions of the parameter @xmath87 which is closely connected with another known classical quantity . namely , in the classical case ( @xmath77 ) we have that @xmath88 , where @xmath89 is so called debye s number , i.e. the mean number of the free particles in the sphere with the radius @xmath78 . the behavior of the coefficients @xmath82 and @xmath90 and the parameters @xmath72 and @xmath73 as functions of @xmath91 is illustrated in fig . [ fig : gamma ] . this figure shows that in the region @xmath92 the values of parameters @xmath72 and @xmath73 are practically same as the values of @xmath93 and @xmath86 and . they values left very closed up to the value @xmath94 . * the charges @xmath97 . * the electro - neutrality condition eq . ( [ eq42 ] ) can be presented in the form : @xmath98 , where @xmath99 , and the quantities @xmath97 are given by expressions @xmath100 with the charge density @xmath101 defined by eq . ( [ eq21 ] ) . one can see that @xmath102 and @xmath103 represent the total charge inside the sphere with radius @xmath12 , centered in the point @xmath104 , and the total charge of the rest of the space , respectively . probably , the quantities @xmath97 could have different applications . so , the fact that in the considered case the total average charge of the probe particle self - sphere , i.e. @xmath105 , is strictly equal to the average charge of the free particles inside this sphere , can be of interest . however , in the further consideration the quantity @xmath102 will play only an accessory role . namely , it will be used only the fact that @xmath102 , obtained by means eqs . ( [ eq21 ] ) and ( [ eq16 ] ) , has the physical sense in the whole space ( including the region @xmath106 ) , and can characterize the degree of the probe particle charge screening . * the connection of @xmath95 with the radius @xmath96 . * in the case of single - component systems the screening constant @xmath107 , defined by eq . ( [ eq181 ] ) , represents one of the main parameters in both methods - the method developed in this work and debye - hckel s ( dh ) method . consequently , the characteristic lengths @xmath108 has to be treated in the similar way . within dh method it is usual to interpret @xmath96 , which is equal to debye s radius @xmath78 in the classical case , as the screening radius . that is based on the fact that @xmath96 is a distance from the probe particle where the dh electrostatic potential ( described in part 1 ) becomes less than coulomb potential of this particle by the factor @xmath109 . here we will consider the role of radius @xmath96 from other aspect in order to clarify its real physical sense in both methods and determine the region of its applicability . for this purpose , we will introduce the radial charge density @xmath110 , where @xmath111 is given by eq . ( [ eq21 ] ) , and the characteristic length @xmath95 defined by relation @xmath112 the behavior of @xmath113 for several values of @xmath81 is shown in fig . [ fig : p ] . this figure illustrates the fact that for @xmath114 the point @xmath115 there is in the region @xmath116 , while for @xmath117 we have that @xmath118 . since in the region @xmath116 the charge density @xmath119 , the parameter @xmath95 for @xmath114 represents the root of equation @xmath120 where @xmath121 . consequently , in this region the parameter @xmath95 is equal to @xmath96 . it means that within the method developed in this work we have the relations @xmath122 which determine the real physical sense of the radius @xmath96 . namely , from eq . ( [ eq241 ] ) it follows that @xmath96 has the physical sense only in the region @xmath114 where @xmath123 , while in the region @xmath124 the length @xmath96 loses any connection with the charge distribution in the probe particle neighborhood and , consequently , loses any physical sense . _ let us emphasize that on the base of above mentioned one can conclude that it is always @xmath125_. then , we will consider the problem of screening of the probe particle as a problem of compensation of its charge @xmath2 within the sphere of a radius @xmath12 centered at the probe particle . for that purpose we will draw attention to the fact that uncompensated part of this charge in the case of such a sphere is equal to the charge @xmath126 . keeping this in mind , we compared the charge @xmath127 with the charge @xmath2 in the region @xmath128 . using the relations ( [ eq16 ] ) and ( [ eq21 ] ) it could be shown that in this whole region @xmath129 . from here it follows that _ the radius _ @xmath96 , _ as well as the radius _ @xmath95 , _ can not be treated as a characteristic length of full screening ( neutrality ) of the probe particle charge_. because of that , such a characteristic length is determined here in another way . * the charge @xmath102 and the quantity @xmath131 . * in accordance with eq . ( [ eq21 ] ) and the condition eq . ( [ eq41 ] ) , the charge @xmath102 can be presented in the form @xmath132 where the quantity @xmath133 is given by the expression @xmath134 \cdot 4\pi r'^{2 } } dr',\ ] ] and @xmath11 is the free particle density in the considered system with the probe particle . although the expressions ( [ eqnun1 ] ) for @xmath133 can be determined for any @xmath135 , this quantity has a special physical meaning in the region @xmath136 . namely , it can be shown that in this region @xmath133 represents the mean number of particles whose surplus into the mentioned sphere of radius @xmath12 and corresponding deficit in the rest of the space causes the deviations of @xmath102 and @xmath103 from zero . because of that only the quantities @xmath102 and @xmath133 in the region @xmath137 will be needed in this paper . from this reason in further considerations we will use the expression for @xmath133 which is applicable in the region @xmath138 , since it is always @xmath69 . in accordance with eq . ( 20 ) for @xmath11 from part 1 we have that @xmath139 where @xmath81 , and @xmath140 is defined by eq ( 30 ) and illustrated by fig . 3 from part 1 . similarly to @xmath102 , the quantity @xmath133 could have also some different applications . for example , knowing of @xmath141 makes possible to estimate the density @xmath142 of pairs of particles with the inter - particle distance less then @xmath21 . namely , in the binary approximation ( not more than one particle inside the self - sphere of any particle in the initial system ) it can be shown that : @xmath143 , where @xmath144 is given by eq . ( [ eq24app ] ) with @xmath145 . however , in further consideration the charge @xmath102 and the quantity @xmath133 will play only an accessory role . * the characteristic length @xmath130 as the neutrality radius . * in order to determine the required `` large '' characteristic length one should consider the charge @xmath146 in the region @xmath136 , where @xmath133 is given by ( [ eq24app ] ) and has the sense which is described above . let us remind that in this region the charge @xmath2 of the probe particle is already completely compensated by background charge of the probe particle self - sphere . consequently , in the region @xmath136 the charge @xmath102 represents a quantitative characteristic of deviation of the neutrality of the sphere with radius @xmath12 , centered at the prove particle , which is exceptionally caused by the charge of @xmath133 particles which are enter in this sphere from the rest of space . from ( [ eqnun ] ) and ( [ eq24app ] ) it could be seen that @xmath147 in the region @xmath148 almost exponentially decreases from the maximum value @xmath149 down to zero , with the increasing of @xmath12 . consequently , as the requested screening length , denoted here as @xmath150 , can be taken the root of equation @xmath151 where @xmath102 is given by ( [ eqnun ] ) and ( [ eq24app ] ) . in accordance with this @xmath130 can be treated as the neutrality radius . here , @xmath130 is the radius of such a sphere centered at the probe particle for which charge exchange with the rest of the space becomes practically negligible in comparison with the similar exchange in the case of probe particle self - sphere , which is illustrated by fig . [ fig : nu ] . in the case of the initial system ( see part 1 ) the radius @xmath130 can be interpreted as a radius of minimal sphere which can be considered as practically neutral one . from ( [ eq44 ] ) and ( [ eq24app ] ) it follows that the characteristic length @xmath150 represents the root of the equation : @xmath152 = e^ { - 1},\ ] ] which can be determined only numerically . here , it is presented in two equivalent form , namely @xmath153 where the coefficients @xmath154 and @xmath155 are taken in the form @xmath156/x , \qquad \eta_{\kappa } = 1 + x + g(x),\ ] ] since the member @xmath157 can be very well approximated by means of the easy expressions @xmath158 where at the point @xmath159 both expressions give the same value @xmath160 . for the coefficients @xmath154 and @xmath161 we obtain then following approximate expressions @xmath162 @xmath163 where @xmath81 . the behavior of @xmath154 and @xmath161 obtained numerically from basic equation ( [ eq45 ] ) and by approximative expressions ( [ eq46a ] ) and ( [ eq46b ] ) , is illustrated in fig . [ fig : eta ] . this figure shows that the mentioned approximative expressions give very good results . from ( [ eq47 ] ) it follows that : @xmath164 and @xmath165 . consequently , on the base of ( [ eq46 ] ) we have it that : @xmath166 for any @xmath167 , and wigner - seitz s radius @xmath168 . then , from ( [ eq48 ] ) it follows that : @xmath169 for @xmath170 . _ on the base of ( [ eq46 ] ) we have it that : @xmath171 for any @xmath167_. * the connection of @xmath49 with landau radii @xmath174. in part 2 it was shown that the simplifications which give the possibility to describe a two - component plasma in dh method by means of unique screening constant are unacceptable and that its ion and electron components have to be described by own screening constant @xmath175 and @xmath176 given by eqs . ( [ eq292 ] ) and ( [ eqalf ] ) . one can see that these screening constants are substantially different from the dh screening constant ( see part 2 ) . in the general case @xmath177 . unique exception is the case of completely classical plasma with @xmath178 , and @xmath179 , but even in this case the common value of ion and electron screening constants is significantly different from dh screening constant . the existence of two special screening constants @xmath175 and @xmath176 causes that in the case of two - component plasma we have two groups of screening parameters , analogous to the screening parameters described in the previous section , but for ion and electron components separately ( see part 2 ) . as first , we will consider `` small '' characteristic lengths @xmath180 and @xmath181 , which are analogous to the characteristic length @xmath59 in the single - component case , and have the sense of radii of the spheres centered on the probe particles which are classically forbidden for the ions in the case ( i ) and for electrons in the case ( e ) . keeping in mind eqs . ( [ eq251 ] ) and ( [ eq702 ] ) , as well as the facts that @xmath182 in the case of weakly non - ideal plasma and that in the classical case @xmath183 , we have the relation @xmath184 where @xmath185 , and @xmath186 and @xmath187 are known ion and electron landau s radii , namely @xmath188 one can see that @xmath186 and @xmath187 represent the approximation of @xmath180 and @xmath181 in the classical case in the regions @xmath189 . consequently , in the case of two - component classical plasma the characteristic lengths @xmath180 and @xmath181 can be treated as the corresponding generalization of landau s radii @xmath186 and @xmath187 . it is important that , contrary to @xmath174 which are principally unlimited , the radii @xmath190 for any @xmath191 , and that wigner - seitz s radii @xmath192 . the connection of @xmath172 and @xmath173 with non - ideality parameters @xmath193 and @xmath194 . * on the base of eqs . ( [ eq251 ] ) , ( [ eq702 ] ) , ( [ eq34b ] ) and ( [ eq33b ] ) in the case of classical plasma , we can obtain the relations @xmath195 where @xmath185 , and the quantities @xmath193 and @xmath194 are often used the classical ion and electron non - ideality parameters , given by relations @xmath196 let us emphasize that the parameters @xmath193 and @xmath194 , similarly to @xmath72 and @xmath73 in the single - component case , are also introduced from `` some physical reasons '' ( see for example @xcite . however , one can see that @xmath193 and @xmath194 represent the approximations of the ion and electron non - ideality parameters @xmath172 and @xmath173 in the region @xmath189 . consequently , @xmath172 and @xmath173 can be treated as the generalization of ion and electron non - ideality parameters of two - component classical plasma . the behavior of @xmath172 and @xmath173 as functions of @xmath197 is similar to the behavior of analogous parameters in the case of single - component system ( see fig . [ fig : gamma ] ) . let us draw attention that , _ contrary to the single - component case , the coefficients @xmath172 and @xmath173 illustrate the physical inadequateness of debye s parameters @xmath78 and @xmath89 for two - component systems . _ namely , in accordance with eqs . ( [ eq702 ] ) and ( [ eq712 ] ) these coefficients represent the functions of the parameters @xmath198 . however , in two - component case we have that @xmath199 and the quantity @xmath200 . * the charges @xmath202 . * similarly to the single - component case , we can take the electro - neutrality condition ( [ eq42 ] ) in the form : @xmath203 , where @xmath99 , and the quantities @xmath204 and @xmath205 are given by relations @xmath206 with the charge density @xmath207 defined by eq . ( [ eq22 ] ) . similarly to the single - component case , @xmath208 is the total charge of the whole sphere with radius @xmath12 , centered at the probe particle , and @xmath205 is the total charge of the rest of space . in all further considerations it is needed to know the charges @xmath204 only in the region @xmath209 . keeping in mind eq . ( [ eq21b ] ) , as well as the fact that @xmath210 under the condition ( [ eq42 ] ) , we have that @xmath211 where @xmath185 , @xmath48 and @xmath212 is the same function as in eq . ( [ eq24app ] ) . for the practical applications of described method it is important that the expression ( [ eqalf ] ) for the coefficient of the electron - ion correlation @xmath37 , which figure not only in eq . ( [ eq64 ] ) , but in expressions for all relevant quantities , can be very well approximated by two simple expressions , namely @xmath213 where the right side represents the first member of the expansion of right side of eq . ( [ eqalf ] ) in the series with the respect to @xmath214 , and @xmath215 the behavior of the coefficient @xmath37 , given by eqs . ( [ eqalf ] ) , ( [ eq50 ] ) and ( [ eq51 ] ) is illustrated by fig . [ fig : alpha ] . * the connection of @xmath201 with the radii @xmath216 . * in the two - component case the ion and electron screening constants @xmath175 and @xmath176 have the sense of basic parameters of the method developed in part 2 . consequently , the corresponding radii @xmath217 and @xmath218 also represent the screening parameters of this method . in order to clarify the real role of @xmath219 we will introduce the characteristic lengths @xmath220 and @xmath221 which in the cases ( i ) and ( e ) have the similar sense as the radius @xmath95 in the single - component case . it can be shown that in the regions @xmath222 the parameters @xmath201 are the roots of equations @xmath223 where @xmath224 , and @xmath225 is given by eq . ( [ eq21b ] ) . consequently , we have that in the region @xmath226 the relations @xmath227 are valid . due to the behavior of @xmath228 , which is described in part 2 , we have it that in the case @xmath229 the parameter @xmath230 . in all examined cases in the region @xmath231 we obtain that @xmath232 except of the points @xmath233 , where it is @xmath234 . from just mentioned it follows that the radii @xmath235 and @xmath236 have clear physical sense only in the region @xmath226 where it is defined by the relation ( [ eq39a ] ) . then , similarly to the single - component case we have to examine the behavior of the charge @xmath237 for @xmath238 . since in this case @xmath239 is given by eq . ( [ eq64 ] ) , we have that @xmath240 . it means that @xmath241 , in accordance with the behavior of @xmath242 , is comparable with the probe particle charge @xmath243 , excluding eventually the region @xmath244 . consequently , excluding the cases when @xmath37 is close to unity , the parameters @xmath220 and @xmath221_cannot be treated as a characteristic length of full screening ( neutrality ) of the probe particle charges in the cases of ( i ) and ( e)_. because of that , such characteristics lengths are determined here in the similar way as in the single - component case . finally , we wish to draw attention to the fact that in the weakly non - ideal plasma ( @xmath245 ) the radii @xmath246 and @xmath247 are very closed to `` medium '' characteristic lengths in the corresponding single - component systems ( ion gas on the negative background and electron gas on the positive background ) . this fact has already been considered in connection with non - applicability of dh method in the case of two - component plasmas . namely , in @xcite , where the case of the classical plasma with @xmath178 and @xmath179 were considered ( see part 1 ) , the screening constants and radius close to @xmath248 and @xmath249 were used , instead of dh screening constant @xmath250 and radius @xmath251 . the results obtained in this work justify such a choice . * the expressions for @xmath252 and @xmath253 . * similarly to part 1 , we will introduce the screening lengths @xmath252 and @xmath253 which represent the roots of the equations @xmath254 where the charge @xmath204 is given by eq . ( [ eq64 ] ) and one of eqs . ( [ eqalf ] ) , ( [ eq50 ] ) and ( [ eq51 ] ) . from it follows that eq . ( [ eq100 ] ) can be taken in the form @xmath255 = e^ { - 1},\ ] ] which is the same as the form of the corresponding equation for the radius @xmath130 from previous section . consequently , we have that @xmath252 and @xmath253 are given by relations @xmath256 @xmath257 where the coefficients @xmath258 and @xmath259 are given by eqs . ( [ eq46a])-([eq48 ] ) . on the base of these expressions we have it that the relations @xmath260 and @xmath261 , as well as @xmath262 , are valid in the whole regions @xmath263 . the behavior of the coefficients @xmath264 and @xmath265 in a wide region of @xmath266 , which are given by eqs . ( [ eq292 ] ) , ( [ eqalf ] ) , ( [ eq712 ] ) , ( [ eq46a])-([eq48 ] ) , ( [ eq104i ] ) and ( [ eq104e ] ) , is displayed in figs . [ fig : eta1 ] and [ fig : eta2 ] . these figures clearly illustrate the changes of the behavior of @xmath264 and @xmath267 with the increasing of deviation of the screening constant @xmath248 from their classical values in the case @xmath268 . namely , fig . [ fig : eta1 ] shows the behavior of @xmath269 and @xmath270 in the region @xmath271 where the classical and general expressions give practically the same results , while fig . [ fig : eta2 ] is related to the region @xmath272 where the classical and general expressions give very different results . * the applications and comparison with the existing experimental data . * as we have already mentioned , the necessity of interpretation of experimental data caused several attempts ( see for example @xcite ) to determine of the characteristic screening length @xmath273 for two - component plasma which was taken as : @xmath274 , where @xmath78 is debye s radius for two - component system ( see part 2 ) . in order to compare the values of @xmath273 and @xmath130 we take here @xmath273 in the form @xmath275 where the correction factor @xmath276 . we will have in mind that the classical plasmas with @xmath178 and @xmath179 has been considered in the above mentioned papers , and that in such plasmas : @xmath277 , @xmath278 , @xmath279 , @xmath280 and @xmath281 . the comparison of our results with the existing experimental data is performed for several cases and is presented in fig . [ fig : etaexp ] . this figure shows the behavior of the parameter @xmath282 determined by ( [ eq104i])-([eq104e ] ) and the correction coefficients @xmath283 taken from @xcite in the region @xmath284 . one can see a good agreement with the correction coefficients obtained from several measurements of conductivity from @xcite . in order to examine usage of @xmath285 as a screening radius in the expressions of spitzer s type for the conductivity of plasma , we performed a calculation of conductivity for a fully ionized plasma with @xmath178 and @xmath286 , where @xmath280 . the cases of plasmas with @xmath287 and @xmath288 in the region @xmath289 were considered . the conductivities were determined by means of corrected spitzer s expression @xcite where debye s radius @xmath78 for two - component plasma is replaced ( in so called coulomb logarithm ) by screening radius @xmath273 . the calculation was performed for @xmath290 , @xmath291 , where @xmath292 is the corrected factor from @xcite , and @xmath293 . the results are presented in figs . [ fig : spe18 ] and [ fig : spe19 ] . in the same figures the corresponding values of conductivity determined by improved rpa method which is applicable for dense non - ideal plasmas @xcite are also presented . the figs . [ fig : spe18 ] and [ fig : spe19 ] show that spitzer s conductivity with @xmath293 apparently converge to rpa conductivity with the increasing of electron density @xmath28 from @xmath294 to @xmath288 , while spitzer s conductivity with @xmath290 and @xmath291 lie significantly above . one can see that use of @xmath293 provides the complete agreement with the results of rpa calculations in the case @xmath295 . it is important that this agreement becomes better when the non - ideality degree increases . the new model method for describing of the electrostatic screening in single- and two - components systems ( electron - ion plasmas , dusty plasmas , some electrolytes , etc . ) developed in part 1 and part 2 of this work , generates a group of new screening parameters . here , we keep in mind `` small '' characteristic length @xmath59 and the non - ideality parameters @xmath75 and @xmath296 in the single - component case , and the corresponding characteristic lengths @xmath49 and the non - ideality parameters @xmath172 and @xmath173 in the two - component case . in connection with the mentioned screening parameters is established that @xmath59 and @xmath49 represent the generalization of classical landau s radii @xmath61 and @xmath174 , and @xmath297 , @xmath172 and @xmath173 - the generalization of known classical non - ideality parameters @xmath72 and @xmath73 in the single - component case , and @xmath193 and @xmath194 in the two - component case . apart of that , in this paper are introduced into consideration `` medium '' and `` large '' characteristic lengths @xmath95 and @xmath130 in the single - component case , and @xmath201 and @xmath285 in the two - component case . the behavior of these radii is examined in the whole regions @xmath298 and @xmath299 , where @xmath107 and @xmath248 are the corresponding screening constants , and @xmath21 and @xmath54- the corresponding wigner - seitz s radii . it was found that the considered characteristic lengths satisfy the relations @xmath300 @xmath301 @xmath302 where @xmath81 and @xmath185 . these relations establish the two hierarchy systems of the characteristic lengths , and causes a redefinition of wigner - seitz s radii as a boundary screening lengths . then , it was found that the radius @xmath108 in the single - component case has the sense only in the region of @xmath303 , where @xmath123 , and the radii @xmath304 have the sense only in the regions @xmath305 , where @xmath306 in the two - component case . the results of application of the characteristic length @xmath252 as the neutrality radius were compared in this paper with existing experimental data in the cases of the classical plasmas with @xmath178 and @xmath179 . it was found their very good agreement . finally , we wish to draw attention that developed method is suitable for some astrophysical applications . here we keep in mind that in outer shells of stars the physical conditions change from those which correspond to the rare , practically ideal plasma , to those which correspond to extremely dense non - ideal one . however , the method presented gives a possibility to describe the electrostatic screening of all such outer shells in the same way , by means of the obtained screening characteristics . the authors are thankful to the university p. et m. curie of paris ( france ) for financial support , as well as to the ministry of science of the republic of serbia for support within the project 141033 `` non - ideal laboratorial and ionospheric plasmas : properties and applications '' . the method developed in part 1 and part 2 serve for describing of the electrostatic screening in the single- and two - component systems which contain only charged particles . however , as it is well known , the presence of the neutral component in plasma can often be neglected from the aspect of inner - plasma screening . it gives the possibility to apply the results obtained in this work on the problem of the lowering of the atomic ionization potential in plasmas with the neutral component ( see part 1 ) . in such a way we will be able to compare the results obtained within the developed and dh method . the results of this comparison will give another important example of the inapplicability of dh method . in this context we will remind that in part 2 the potential energies @xmath307 and @xmath308 of the ion and electron in plasma were determined . in the case @xmath178 and @xmath179 , when @xmath309 , the ion potential energy @xmath307 was compared with the corresponding dh ion potential energy @xmath310 . then , we will denote with @xmath311 and @xmath312 the atomic ionization potential determined within the developed and dh method , and take into account that @xmath313 and @xmath314 where the proportionality coefficients are equal or at least very closed . from here it follows the relation @xmath315 the behavior of the right side of this relation is shown in fig . 5 of part 2 . accordingly to this figure we can conclude that @xmath316 in the whole region @xmath317 . it is important to noted this fact , since the handbooks ( see also @xcite ) , often used in laboratories , recommend just dh lowering of the atom ionization potential @xmath318 , which would not be taken as one of screening characteristics of electron - ion plasmas .	the physical sense and the properties of a group screening parameters which are determined in the previous papers ( part 1 and part 2 ) for single- and two - component systems is discussed in this paper . on the base of data from the mentioned papers are determined two new characteristic lengths which complete a new hierarchy system of screening lengths in plasma . it was shown that the methods developed in part 1 and part 2 generates results which are applicable to the strongly non - ideal gaseous and dusty plasmas , and manifest a very good agreement with the existing experimental data .
grs1758258 , discovered by art - p and sigma on granat in the spring of 1990 , is one of the brightest persistent sources above 100 kev in the vicinity of the galactic center ( @xcite ) . its spectrum extends up to 300 kev , and because this hard spectrum resembles those of other black hole candidates ( bhc ) in the hard (= soft x - ray low ) state , grs1758258 is generally considered to be a bhc ( @xcite ; @xcite ; @xcite ; @xcite ) . the nature of the donor star orbiting around the black hole is still controversial , although a high mass companion is certainly ruled out ( e.g. @xcite ) . monitoring has shown that the source was in this hard state for almost a year during a three - year period before the hard x - ray flux ( above 35 kev ) decreased by a factor of 10 ( @xcite ; @xcite ) . this decrease in the hard x - ray flux may indicate a transition from the hard state to the soft (= soft x - ray high ) state . unfortunately , the lack of simultaneous soft x - ray observations prevents us from drawing a definitive conclusion about the transition . a significant soft x - ray excess was suggested by simultaneous rosat - sigma observations in 1993 , when the hard component had a lower intensity ( @xcite ) . but a reanalysis suggests that the excess was due to the contamination from gx 51 ( @xcite ) . from asca observations , it was inferred that the soft component contributes only a few percent of the total x - ray and gamma - ray flux when the source is in its hard state although there were no simultaneous gamma - ray observations ( @xcite ) . a significant soft component for grs 1758258 was also reported by @xcite , and in 1999 it has oscillated between the hard and soft states(@xcite ) . to be able to understand the nature of the soft component when the spectrum is in its hard state , it is essential to have good simultaneous coverage of grs 1758258 over as broad an energy range as possible . here we report the results of our simultaneous radio , millimeter , x - ray and gamma - ray observations of grs 1758258 in august 1997 . the radio observations were performed by the very large array ( vla ) . the millimeter observations were performed by iram , and we compared to the early observations made using the swedish eso sub millimeter telescope ( sest ) at la silla ( chile ) . the x - ray and gamma - ray observations involved the proportional counter array ( pca ) and high energy x - ray timing experiments ( hexte ) on the _ rossi x - ray timing explorer ( rxte ) _ and the oriented scintillation spectrometer experiment ( osse ) on the _ compton gamma - ray observatory ( cgro)_. since the energy ranges of these three instruments overlap , we get a good measurement of the complete x- and gamma - ray spectrum . the rapid x - ray variability can also provide direct information about the accretion disk . observations of grs 1758258 made by _ rxte _ in 1996 ( @xcite ) found that the source had a typical power spectrum for bhcs in their hard states . the integrated fractional rms amplitude ( ifra ) was 24% . unlike the case in 1996 when the power spectrum had little energy dependence , our observations show that the 5 10 kev band was more variable than the 2 5 kev and 10 40 kev bands . in section 2 , we describe the details of the observations and data reductions . daily and short time scale light curves are shown in section 3 . in section 4 , we present the analysis of the spectral data . the timing analysis is discussed in section 5 . we summarize the results in section 6 . at radio wavelengths , grs 1758258 has been found to be a `` microquasar '' with relativistic jets emanating from a central variable source vla - c ( @xcite ; @xcite ) . we performed a series of nine vla observations between 1997 august 3 24 , spanning our entire _ some of these were simultaneous with the _ rxte _ observations discussed in 2.2 . all the observations were made in the c configuration at the 6 and 3.5 cm wavelengths ( 4860 and 8460 mhz ) , with a bandwidth of 100 mhz . the vla runs were all 2 hours long except for 1.5 hours on august 15 . this included time for both amplitude and phase calibration : the amplitude calibrator was 1331 + 305 and the phase calibrator 1751253 , whose bootstrapped flux density was always close to 0.48 jy at 6 cm and 0.27 jy at 3.5 cm , respectively . table 1 gives the flux densities for the central radio source for each of our separate observations . the flux densities and errors have been measured by fitting an elliptical gaussian to the source with the aips task jmfit . the upper limits are @xmath3 based on the map rms noise , measured using the aips task imsta . the radio light curves are also plotted in figure 1 . as has been found previously , vla - c is often a weak radio source close to the detection threshold in our daily observations . the central source is clearly detected ( @xmath4 ) when we concatenate the vla data from all our 1997 august 324 observations . the average flux density values were @xmath5 mjy at 4860 mhz and @xmath5 mjy at 8460 mhz , giving a flat spectral index of @xmath6 . we combine the averaged radio spectrum with our averaged x - ray and gamma - ray spectra in figure 8 . the radio emission suggests the continual presence of an outflow from vla - c . the approximately flat spectrum is not consistent with optically thin emission ( unless the electron distribution is exceptionally hard ) and is indicative of some synchrotron self - absorption in the radio - emitting regions . one possible geometry would be a conical jet such as discussed in @xcite . for comparison , in 1997 august the nearby source vla - d had average flux densities of @xmath5 mjy at 4860 mhz and @xmath7 mjy at 8460 mhz , giving a spectral index of @xmath8 . the 6 cm flux density of vla - d is @xmath9 the value it had in 1992 which was also measured in the c configuration ( @xcite ) . maps and further details of these and other vla observations of grs 1758258 will be presented elsewhere ( j. mart et al . 2000 , in preparation ) . the discovery of a molecular cloud ( bally & leventhal 1991 ; mirabel et al . 1991 ) in the direction of 1e 1740.72942 raised the picture of a compact source embedded inside a molecular cloud that is accreting material from the cloud . grs 1758258 has similar x - ray properties to 1e 1740.72942 , but the lower value of @xmath10 @xmath11 suggests that grs 1758258 is not inside a dense molecular cloud ( chen , gehrels , & leventhal 1994 ) . to determine whether there is a molecular cloud in the direction of grs 1758258 , we performed millimeter observations during our multi - wavelength campaign on 1997 august 7 using the iram 30-m telescope at pico veleta ( spain ) . we also report here on earlier millimeter observations of grs 1758258 that were conducted in april 1995 and october 1995 using the 15-m sest . as well as observing the core radio source , we also made millimeter observations along the radio jets in order to look for signatures of the jet interaction with a molecular cloud such as an increase in the width of the line or the detection of wings in the peak . for example , we previously detected a correlation between the rosat x - ray hot spots and the co mm measurements in observations of ss 433 ( durouchoux et al . the higher ionization environments in shocked molecular gas and near x - ray sources can also enhance the emission from molecules such as @xmath12 and hcn ( krolik & kallman 1983 ; phillips & lazio 1995 ; lepp & dalgarno 1996 ; yan & dalgarno 1997 ) . we performed observations using the transition at 115.3 ghz and the transition at 230.5 ghz . observations were made in the direction of the core source ( @xmath13 17:58:07 , @xmath14 25:44:30 , epoch 1950.0 ) and at one location in the northern radio jet ( @xmath13 17:58:06.8 , @xmath14 25:43:45 ) and one in the southern jet ( @xmath13 17:58:05.5 , @xmath14 25:45:50 ) . the observations were carried out in a position beam switching mode , choosing an absolute reference position @xmath15 in @xmath16 and @xmath17 in @xmath18 away from the source . such a distant location was required to find a region without radio emission in the crowded galactic center area . the receivers yielded an overall system temperature ( including sky noise ) @xmath19 and 300 k at 115 and 230 ghz respectively . the backend was an acousto - optical spectrometer . the low resolution spectrometer with a channel separation of 0.7 mhz ( or 2.4 @xmath20 ) was used for these observations . the antenna beam - widths were 45 and 23 for the and transitions respectively . all the spectral intensities were converted to the main brightness temperature scale , correcting for a mean beam efficiency of 0.7 at 115 ghz and 0.5 at 230 ghz . the calibration was checked by monitoring m17sw , and was found to be consistent between the different observing periods to within 19% . the pointing accuracy , obtained from measurements of the sio maser w hya , was better than 5 . a molecular cloud was found in and in the direction of grs 1758258 with a velocity @xmath21 . the results are given in table 2 . the peak area / ratio of @xmath22 is consistent with the values found in other molecular clouds ( oka et al . 1998 ) . a line with the same velocity was also detected in the northern and southern parts of the jet . for the northern direction , the flux density was slightly lower than in the core direction , but not significantly . for the southern direction , the width of the line is slightly larger than for the core source , but again not significantly . was also detected in the north and south jets with velocities , peak areas , and line widths consistent with the ones measured for this transition in the core source . we conclude that there are no obvious signatures for the interaction of the jet with the molecular cloud . we also searched for the @xmath23 line in the and transitions , but no significant detections were made : integrating over 10 minutes the mean @xmath24 upper limit was @xmath25 k for both the transitions . we also did not detect the or lines : the mean @xmath24 upper limit was @xmath26 k. we observed simultaneously in the transition at 89.2 ghz with the sis 3mm-1 receiver , in the transition at 88.6 ghz with the sis 3mm-2 receiver , in the at 220.4 ghz with the 230g1 receiver , and the transition at 230.5 ghz with the 230g2 receiver . the half power beams varied from 27 at 90 ghz to 10.4 at 230 ghz . as for the sest runs , the observations were carried out in a position beam switching mode , choosing an absolute reference position @xmath15 in @xmath16 and @xmath17 in @xmath18 away from the source . the main system temperatures were 356 , 389 , 912 , and 1000 k , and the opacities 0.16 , 0.18 , 0.58 , and 0.62 at 89.2 , 88.6 , 220.4 , and 230.5 ghz respectively . we observed grs 1758258 at 12 different positions on the jet axis separated by 24 . the integration time was 1 minute on each location . the individual observations did not have sufficient significance to study the line emissions , and so we added them together to make an average result for each transition . we did not detect any lines in the ( 0.04 k , @xmath24 ) and ( 0.6 k , @xmath24 ) spectra . the transition was clearly detected at @xmath27 with a line width of @xmath28 , similar to that seen earlier by sest . however , the intensity of @xmath29 k and peak area of @xmath30 are almost double those seen earlier by sest . the transition was also detected in our iram observation at @xmath31 with a line width of @xmath32 , intensity @xmath33 k and peak area of @xmath34 . although the statistical significance is weak , this also seems to be brighter than was seen in our sest observations . the comparison between the sest and iram results suggests a brightening in the outflow in the 2 years between the observations . the fact that the iram beam size is less than half that of sest implies that if the emission is diffuse , there has actually been an increase of @xmath35 in the line emissions . however , we caution that a longer - term monitoring program is required to prove that the line emissions are variable . we conclude that while there is a molecular cloud in the direction of grs 1758258 , we have not yet found any conclusive signatures that there is an interaction between the jet and this molecular cloud . we suggest that longer observations using the transition would also be worthwhile to search for an interaction . the pca consists of five xe proportional counters with a total effective area of about 6500 @xmath36 ( @xcite ) . its sensitive energy range is from 2 to 60 kev with an energy resolution of @xmath37 18% at 6 kev , and a time resolution of 1 @xmath38s . the hexte consists of two independent clusters ( clusters a and b ) each containing four phoswich scintillation detectors ( @xcite ) . it has a total effective area of 1600 @xmath39 and a sensitive energy range of 15 250 kev with an energy resolution of 15 % at 60 kev and a time resolution of 8 @xmath38s . both the pca and the hexte have a field of view of @xmath40 ( fwhm ) . both the pca and hexte made observations on 5 days during the period of august 13 - 24 , 1997 . during the observations , the on - source pointing direction ( ra = 270.315@xmath41 , dec = 26.245@xmath41 , j2000 ) was offset from grs 1758258 by half a degree to avoid the x - ray source gx 51 , which is @xmath42 away from the on - source location . the flux of gx 51 during the observation was about 1 crab based on the observations by the all sky monitor ( asm ) on the _ rxte_. however , the offsetting effectively reduced the contamination from gx 51 , whose contribution is estimated to be much less than 0.033% of the total count rate in the energy band of 2 12 kev . the hexte cluster a was rocked by @xmath43 and cluster b by @xmath44 to make the background measurements with a switching time of 16 seconds . the pca has no direct background measurements . the pca and hexte data were reduced using ftools 4.1.1 . both data sets were screened so that only those data were used when the spacecraft was not passing through the south atlantic anomaly and when the source was observed at elevations @xmath45 above the earth s limb and the pointing was offset @xmath46 from the on - source pointing to filter out the slew data . after the screening , 14 observation segments have valid data . the start and stop times of each segment are listed in table 3 . figure 2 shows the pca 16-second resolution light curves for the 14 observation segments . for the pca , standard 2 production data from all layers of the five pcu detectors were used to generate an average spectrum over our entire _ the ftools program pcarsp ( version 2.36 ) was used to generate the response matrix file with corrections made for the off - axis pointings . the total exposure time with dead - time correction is 33399 seconds . a scheme of adding systematic errors to the pca data was also applied ( @xcite ) . the added errors are 0.5% for 2.5 5.5 kev , 1% for 5.5 8 kev , 0.5% for 8 20 kev , and 3% for 20 90 kev . the estimation of the pca background is complicated by the fact that the current pca background models do not include the galactic diffuse emission , which significantly contaminates our observations . we used the previous _ rxte _ observations in the vicinity of grs 1758258 ( @xcite ) to estimate the contribution from the diffuse emission . by combining these background data with the detector internal background ( q6 + activation ) , we are able to estimate the true background for our pca measurements . in the timing analysis , we neglect the time variation of the diffuse emission , which is estimated to be less than 10% of that of grs 1758258 based on the background observations . therefore , we remove the effects of the diffuse emission by subtracting from the light curves constant count rates of 15.1 counts / s for the 2 5 kev band , 17.9 counts / s for 5 10 kev , and 8.4 counts / s for 10 40 kev . for the hexte , we only used the cluster a data for the spectral analysis , because the data from cluster b are less reliable due to the count piling problem and the off - axis pointing . the data were taken using the e_8us_256_dxif mode . the hexte background measurements agree very well with the on source measurement in the energy channels above 125 kev . the response matrix file hexte_97mar20c_pwa.rmf was used . the collimator effective area has been corrected with the calibration file hexte_pwa.fov for the off - axis pointing . the total exposure time with dead - time correction is 9683 seconds . the osse instrument , consisting of four nearly identical nai - csi phoswich detectors , has a field of view of 3.8@xmath41 @xmath47 11.4@xmath41 ( fwhm ) , and has a sensitive energy range of 0.05 10 mev with an energy resolution of 7.8% at 661 kev ( @xcite ) . the current osse observations were carried out on august 5 19 , 1997 . during the observations , the four detectors made alternating on - source and off - source pointings with 2.048 minute dwell time on each pointing . the rocking direction , which is perpendicular to the long dimension of the collimator , was at an angle of 41@xmath41 to the galactic plane to avoid confusing sources . the on - source pointing was at ra = 270.45@xmath41 , dec = 26.12@xmath41 . detectors 1 and 2 did the standard @xmath48 off - source pointings ( ra = 257.45@xmath41 , dec = 29.59@xmath41 ; ra = 282.5@xmath41 , dec = 21.37@xmath41 ) . however , due to the instruments configuration , detectors 3 and 4 could not be rocked to the regular off - source pointings , and instead had off - source pointings at + 12@xmath41 and 7@xmath41 ( ra = 277.60@xmath41 , dec = 23.48@xmath41 ) . unfortunately , the known osse source gs 182624 ( @xcite ) sits in the center of the 7@xmath49 background pointing , and so we did not use the data from detectors 3 and 4 . the osse data were reduced with the igore data analysis package . a count spectrum and a response matrix were generated from the data of detectors 1 and 2 with the background subtracted . both the count spectrum and the response matrix could be read by xspec ( version 10.0 ) to make a joint fit with the pca and the hexte data . the total osse exposure time is 320078 seconds . since the background pointings are 12 degrees away from the on - source pointing , the measured background , which igore automatically subtracts from the data during the processing , may be different from the real background at the on - source location . we have checked all the sources monitored by the asm/_rxte_. no other significant hard x - ray point sources are in the fields of view of detectors 1 and 2 . however , an important problem is the spatial variation in the galactic diffuse emission ( gde ) . since our background pointings are farther away from the galactic center than the on - source pointing , we need to make a correction to get the true on - source background . based on several osse observations of the gde and coordinated sigma observations ( @xcite ) , a model of the gde distribution in and around the galactic plane is : @xmath50 e^{-\frac{b^{2}}{b_{0}^{2}}}\ ] ] where @xmath51 is the photon flux emitted from a unit area in the galactic coordinates , @xmath52 is the photon energy in kev , @xmath53 and @xmath54 are the galactic longitude and latitude respectively , posm(@xmath52 ) is the positronium continuum spectrum defined in xpsec , @xmath55 is the normalization factor of posm(@xmath52 ) , @xmath56 = 0.78 photons/@xmath57/s / kev / degree@xmath58 , @xmath16 = 2.55 , @xmath59 = @xmath60 , @xmath61 = @xmath62 = @xmath63 , and @xmath64 = @xmath65 . these numbers were determined by folding the observations at l=0@xmath41 , 25@xmath41 and 339@xmath41 through the detector response function , which is assumed to be a gaussian function with a two - dimension fwhm of 3.8@xmath41 @xmath47 11.4@xmath41 . the lack of measurements at locations other than these three longitudes prevents us from determining the exact form of @xmath55 . but we can still set upper and lower limits to the background correction that should be made to our osse data . the lower limit is when @xmath55 = @xmath61 . the upper limit is when @xmath55 = @xmath59 for the on - source measurements and @xmath55 = @xmath61 for the off - source measurements . the two limiting versions of the correction spectrum are shown in figure 4 , which can be simulated with xspec and then subtracted from the osse data . daily light curves of the vla , _ rxte _ , and osse observations are shown in figure 1 . we fit constants to the light curves . the fitting results , listed in table 4 , show that grs 1758258 was stable on the day time scale , although there was a small decline in the flux observed by the pca . on shorter time scales , grs 1758258 is clearly variable . significant variations are seen in the pca 16-second light curves ( figure 2 ) , which show the 14 valid observation segments . some small bursts are seen in the light curves . in figure 3 , the observation segment 11 is examined at higher time resolutions . the peak near 650 seconds in the 16-second resolution light curve actually consists of several small bursts on the 1-second time scale . moreover , a narrow burst near 630 seconds in the 0.1-second light curve still has substructure in the 0.01-second resolution light curve . however , these bursts are relatively small compared to those in other sources such as gx 3394 ( @xcite ) . for all three instruments , we combined all the data to make one spectrum . though the source is highly variable on small time scales , we will show in section 5.3 that there are no significant trends in the hardness ratio . thus the average spectrum shown here should be a good representative shape , though the instantaneous amplitude varies significantly with time . since the hexte data has fewer problems with the background , in the following discussion we will normalize the osse and pca spectra to the hexte spectrum . the hexte data can be well fit by a power law with photon index of [email protected] in the energy range of 15 125 kev ( figure 5 ) . above 125 kev , the background dominates the data . the photon flux between 20 100 kev is ( 2.0 @xmath66 0.4 ) @xmath47 10@xmath67 photons/@xmath57s . this is consistent with the average flux of ( 2.8 @xmath66 0.2 ) @xmath47 10@xmath67 photons/@xmath57s detected by the batse earth occultation technique in the same period . the errors are 1-sigma throughout this paper . figure 6 shows the joint spectra of the osse and hexte data . the fit model is a power law with an exponential cutoff ( ple ) : @xmath68 where @xmath69 is the photon flux , @xmath70 is the normalization flux in the unit of photons/@xmath57/s / kev , e is the photon energy in kev , @xmath16 is the photon power - law index , and @xmath71 is the cutoff energy . the higher curve in the figure , which has been scaled up by a factor of 10 for illustrative purposes , has the minimum background correction applied to the osse data , while the lower curve is the maximum background correction case . the most significant spectral difference between the minimum and maximum corrections is the difference in the cutoff energies . this is not surprising , since the effect of the positronium continuum is significant only when the energy is close to and less than 511 kev . however , for a joint fit to the pca , hexte and osse data , the ple model alone ( figure 7(e ) ) is not as good as the sunyaev - titarchuk ( st ) compton scattering model ( @xcite ) . an example of a st fit is shown in figure 7 ( a ) & ( b ) . in table 5 , we give the fit parameters for two cases , one using the maximum osse background correction , the other using the minimum . since these two extreme cases set the limits for the parameters , we use the average values over the two cases ( the average column in table 5 ) as the true values . assuming a distance of 8.7 kpc , we get a soft x - ray luminosity ( 2 20 kev ) of @xmath72 ergs / s and a hard x - ray luminosity ( 20 200 kev ) of @xmath73 ergs / s . these values place grs 1758258 outside the `` x - ray burst box '' in figure 1 of @xcite . this further supports that grs 1758258 is a black hole . the hydrogen column density is ( 2.0 @xmath66 0.1 ) @xmath74 @xmath1 , which is consistent with the asca and exosat observations ( @xcite ; @xcite ) . the residuals in figure 7 suggest that no significant iron lines are seen in the spectrum . @xcite and @xcite generalized the st model to cover a wide range of the plasma parameters , such as low thomson depths and relativistic effects . the best ht model fit to the data gives a negative hydrogen column density ( @xmath75 ) and a seed photon energy @xmath76 kev ( or the cutoff energy in the lower end of the spectrum ) . the negative @xmath77 suggests that the true value of @xmath78 must be much smaller than 0.85 kev . in the further model fits , we set @xmath78 = 0.01 kev . because we do nt have valid data below 2.5 kev , the exact value of @xmath78 does not affect the fit results as long as it is much smaller than 0.85 kev . with the fixed @xmath78 , the ht model alone can not fit the data well . adding a back - body component to the model gives a good fit ( figure 7(c ) ) . the best fit parameters are listed in table 6 . compared with the st model , the ht model gives a slightly higher plasma temperature due to the relativistic corrections . the black body temperature is about 1.18 kev , which is much higher that the seed photon energy ( @xmath79 kev ) . therefore , it is very likely that the soft component and the seed photons have different origins . the total flux of the soft black - body componnent is @xmath80 ergs/@xmath81/s , which is about 0.9% of the total continuum flux from 2 kev to 10 mev . the low flux of the soft component in the hard state has been reported in the asca observations ( @xcite ) . the hydrogen column density given by the ht model is lower than that by the st model , but is close to the asca results ( @xcite ) . adding a soft component to the ple model also improves the fit , in particular to the shape of the residuals ( figure 7(d ) ) . the fit parameters are listed in table 7 . the soft black - body component ( bb ) has a temperature of @xmath82 kev and a total flux of @xmath83 ergs/@xmath81/s , which is about 1.3 % of the total continuum flux from 2 kev to 10 mev and is slightly higher than that given by the ht model . the hydrogen column density obtained from the ( ple + bb ) model is lower than that from previous exosat and asca observations . we therefore conclude that the ( ple + bb ) model is less favored than the compton scattering models ( st and ht models ) . figure 7(f ) shows that a power - law model can not fit the data . this means that the exponential cut - off is real . in figure 8 we show the unfolded and normalized pca , hexte and osse spectrum using the st model . the timing analysis was done only on the goodxenon pca data using all five pcus . after the basic steps mentioned in section 2 , background - subtracted light curves were generated for three energy bands : 2 5 kev , 5 10 kev , and 10 40 kev . from these light curves , power spectra , phase lags , coherences and hardness ratios were calculated . in the calculations , the counts have been summed into 46.875 ms bins , the power spectra etc . were generated for intervals that are 384 seconds long , and the results were then averaged . we performed the power density spectra ( pds ) analysis on the 14 separate segments . in all cases , the power spectrum had a white noise at high frequencies , and this has been subtracted from the results shown here . we normalized our plots so that their integral gives the squared fractional rms variability , i.e. the power spectrum is in the unit of rms@xmath58/hz . the power spectra in figure 9 , which have been averaged over all the observation segments , resemble the typical low - state power spectrum of gbhc : a flat noise at the low frequencies , rolling over to approximately a power law at higher frequencies ( @xcite ) . we find that the power spectrum varies between the energy bands . the ifra ( 0.0025 10 hz ) is @xmath84 , @xmath85 , and @xmath86 for the energy bands 2 5 kev , 5 10 kev , and 10 40 kev respectively . our result is interestingly different from the previous _ rxte _ observation in 1996 aug . 3 5 when the ifra ( 0.002 60 hz ) was nearly independent of energy : 26% ( 2 6 kev ) , 27% ( 6 13 kev ) , 25% ( 13 28 kev ) ( @xcite ) . it is also interesting to compare our results with other sources in their low states . gx 3394 , for example , had the highest variability in the 2 5 kev band during an _ rxte _ observation ( @xcite ) . this energy dependence of the variability may be useful in characterizing different x - ray sources , and its implication is explored elsewhere ( @xcite ) . we fit the general shape of the average pds in figure 9 using two zero - centered lorentzians . though a broken power law is also commonly used in the fitting , the double - lorentzian model gives better fits in our case . an additional pair of lorentzians qpos are necessary to obtain a good fit . the lorentzian function is defined as : @xmath87 where @xmath88 is the frequency , @xmath89 is the central frequency , @xmath90 is the half width at half maximum ( hwhm ) , and a is the area under the function ( @xmath91 ) . the fit parameters are listed in table 8 . unlike the previous observations of grs 1758258 ( @xcite ) , the current observations detected the qpos at 0.21 hz and 0.42 hz instead of at 0.4 hz and 0.8 hz . our power spectra are typical of low state power spectra , with two breaks at @xmath92 0.1 hz and 1 hz . according to an inhomogeneous corona model ( @xcite ) , which assumes that soft photons are injected into an inhomogeneous corona from a blob spiraling into the central black hole , the two breaks are due to different mechanisms , and they can have different energy dependencies . we performed @xmath93 tests on the widths of the two zero - centered lorentzians to study their dependence on energy . if we assume that the width of the lorentzian is constant , a @xmath94 test on @xmath95 gives a reduced @xmath94 of 0.67 for 2 degrees of freedom and a q value of 0.51 , while the test on @xmath96 has a reduced @xmath94 of 3.5 for 2 degrees of freedom and a q value of 0.021 . therefore , at the 95% confidence level , the width of the lower zero - centered lorentzian @xmath95 has no significant difference between the three energy bands , but this may not be the case for the width of the higher lorentzian @xmath96 . further analysis using pds ratios shows that the general pds shapes are energy independent for grs 1758258 ( @xcite ) , although this technique is less sensitive to the detailed shape of the pds . in addition to studying the averaged power spectra , we also generated the power spectra integrated over all energies and searched for qpos in each separate observation segment . a broad qpo ( or `` peaked noise '' ) at @xmath97 hz is present in all the segments . it is a very stable feature in the power spectrum . a pair of harmonically - spaced narrow qpos at 1.4 hz and 2.8 hz with a total ifra of 5% were found in observation segment 7 ( figure 10 ) . they are not positively identified in the other observation segments . only two cases of these narrow significant x - ray qpos have previously been observed in black hole candidates in the hard state ( @xcite ) . another known characteristic of bhcs is that the phase lag between hard photons and soft photons changes very little with frequency ( @xcite ) . this is supported by our observations ( figure 11 ) . constant phase lags between the energy bands can fit the data well . the fits give a phase lag of @xmath98 radians between the energy bands 2 5 kev and 5 10 kev ( @xmath99 for 19 degrees of freedom , q = 0.66 ) and @xmath100 radians between 2 5 kev and 10 40 kev ( @xmath101 for 19 degrees of freedom , q = 0.042 ) . the apparent increase of the phase lags between the 2 5 kev and 10 40 kev bands may suggest a constant time delay over the frequencies , but the constant time delay model @xmath102 can not fit the data well , where @xmath103 is the phase lag , @xmath88 is the frequency , and @xmath104 is the time delay . the fit gives @xmath105 for 19 degrees of freedom and q = @xmath106 . this flat distribution of phase lags over a wide range of frequencies ( @xmath107 10 hz ) was predicted by the extended atmosphere model ( @xcite ) that assumes that the phase lags are generated by compton scattering . the atmosphere has a density profile of @xmath108 , and the size has to be as large as @xmath109 , where @xmath110 is the schwarzchild radius . bttcher & liang ( 1999 ) proposed a spiraling blob model that assumes that the seed photons are from a blob spiraling into the central object . it also predicted a flat phase lag distribution , but it does not need a huge atmosphere to account for the phase lags because they are caused by the motion of the blob from the cool outer disk to the hot inner disk . like the previous measurements of grs 1758258 ( @xcite ) and cyg x1 ( @xcite ; @xcite ) , the current observations found a high coherence between the hard and soft photons ( figure 12 ) . the coherence between 2 5 kev and 5 10 kev is @xmath111 , and @xmath112 between 2 5 kev and 10 40 kev over the frequency range of 0.01 1.0 hz . as @xcite pointed out , the high coherence ( @xmath113 ) between the photon fluxes at different energies can be produced by the extended atmosphere model ( hua & titarchuk 1995 ) . the spiraling blob model also predicted such a high coherence . the hardness ratio vs intensity plots for observation segment 7 are shown in figure 13 . similar results are seen in the other segments . the hardness ratio has little dependence on the photon flux , although the poor statistics prevent us from drawing any significant conclusions . the fact that the hardness ratios show little dependence on the time or brightness indicates that the spectral shape we have found did not change markedly in 1997 august . our simultaneous observations of grs 1758258 in radio , millimeter , x - ray and gamma - ray add support to the idea that the source is a black hole . it has a very hard spectrum with a photon power law index of 1.40 , but there is a cutoff energy around 200 kev . for compton scattering models , the electron temperature is about 45 kev without considering relativistic effects , and 52 kev with relativistic corrections . at a distance of 8.7 kpc , the hard x - ray ( @xmath114 kev ) luminosity is @xmath73 ergs / s , which is higher than the upper limit for a neutron star binary ( nsb ) . the power spectra resemble those of the bhcs , flat at low frequencies and breaking into a power law at higher frequencies . the phase lags between the hard and soft photons have a canonical frequency dependence for bhcs ( @xcite ) . a high coherence of about 1.0 is also found among the different energy bands . the flat distribution of phase lags and the high coherence were predicted by the extended atmosphere model and the spiral blob model . the radio flux was stable during the observations , and had a flat energy spectrum . the millimeter observations did not reveal any conclusive signatures of an interaction between the jet from grs 1758258 and the molecular cloud that lies in the direction of grs 1758258 . the spectrum can be fit with st , ( ht + bb ) , and ( ple + bb ) models . the hydrogen column density deduced from the st fitting is @xmath115 @xmath1 , @xmath116 @xmath1 from ( ht + bb ) , and @xmath117 @xmath1 from ( ple + bb ) . it is not surprising that the value of the hydrogen column density is model dependent because we have valid data only down to 2.5 kev . though the presence of a soft component is also model dependent the ( ht + bb ) and ( ple + bb ) models need a soft component while the st model does not we can certainly conclude that the flux of the soft component is low . the low flux of the soft component implies that the soft photon injection regions are either small or mostly covered by the hot corona . one interesting result from the ht model fitting is that the seed photons and the soft component may have different origins . the soft component is usually believed to originate from the cool optically thick disk ( e.g. @xcite ) . the seed photons may have two origins . one is the cool optically thick disk , and the other is the internal emission processes in the corona such as synchrotron emission ( e.g. @xcite ) . therefore , the seed photons are very likely from the internal sources . we find that grs 1758258 is more variable in the 5 10 kev energy band than in the other two energy bands ( 2 5 kev and 10 40 kev ) . this property is interestingly different from other sources , such as gx 3394 . systematic studies show that the property is indeed significantly different for the four hard x - ray sources : cygnus x1 , gx 3394 , grs 1758258 and 1e 1740.72942 ( @xcite ) . however , further observations of the sources are necessary to correctly understand the energy dependence of the variability when the sources are in different flux and spectral states . cll 03.313 & @xmath118 & @xmath119 05.308 & @xmath120 & @xmath121 08.176 & @xmath122 & @xmath123 11.211 & @xmath124 & @xmath125 14.182 & @xmath126 & @xmath127 15.187 & @xmath122 & @xmath122 18.149 & @xmath128 & @xmath125 20.123 & @xmath129 & @xmath125 24.174 & @xmath130 & @xmath131 lllcclr core & & @xmath132 & @xmath133 & @xmath134 & @xmath135 & @xmath136 & & @xmath137 & @xmath138 & @xmath139 & @xmath140 & @xmath141 north jet & & @xmath142 & @xmath143 & @xmath144 & @xmath145 & @xmath146 south jet & & @xmath147 & @xmath148 & @xmath149 & @xmath150 & @xmath151 lllr 1 & 13.501 & 13.540 & 10673 + 2 & 13.568 & 13.606 & 10673 + 3 & 13.635 & 13.673 & 10673 + 4 & 13.701 & 13.740 & 10673 + 5 & 13.768 & 13.776 & 10673 + 6 & 13.788 & 13.807 & 10673 + 7 & 14.502 & 14.540 & 10674 + 8 & 14.568 & 14.607 & 10674 + 9 & 14.635 & 14.673 & 10674 + 10 & 14.702 & 14.729 & 10674 + 11 & 18.103 & 18.121 & 10678 + 12 & 18.170 & 18.184 & 10678 + 13 & 20.103 & 20.123 & 10680 + 14 & 24.172 & 24.192 & 10684 + vla 4860 mhz & 0.123 & 2.28 & 8 & 0.02 + vla 8460 mhz & 0.139 & 1.57 & 8 & 0.13 + _ rxte _ asm & 1.93 & 1.59 & 16 & 0.062 + pca & 118.4 & 34.8 & 4 & 4.2@xmath152 + hexte & 5.88 & 1.156 & 4 & 0.33 + osse & 35.6 & 0.66 & 14 & 0.83 + @xmath153 @xmath1 ) & @xmath154 & @xmath154 & @xmath155 + @xmath156(kev ) & @xmath157 & @xmath158 & @xmath159 + @xmath160 & @xmath161 & @xmath162 & @xmath163 + @xmath164 & @xmath165 & @xmath166 & @xmath167 + @xmath93 ( dof ) & 0.91 ( 620 ) & 0.90 ( 620 ) & + q & 0.94 & 0.95 & + @xmath153 @xmath1 ) & @xmath168 & @xmath169 & @xmath170 + @xmath156(kev ) & @xmath171 & @xmath172 & @xmath173 + @xmath160 & @xmath174 & @xmath175 & @xmath176 + @xmath177 & @xmath178 & @xmath179 & @xmath180 + @xmath181(kev ) & @xmath182 & @xmath183 & @xmath184 + @xmath185 & @xmath186 & @xmath187 & @xmath188 + @xmath93 ( dof ) & 0.89 ( 618 ) & 0.89 ( 620 ) & + q & 0.98 & 0.98 & + @xmath153 @xmath1 ) & @xmath189 & @xmath190 & @xmath191 + @xmath16 & @xmath192 & @xmath193 & @xmath194 + @xmath71(kev ) & @xmath195 & @xmath196 & @xmath197 + @xmath198 & @xmath199 & @xmath200 & @xmath201 + @xmath181(kev ) & @xmath82 & @xmath82 & @xmath202 + @xmath185 & @xmath203 & @xmath204 & @xmath205 + @xmath93 ( dof ) & 0.91 ( 618 ) & 0.92 ( 618 ) & + q & 0.95 & 0.92 & + @xmath206 & @xmath207 & @xmath208 & @xmath209 + @xmath210 & @xmath211 & @xmath212 & @xmath213 + @xmath214 & @xmath215 & @xmath216 & @xmath217 + @xmath218 & @xmath219 & @xmath220 & @xmath221 + @xmath222 & @xmath223 & * * * & @xmath224 + @xmath225 & @xmath226 & * * * & @xmath227 + @xmath228 & @xmath229 & @xmath230 & @xmath231 + @xmath232 & @xmath233 & @xmath234 & @xmath235 + @xmath236 & @xmath237 & @xmath238 & @xmath239 + @xmath240(dof ) & 1.43 ( 64 ) & 1.31 ( 64 ) & 0.95 ( 64 ) + q & 0.013 & 0.049 & 0.59 +	we report the results of our multi - wavelength observations of grs 1758258 made in august 1997 . the energy bands include radio , millimeter , x - ray , and gamma - ray . the observations enable us to obtain a complete spectrum of the source over an energy range of 2 500 kev . the spectrum shows that grs 1758258 was in its hard state . it is well fitted by the sunyaev - titarchuk ( st ) compton scattering model with a plasma temperature of 45 kev and a thomson depth of 3.3 . taking relativistic effects into account , we get a little higher plasma temperature ( 52 kev ) by using the improved version of the st model ( ht model ) plus a soft black - body component . the spectrum is also fit by a power law with an exponential cutoff ( ple ) plus a soft black - body component . the temperature of the soft components in both models is about 1.2 kev , and the energy flux is less than 1.5% of the total x- and gamma - ray flux . the deduced hydrogen column density is in the range of @xmath0 @xmath1 . no significant iron lines are detected . the radio emission has a flat energy spectrum . the daily radio , x - ray and gamma - ray light curves show that grs 1758258 was stable during the observation period , but was highly variable on smaller time scales in x- and gamma - rays . the power density spectra are typical for the low - state , but we find the photon flux for the 5 to 10 kev band to be more variable than that in the other two energy bands ( 2 5 kev and 10 40 kev ) . harmonically spaced quasi - periodic oscillations ( qpos ) are observed in the power spectra . the phase lags between the hard photons and the soft photons have a flat distribution over a wide range of frequencies . a high coherence of about 1.0 ( 0.01 1 hz ) between the hard photons and the soft photons is also obtained in our observations . we compare these results with two variation models . our millimeter observations did not reveal any conclusive signatures of an interaction between the jet from grs 1758258 and the molecular cloud that lies in the direction of grs 1758258 . # 1#2#3@xmath2
in recent years there has been considerable interest in the cosmological models with variable gravitational constant @xmath7 and the cosmological constant @xmath0 . variation of the gravitational constant was first suggested by dirac @xcite in an attempt to understand the appearance of certain very large numbers , when atomic and cosmic worlds are compared . he postulates that the gravitational constant @xmath7 decreases inversely with cosmic time . canuto et al . @xcite made numerous suggestions based on different arguments that @xmath7 is indeed time dependent . beesham @xcite has studied the creation with variable @xmath7 and pointed out the variation of the form @xmath8 , originally proposed by dirac @xcite . on the other hand , einstein introduced the cosmological constant @xmath0 to account for a stable static universe as appeared to him at the time . when he later knew of the universal expansion he regretted its inclusion in his field equations . now cosmologist believe that is not identically , but very close to zero . they relate this constant to the vacuum energy that first inflated a universe causing it to expand @xcite . from the point of view of particle physics a vacuum energy could correspond to quantum field that is diluted to its present small value . however , other cosmologists dictate a time variation of this constant in order to account for its present smallness @xcite . the variation of its constant could resolve some of the standard model problems like @xmath7 , the constant @xmath0 is a gravity coupling and both should therefore be treated on an equal footing . the generalized einstein s theory of gravitation with time dependent @xmath7 and @xmath0 has been proposed by lau @xcite . the possibility of variable @xmath7 and @xmath0 in einstein theory has also been studied by dersarkissian @xcite . this relation plays an important role in cosmology . berman @xcite and sistero @xcite have considered the einstein field equations with perfect fluid and variable @xmath7 and for robertson - walker line element . kalligas et al . @xcite have studied frw models with variable @xmath0 and @xmath7 and discussed the possible connection with power - law time dependence of @xmath7 . abdussattar and vishwakarama @xcite presented r - w models with variable @xmath0 and @xmath7 by admitting a contracted ricci collineation along the fluid flow vector . recently some of us and others have studied cosmological models with variable @xmath7 and @xmath0 in a diversified fields @xcite . thus the implication of time varying @xmath0 and @xmath7 are important to study the early evolution of the universe . in the present paper a five dimensional kaluza - klein cosmological model is considered with variable @xmath7 and @xmath0 which provides an expanding universe . it is found that the gravitational constant decreases with time as @xmath9 whereas the cosmological constant decreases as @xmath2 , @xmath10 and @xmath4 where @xmath5 is the equation of state parameter , @xmath6 is the scale factor . however , we have come across with some awkward situations in connection to the deceleration parameter and the age of the universe as far as the scenario of present accelerating universe is concerned . in the present investigation , our approach is similar to that of collins et al . @xcite for solving the einstein field equations . the paper is organized as follows : in sections 2 and 3 we present the basic field equations governing the models and their solutions . a critical discussion and conclusion are provided in section 4 . we consider five dimensional kaluze - klein space - time given by @xmath11 where @xmath12 and @xmath13 are function of the temporal coordinate @xmath14 only . the universe is assumed to be filled with distribution of matter represented by energy - momentum tensor for a perfect fluid @xmath15 where @xmath16 is the energy density of the cosmic matter and @xmath17 is its pressure , @xmath18 is the unit flow vector such that @xmath19 . we assume that the matter content obeys an equation of state @xmath20 the field equations are those of einstein but with time dependent cosmological and gravitational constants and are given by weinberg @xcite @xmath21 the spatial average scale factor @xmath22 is given by @xmath23 and average volume scale factor @xmath24 . the average hubble parameter h may be generalized in anisotropic cosmological model as @xmath25 where @xmath26 and @xmath27 and an over head dot denotes ordinary differentiation with respect to cosmic time @xmath14 . we also have @xmath28 where @xmath29 and @xmath30 are hubble s factor in the directions of @xmath31 , @xmath32 , @xmath33 and @xmath34 respectively . for the line element ( 1 ) the einstein field equations ( 4 ) yield the following equations @xmath35 @xmath36 @xmath37 if we use the equivalent energy conservation of general relativity by taking the covariant derivative of einstein field equations ( 4 ) we find that @xmath38 this provides @xcite @xmath39 as well as @xmath40 which will be satisfied by the solutions of the field equations as can be verified easily . the equations ( 3 ) , ( 8) - ( 10 ) and ( 12 ) are five independent equations with six unknowns @xmath12 , @xmath13 , @xmath16 , @xmath17 , @xmath7 and @xmath0 . hence to get a realistic solution we assume that the expansion scalar @xmath41 in the model is proportional to the shear @xmath42 @xcite . this condition leads to the relation between metric potential @xcite @xmath43 where @xmath44 is a constant . from equations ( 9 ) and ( 10 ) , we have @xmath45 where @xmath46 is a constant of integration . using equation ( 14 ) in equation ( 15 ) and then integrating it we get the lime element ( 1 ) as @xmath47^{2/(n+3)}(dx^2 + dy^2 + dz^2 ) \nonumber \\ -\left[k_1 \frac{(n+3)}{(1-n)}t + k_2(n+3)\right]^{2n/(n+3 ) } d\psi^{2}.\end{aligned}\ ] ] by suitable changes of constants the line element ( 1 ) can be written as @xmath48 where @xmath49 and @xmath50 are two positive constants . for the model ( 17 ) , the spatial volume @xmath51 , matter density @xmath16 , pressure @xmath17 , gravitational parameter @xmath7 , cosmological parameter @xmath0 are given by @xmath52 @xmath53 @xmath54 @xmath55 @xmath56 the expansion scalar @xmath41 and shear @xmath42 are @xmath57 @xmath58 where @xmath59 is a positive constant . the critical density @xmath60 , vacuum density @xmath61 and density parameter @xmath62 are given by @xmath63 @xmath64 @xmath65 keeping in mind the form of the solution ( 18 ) , i.e. @xmath66 , one can express all the above parameters as function of the scale factor @xmath6 as well which we will follow in the following two sub - cases . for the radiation epoch , we have @xmath67 . in this case @xmath68 @xmath69 @xmath70 @xmath71 @xmath72 we observe that in this phase matter density is three times the vacuum density . the cosmic density parameter @xmath62 in this phase is @xmath73 for the phase dominated by dust matters , we have @xmath74 . in this epoch @xmath75 @xmath76 @xmath77 @xmath78 @xmath79 in the present phenomenological study five dimensional kaluza - klein cosmological models with varying @xmath7 and @xmath0 are obtained . the models obtained present an expansion scalar @xmath41 bearing a constant ratio to the anisotropy in the direction of space like unit vector @xmath80 and with equation of state @xmath81 . if we inspect the solution set then it can be observed from the equation ( 18 ) that the spatial volume @xmath51 is zero at @xmath82 where @xmath83 and expansion scalar @xmath41 is infinite at @xmath84 which shows that the universe starts evolving with zero volume and infinite rate of expansion at @xmath82 . the scale factors also vanish at @xmath82 and hence the model has a _ point type _ singularity at the initial epoch . since @xmath85 is constant , as evident from the solutions ( 23 ) and ( 24 ) , the anisotropy does not die out asymptotically . the model admits a negative @xmath0 unless @xmath86 and these type of models for the universe with negative @xmath0 are available in the literature @xcite . we observe from the general solution set that for the stiff fluid ( @xmath86 ) the model reduces to standard kaluza - klein type zeldovich universe @xcite with @xmath87 , @xmath88 constant and @xmath89 for the limiting case @xmath90 ( figs . 1 & 2 ) . at @xmath91 limit all the parameters @xmath7 , @xmath0 , @xmath92 and @xmath41 are infinite in value , whereas all these become zero as @xmath93 . the ratio between cosmic matter density parameter @xmath94 and cosmic vacuum - energy density parameter @xmath95 scales as @xmath96 whereas cosmic vacuum - energy density parameter can be given as @xmath97 the above two expressions related to cosmic matter and vacuum density parameters do vanish for the stiff fluid condition @xmath86 . if we look at the solution expressed in the equation ( 21 ) then we can observe that in general @xmath9 . thus for the dust case ( @xmath74 ) this case is similar to the result of dirac @xcite , canuto et al . @xcite and beesham @xcite . on the other hand , equations ( 22 ) and ( 23 ) at once imply that @xmath98 for any constant values of @xmath5 and @xmath99 . again , from the solutions ( 18 ) and ( 22 ) one can find @xmath100 . on the other hand for the limiting case @xmath101 the equation ( 22 ) yields the relation @xmath2 . thus we obtain @xmath102 , @xmath100 and @xmath2 which is in accordance with the main dynamical laws one finds in literature proposed for the decay of @xmath0 . note that we find all these phenomenological relations for @xmath0 directly from the solution set which generally consider as ad hoc basis in cosmological research . the dynamical law @xmath103 , has been proposed by carvalho et al . @xcite and considered by salim and waga @xcite , arbab and abdel - rahman @xcite , wetterich @xcite , arbab @xcite and very recently by ray and his collaborators as well as pradhan and his collaborators @xcite . the decay law of the form @xmath4 has been considered by arbab @xcite , khadekar et al . @xcite , ray et al . @xcite , ray and mukhopadhyay @xcite and ray et al . @xcite . the dynamical law @xmath2 has been considered by several authors , e.g. bertolami @xcite , berman and som @xcite , berman @xcite , beesham @xcite , pradhan et al . @xcite to mention a few . we observe that @xmath104 decays faster than @xmath7 where as @xmath17 , @xmath16 , @xmath105 , @xmath60 scales @xmath106 as can be seen from the figs . 1 - 4 . in the model we see that the quantity @xmath7 satisfies the condition for a machian cosmological solution i.e. @xmath107 @xcite . however , as far as the scenario of present accelerating universe is concern our models seems suffer a lot in connection to the different physical parameters , such as the deceleration parameter and the age of the universe . the deceleration parameter @xmath108 for the present model is @xmath109 and hence does not correspond to the observational result @xmath110 @xcite . however , we have observed that for the scale factor of the form @xmath111 we can have a negative deceleration parameter when @xmath112 . in our present case @xmath113 which yields @xmath114 . here the age of the universe is given by @xmath115 which is also differs from the present estimate @xmath116 gyr @xcite . this value , therefore , obviously suffers from the low age problem @xcite as even the oldest globular clusters show a value which is @xmath117 gyr @xcite . in this connection it is argued that dark energy models without @xmath0 suffer from low age problem @xcite though even in the presence of @xmath0 we are not in a hopeful position . one of the reasons of these discrepancies may be due to the constant nature of the equation of state parameter @xmath5 which can not accommodate with the present accelerating phase of the universe . however , this needs further investigation . the authors ( fr & sr ) is thankful to the authority of inter - university centre for astronomy and astrophysics , pune , india for providing visiting associateship under which a part of this work was carried out . we all are thankful to the referee for valuable suggestions which have enabled us to improve the manuscript substantially .	a five dimensional kaluza - klein space - time is considered in the presence of a perfect fluid source with variable g and @xmath0 . an expanding universe is found by using a relation between the metric potential and an equation of state . the gravitational constant is found to decrease with time as @xmath1 whereas the variation for the cosmological constant follows as @xmath2 , @xmath3 and @xmath4 where @xmath5 is the equation of state parameter and @xmath6 is the scale factor . . . variable @xmath7 and @xmath0 , kaluza - klein model , phenomenological cosmology . 04.20.-q , 04.20.jb , 98.80.jk
for a banach space @xmath11 , we denote its norm by @xmath12 . a map @xmath0 from one banach space @xmath13 to another banach space @xmath14 is said to have _ bi - lipschitz property _ if there exist two positive constants @xmath15 and @xmath16 such that @xmath17 our models of bi - lipschitz maps between banach spaces are analysis operators associated with hilbert frames , @xmath1-frames , banach frames , @xmath18-frames and fusion frames @xcite . our study is also motivated by nonlinear sampling theory and phase retrieval , which have gained substantial attention in recent years @xcite . the framework developed in the first part of this paper could be considered as a nonlinear extension of frame theory . denote by @xmath19 the banach space of all bounded linear operators from one banach space @xmath13 to another banach space @xmath14 . a continuous map @xmath0 from @xmath13 to @xmath14 is said to be _ differentiable at @xmath20 _ if there exists a linear operator , denoted by @xmath21 , in @xmath19 such that @xmath22 and to be _ differentiable on @xmath13 _ if it is differentiable at every @xmath20 @xcite . for a differentiable bi - lipschitz map @xmath0 from @xmath13 to @xmath14 , one may easily verify that its derivatives @xmath23 , are _ uniformly stable _ , i.e. , there exist two positive constants @xmath15 and @xmath16 such that @xmath24 the converse is not true in general . then we have the following natural question . * question 1 * : _ when does a differentiable map with the uniform stability property have the bi - lipschitz property ? _ we say that a linear operator @xmath25 from one banach space @xmath13 to another banach space @xmath14 is _ bounded below _ if @xmath26 for a continuously differentiable map @xmath0 not too nonlinear , particularly not far away from a bounded below linear operator @xmath10 , a sufficient condition for is that for any @xmath27 , the set @xmath28 of unit vectors @xmath29 , is contained in a ball of radius @xmath30 with center at @xmath31 , where @xmath32 the above geometric requirement on the radius @xmath33 is optimal in banach space setting , but it could be relaxed to @xmath34 in hilbert space setting , which implies that for any @xmath27 , the set @xmath28 is contained in a right circular cone with axis @xmath35 and angle strictly less than @xmath36 . detailed arguments of the above conclusions on a differentiable map are given in appendix [ bilipschitz.appendix ] . denote by @xmath37 the image of a map @xmath0 from one banach space @xmath13 to another banach space @xmath14 . for a bi - lipschitz map @xmath38 , as it is one - to - one , for any @xmath39 there exists a unique @xmath20 such that @xmath40 . our next question is as follows : * question 2 * : _ given noisy observation @xmath41 of @xmath42 corrupted by @xmath43 , how to construct a suboptimal approximation @xmath20 such that @xmath44 where @xmath45 is an absolute constant independent of @xmath42 and @xmath43 ? _ for a differentiable bi - lipschitz map @xmath0 not far away from a bounded below linear operator @xmath10 , define @xmath46 , iteratively with arbitrary initial @xmath47 by @xmath48 where @xmath49 is a bounded left - inverse of the linear operator @xmath10 , and the relaxation factor @xmath50 satisfies @xmath51 . in theorem [ banachalgorithm.thm ] of section [ algorithm.section ] , we show that the sequence @xmath46 , in the iterative algorithm converges exponentially to a suboptimal approximation element @xmath20 satisfying , provided that @xmath52 the above requirement about @xmath33 to guarantee convergence of the iterative algorithm is stronger than the sufficient condition for the bi - lipschitz property of the map @xmath0 . in theorem [ hilbertalgorithm.thm ] , we close that requirement gap on @xmath33 in hilbert space setting by introducing an iterative algorithm of van - cittert type , @xmath53 where @xmath54 is the conjugate of the linear operator @xmath10 and @xmath55 is a small relaxation factor . in the iterative algorithm , a left - inverse @xmath49 of the bounded below linear operator @xmath10 is used , but its existence is not always assured in banach space setting and its construction is not necessarily attainable even it exists . this limits applicability of the iterative reconstruction algorithm . in fact , for general banach space setting , a stable reconstruction algorithm may not exist @xcite . on the other hand , a stable iterative algorithm is proposed in @xcite to find sub - optimal approximation for well - localized nonlinear maps on sequence spaces @xmath56 . so we have the following question . * question 3 * : _ for what types of banach spaces @xmath13 and @xmath14 and nonlinear maps @xmath0 from @xmath13 to @xmath14 does there exist a stable reconstruction of @xmath42 from its nonlinear observation @xmath57 ? _ we say that a banach space @xmath11 with norm @xmath12 is _ hilbert - dense _ ( respectively _ weak - hilbert - dense _ ) if there exists a hilbert subspace @xmath58 with norm @xmath59 such that @xmath60 is dense in @xmath11 in the strong topology ( respectively in the weak topology ) of @xmath11 and @xmath61 our models of the above new concepts are the sequence spaces @xmath62 , for which @xmath63 with @xmath64 are hilbert - dense and @xmath65 is weak - hilbert - dense . for ( weak-)hilbert - dense banach spaces @xmath13 and @xmath14 and a nonlinear map @xmath66 that has certain localization property , a stable reconstruction of @xmath42 from its nonlinear observation @xmath57 is proposed in theorem [ banachwiener.algorithm.thm ] of section [ localizedmaps.section ] . the crucial step is a new fixed point theorem for a well - localized differentiable map whose restriction on a dense hilbert subspace is a contraction , see theorem [ fixedpoint.thm ] . let @xmath67 and @xmath68 be hilbert spaces , and let @xmath69 be union of closed linear subspaces @xmath70 , of the hilbert space @xmath67 . the second topic of this paper is to study the _ restricted bi - lipschitz property _ of a map @xmath71 on @xmath9 , which means that there exist two positive constants @xmath15 and @xmath16 such that @xmath72 this topic is motivated by sparse recovery problems on finite - dimensional spaces @xcite . as we use the union @xmath4 of closed linear spaces @xmath70 , to model sparse signals , the restricted bi - lipschitz property of a map @xmath0 could be thought as nonlinear correspondence of restricted isometric property of a measurement matrix . so the framework developed in the second part is nonlinear banach space extension of the finite - dimensional sparse recovery problems . in the classical sparse recovery setting @xcite , the set of all @xmath73-sparse signals for some @xmath74 is used as the set @xmath9 . in this case , elements in @xmath9 can be described by their @xmath75-quasi - norms being less than or equal to @xmath73 , and the sparse recovery problem could reduce to the @xmath75-minimization problem . due to numerical infeasibility of the @xmath75-minimization , a relaxation to ( non-)convex @xmath76-minimization with @xmath77 was proposed , and more importantly it was proved that the @xmath76-minimization recovers sparse signals when the linear measurement operator has certain restricted isometry property in @xmath78 @xcite . this leads to the following question . * question 4 * : _ how to create a general optimization framework to recover sparse signals ? _ given a banach space @xmath79 , we say that a subset @xmath80 of @xmath81 is _ proximinal _ ( @xcite ) if every element @xmath82 has a best approximator @xmath83 , that is , @xmath84 given hilbert spaces @xmath67 and @xmath68 , a union @xmath69 of closed linear subspaces @xmath70 , of @xmath67 , and a continuous map @xmath0 from @xmath67 to @xmath68 , consider the following minimization problem in a banach space @xmath79 , @xmath85 for any given observation @xmath86 for some @xmath87 . to make the above minimization problem suitable for stable reconstruction of @xmath87 from its observation @xmath88 , we introduce the concept of a _ sparse approximation triple _ @xmath89 : * ( continuous imbedding property ) the banach space @xmath79 contains all elements in @xmath9 and it is contained in the hilbert space @xmath67 , that is , @xmath90 and the imbedding operators @xmath91 and @xmath92 are bounded . * ( proximinality property ) the banach space @xmath79 has @xmath4 as its closed subset , and all closed subsets of @xmath81 being proximinal . * ( common - best - approximator property ) given any @xmath93 , a best approximator @xmath94 of @xmath95 in the norm @xmath96 is also a best approximator in the norm @xmath97 , that is , @xmath98 * ( norm - splitting property ) for the best approximator @xmath99 of @xmath95 in the norm @xmath96 , @xmath100 and @xmath101 * ( sparse density property ) @xmath102 is dense in @xmath103 , where @xmath104 one may easily verify that these five properties are satisfied for the triple @xmath89 in the classical sparse recovery setting , where @xmath9 is the set of all @xmath73-sparse vectors , @xmath79 is the set of all summable sequences , and @xmath67 is the set of all square - summable sequences @xcite . in this paper , we rescale the norm @xmath96 in the sparse approximation triple @xmath105 so that the imbedding operator @xmath106 has norm one , @xmath107 otherwise replacing it by @xmath108 . next we introduce two quantities to measure _ sparsity _ @xmath109 for signals in @xmath9 , and _ sparse approximation ratio _ @xmath110 for elements in @xmath79 , where @xmath111 and @xmath112 are the first and second best approximators of @xmath2 respectively , @xmath113 the upper bound estimate in holds , since @xmath114 by the norm splitting property . in the classical sparse recovery setting with @xmath9 being the set of all @xmath73-sparse signals , one may verify that @xmath115 and @xmath116 , see appendix [ sat.appendix ] for additional properties of sparse approximation triples . having introduced the sparse approximation triple @xmath89 , our next question is on optimization approach to sparse signal recovery , see @xcite for the classical setting . * question 5 * : _ given a sparse approximation triple @xmath105 , for what type of maps @xmath71 does the solution @xmath117 of the optimization problem @xmath118 is a suboptimal approximation to the sparse signal @xmath119 in @xmath9 ? _ in this paper , without loss of generality , we assume that @xmath120 . we say that @xmath71 has _ sparse riesz property _ if @xmath121 with @xmath122 , and that @xmath0 is _ almost linear on @xmath4 _ if @xmath123 with @xmath124 . combining the sparse riesz property and almost linear property of a map @xmath0 gives @xmath125 and hence @xmath0 has the restricted bi - lipschitz property on @xmath4 , @xmath126 when @xmath127 and @xmath128 satisfy @xmath129 . in section [ optimization.section ] , we show that the solution @xmath117 of the optimization problem is a suboptimal approximation to the signal @xmath119 in @xmath79 , i.e. , there exist positive constants @xmath130 and @xmath131 such that @xmath132 provided that @xmath0 has the sparse riesz property and almost linear property with @xmath133 and @xmath134 satisfying @xmath135 we remark that the approximation error estimate implies the sparse riesz property for the map @xmath0 , @xmath136 which follows from by taking @xmath137 and @xmath138 . the sparse riesz property was introduced in @xcite with a different name , sparse approximation property , for the classical sparse recovery setting ; and the almost linear property was studied in @xcite for bi - lipschitz maps between banach spaces . in section [ sparserieszproperty.section ] , we consider the following question . * question 6 * : _ when does a map @xmath71 have the sparse riesz property and the almost linear property ? _ we say that a linear operator @xmath139 has the _ restricted isometry property _ ( rip ) on @xmath140 if @xmath141 where @xmath142 @xcite . a nonlinear map @xmath143 , not far away from a linear operator @xmath10 with the restricted isometry property in the sense that @xmath144 has the restricted bi - lipschitz property on @xmath9 , where @xmath145 in section [ sparserieszproperty.section ] , we show that @xmath0 has the sparse riesz property when @xmath146 and the almost linear property when @xmath147 . therefore signals @xmath82 could be reconstructed from their nonlinear measurements @xmath88 when @xmath0 is not far from a linear operator @xmath10 with small restricted isometry constant @xmath148 , see theorem [ optimization.thm2 ] . for a banach / hilbert space @xmath11 , we also denote its norm by @xmath149 for brevity . in this section , we establish exponential convergence of the iterative reconstruction algorithms and . [ banachalgorithm.thm ] let @xmath13 and @xmath14 be banach spaces , @xmath0 be a differentiable map from @xmath13 to @xmath14 with its derivative being continuous and uniformly stable , and let @xmath25 be bounded below . assume that holds for some bounded left - inverse @xmath150 of the linear operator @xmath10 . given positive relaxation factor @xmath55 , an initial @xmath47 and a noisy observation data @xmath151 for some @xmath42 with additive noise @xmath43 , define @xmath152 , iteratively by . then @xmath46 , converges exponentially to some @xmath153 with @xmath154 provided that @xmath155 moreover , @xmath156 where @xmath157 set @xmath158 . then for @xmath159 , @xmath160 by , and . this proves the exponential convergence of @xmath46 , to its limit @xmath161 . taking limit in gives @xmath162 because @xmath163 then it follows from , and that @xmath164 which proves . observe that @xmath165 then the estimate follows . the iterative algorithm in theorem [ banachalgorithm.thm ] provides a stable reconstruction of @xmath20 from its noisy observation @xmath166 when @xmath167 , a requirement stronger than @xmath168 that guarantees the bi - lipschitz property for the map @xmath0 , see theorem [ frame.sufficientcondition ] in appendix [ bilipschitz.appendix ] . next we close that requirement gap on @xmath33 in hilbert space setting , cf . theorem [ frame.sufficientcondition.hilbert ] . [ hilbertalgorithm.thm ] let @xmath67 and @xmath68 be hilbert spaces , @xmath0 be a differentiable map from @xmath67 to @xmath68 with its derivative being continuous and satisfying , and let @xmath139 satisfy and . given relaxation factor @xmath169 , an initial @xmath170 , and noisy data @xmath171 for some @xmath172 with additive noise @xmath173 , define @xmath174 , iteratively by . then @xmath175 , converges exponentially to some @xmath176 with @xmath177 provided that @xmath178 define @xmath179 . observe that @xmath180 therefore @xmath181 where @xmath15 is the lower stability bound in . also one may easily verify that @xmath182 therefore by standard arguments ( see for instance @xcite ) , we obtain from and that @xmath183 where @xmath184 this proves the exponential convergence of @xmath175 , in the iterative algorithm . taking limit in the algorithm leads to @xmath185 where @xmath186 is the limit of the sequence @xmath175 . thus @xmath187 by and . this proves and completes the proof . in this section , we develop a fixed point theorem for a well - localized map on a banach space whose restriction on its dense hilbert subspace is a contraction , and we establish exponential convergence of the iterative algorithm for certain localized maps between ( weak-)hilbert - dense banach spaces . to state our results , we recall the concept of differential subalgebras . given two unital banach algebras @xmath188 and @xmath189 , @xmath188 is said to be a _ banach subalgebra _ of @xmath189 if @xmath190 , @xmath188 and @xmath189 share the same identity and @xmath191 holds ; and a banach subalgebra @xmath188 of @xmath189 is said to be a _ differential subalgebra _ of order @xmath192 $ ] if there exists a positive constant @xmath128 such that @xmath193 for all nonzero @xmath194 @xcite . we remark that differential subalgebras include many families of banach algebras of infinite matrices with certain off - diagonal decay and localized integral operators @xcite , and they have been widely used in operator theory , non - commutative geometry , frame theory , algebra of pseudodifferential operators , numerical analysis , signal processing , control and optimization etc , see @xcite , the survey papers @xcite and references therein . next we define the conjugate @xmath54 of a localized linear operator @xmath10 between ( weak- ) hilbert - dense banach spaces . given banach spaces @xmath195 and their dense hilbert subspaces @xmath196 , we assume that linear operators @xmath10 reside in a banach subspace @xmath197 of @xmath198 and also of @xmath19 . for a linear operator @xmath199 , its restriction @xmath200 to @xmath67 is a bounded operator from @xmath67 to @xmath68 , hence its conjugate @xmath201 is well - defined on @xmath68 , and the conjugate " of @xmath10 is well - defined if the conjugate @xmath201 can be extended to a bounded operator from @xmath14 to @xmath13 . the above approach to define the conjugate requires certain localization for linear operators in @xmath197 , which will be stated precisely in the next theorem , cf . . [ banachwiener.algorithm.thm ] let @xmath13 and @xmath14 be banach spaces , @xmath67 and @xmath68 be hilbert spaces with the property that for @xmath202 , @xmath203 , @xmath204 is dense in @xmath195 , and @xmath205 assume that banach algebra @xmath206 with norm @xmath207 is a unital banach subalgebra of @xmath208 and a differential subalgebra of @xmath209 of order @xmath210 $ ] . let @xmath197 be a banach subspace of both @xmath198 and @xmath19 , and @xmath211 be a banach subspace of both @xmath212 and @xmath213 such that * @xmath214 for all @xmath215 and @xmath199 . moreover , @xmath216 * for any @xmath199 and @xmath215 there exist unique @xmath217 and @xmath218 with the property that @xmath219 and @xmath220 assume that @xmath0 is a differentiable map from @xmath13 to @xmath14 such that its derivative @xmath221 is continuous and bounded from @xmath13 into @xmath197 , and @xmath222 for some linear operator @xmath199 . take an initial @xmath47 and a noisy observation data @xmath41 for some @xmath42 and @xmath43 , define @xmath152 , iteratively by @xmath223 where the relaxation factor @xmath50 satisfies @xmath224 and @xmath217 is the conjugate operator defined by . then @xmath46 , converges exponentially to some @xmath153 with @xmath225 where @xmath45 is an absolute positive constant . given a banach space @xmath11 , we say that a map @xmath226 is a _ contraction _ if there exists @xmath227 such that @xmath228 for a contraction @xmath229 on a banach space @xmath11 , the banach fixed point theorem states that there is a unique fixed point @xmath230 for the contraction @xmath229 ( i.e. , @xmath231 ) , and for any initial @xmath232 , the sequence @xmath233 , converges exponentially to the fixed point @xmath230 @xcite . to prove theorem [ banachwiener.algorithm.thm ] , we need a fixed point theorem for differentiable maps on a banach space with its derivative being continuous and bounded in a differential banach subalgebra and its restriction on a dense hilbert subspace being a contraction . [ fixedpoint.thm ] let @xmath11 be a banach space , @xmath60 be a hilbert space such that @xmath58 is dense in @xmath11 and @xmath234 and let @xmath206 be a banach subalgebra of @xmath235 and also a differential subalgebra of @xmath236 of order @xmath192 $ ] . if @xmath229 is a differentiable map on @xmath11 whose derivative @xmath237 is continuous and bounded from @xmath11 into @xmath206 and there exists @xmath227 such that @xmath238 then there exists a unique fixed point @xmath230 for the map @xmath229 . furthermore given any initial @xmath232 , the sequence @xmath46 , defined by @xmath239 converges exponentially to the fixed point @xmath230 . let @xmath46 , be as in . it follows from the continuity of @xmath237 in the banach subalgebra @xmath206 of @xmath235 that @xmath240 observe that @xmath241 and @xmath242 where @xmath243 by the assumption on the map @xmath229 . set @xmath244 then we obtain from , and that @xmath245 and @xmath246 for all @xmath159 . thus @xmath247 where @xmath248 with @xmath249 and @xmath250 . therefore @xmath251 if @xmath252 , and @xmath253 if @xmath254 . by and , for any @xmath255 there exists a positive constant @xmath45 such that @xmath256 recall that @xmath206 is a banach subalgebra of @xmath235 . we then obtain from and that @xmath257 which proves the exponential convergence of the sequence @xmath46 . finally we prove the uniqueness of the fixed point for the map @xmath229 . let @xmath230 and @xmath258 be fixed points of the map @xmath229 . then @xmath259 is a fixed point of the linear operator @xmath260 , because @xmath261 following the argument to prove , we obtain that @xmath262 . this together with implies that @xmath263 , the uniqueness of fixed points for the map @xmath229 . now we apply theorem [ fixedpoint.thm ] to prove theorem [ banachwiener.algorithm.thm ] . [ proof of theorem [ banachwiener.algorithm.thm ] ] define @xmath264 by @xmath265 then @xmath229 is differentiable on @xmath13 and its derivative @xmath266 , is continuous and bounded in @xmath206 by the assumption on @xmath0 and the banach spaces @xmath197 and @xmath211 . set @xmath267 and @xmath268 observe that @xmath269 and @xmath270 where the second inequality holds as @xmath271 and the third inequality follows from . combining the above two estimates about @xmath272 , with theorem [ fixedpoint.thm ] proves the exponential convergence of @xmath46 , in @xmath13 . denote by @xmath273 the limit of @xmath46 , in @xmath13 . then taking limit in the iterative algorithm yields @xmath274 thus @xmath275 where @xmath276 . following the argument to prove theorem [ fixedpoint.thm ] and applying and , there exists a positive constant @xmath277 for any @xmath278 such that @xmath279 thus @xmath280 is invertible in @xmath206 and @xmath281 combining and leads to @xmath282 this proves the error estimate . our model of hilbert - dense banach spaces in theorem [ banachwiener.algorithm.thm ] is @xmath283 , the space of @xmath1-summable sequences @xmath283 , with @xmath64 . for that case , exponential convergence of the van - cittert algorithm , which is similar to the iterative algorithm in theorem [ banachwiener.algorithm.thm ] , is established in @xcite under slightly different restriction on the relaxation factor @xmath50 . for weak - hilbert - dense banach spaces , the iterative algorithm in theorem [ banachwiener.algorithm.thm ] still has exponential convergence if operators in @xmath197 and @xmath211 are assumed additionally to be uniformly continuous in the weak topologies of banach spaces , that is , @xmath284 for any bounded linear functional @xmath285 on @xmath14 if @xmath286 tends to @xmath273 in the weak topology of @xmath13 ; and @xmath287 for any bounded linear functional @xmath18 on @xmath13 if @xmath288 tends to @xmath289 in the weak topology of @xmath14 . we leave the detailed arguments to interested readers . in this section , we show that sparse signals @xmath290 could be reconstructed from their nonlinear measurements @xmath88 via the optimization approach . [ optimization.thm1 ] let @xmath67 and @xmath68 be hilbert spaces , @xmath79 be a banach space , @xmath69 be union of closed linear subspaces of @xmath67 , @xmath291 and @xmath292 be in and respectively , and let @xmath0 be a continuous map from @xmath67 to @xmath68 normalized so that @xmath120 . if @xmath89 forms a sparse approximation triple , and if @xmath0 has the sparse riesz property and the almost linear property with @xmath293 satisfying @xmath294 then given @xmath295 and @xmath296 , the solution @xmath117 of the optimization problem provides a suboptimal approximation to @xmath119 , @xmath297 and @xmath298 to prove theorem [ optimization.thm1 ] , we need the following approximation property for sparse approximation triples . [ aa.prop ] let @xmath105 be a sparse approximation triple and @xmath299 be as in . then @xmath300 where @xmath301 is a best approximator of @xmath82 . we postpone the proof of the above proposition to appendix [ sat.appendix ] and start the proof of theorem [ optimization.thm1 ] . let @xmath302 be a best approximator in @xmath9 to @xmath119 , where the existence follows from the proximinality property of the triple @xmath105 . denote by @xmath303 the linear space in @xmath9 containing @xmath304 . then @xmath305 by the common best approximator property ; and @xmath306 by the norm splitting properties and . let @xmath307 be a best approximator to @xmath117 in @xmath303 . then @xmath308 and @xmath309 by the common best approximator property and norm splitting property of the triple @xmath105 . set @xmath310 . we then obtain from , and that @xmath311 let @xmath312 be a best approximator of @xmath313 . then @xmath314 @xmath315 and @xmath316 by , and . from , , and , it follows that @xmath317 by the definition of @xmath318 , we have that @xmath319 where @xmath320 is the projection operator from @xmath67 to its closed subspace @xmath321 . therefore @xmath322 by , , and we get @xmath323 and @xmath324 hence by , , , and , we have @xmath325 and @xmath326 combining and and using @xmath327 lead to @xmath328 and @xmath329 on the other hand , @xmath330 by , and . therefore the error estimates and follow from , , , and . as a corollary , we have the following result for linear mapping @xmath0 , cf . * theorem 1.1 ) in the classical sparse recovery setting . [ optimization.cor1 ] let @xmath331 be as in theorem [ optimization.thm1 ] , and let @xmath143 be linear and have the sparse riesz property with @xmath332 and @xmath333 . given @xmath295 and @xmath296 , the optimization solution of satisfies @xmath334 and @xmath335 where @xmath336 . in this section , we consider the sparse riesz property and almost linear property for nonlinear maps not far from a linear operator with the restricted isometry property . [ sparsereconstruction.optimization.thm ] let @xmath67 and @xmath68 be hilbert spaces , @xmath79 be a banach space , and @xmath69 be a union of closed linear subspaces of @xmath67 . assume that @xmath89 is a sparse approximation triple , and @xmath139 has the restricted isometry property on @xmath140 with @xmath337 . if @xmath0 is a continuous map from @xmath67 to @xmath68 with @xmath120 and @xmath338 then @xmath0 has the sparse riesz property , @xmath339 for any @xmath95 , define @xmath340 with initial @xmath341 . to prove theorem [ sparsereconstruction.optimization.thm ] , we need convergence of the above greedy algorithm . [ greedy.prop ] let @xmath89 be a sparse approximation triple . then @xmath342 , in the greedy algorithm converges to @xmath95 , @xmath343 we postpone the proof of the above proposition to appendix b and start the proof of theorem [ sparsereconstruction.optimization.thm ] . take @xmath95 , let @xmath342 , be as in the greedy algorithm . then from proposition [ greedy.prop ] , the continuity of @xmath0 on @xmath67 , and the continuous imbedding of @xmath79 into @xmath67 it follows that @xmath344 write @xmath345 . then @xmath346 , and @xmath347 by , , and the assumption @xmath120 . observe that @xmath348 , where @xmath349 . then @xmath350 by the restricted isometry property . we remark that in the classical sparse recovery setting , the inner product @xmath351 between different @xmath352 and @xmath353 is always zero , but it may be nonzero in our setting . hence for @xmath354 , @xmath355 and similarly @xmath356 therefore combining and , and then applying and when taking limit as @xmath357 , we obtain @xmath358 which implies that @xmath359 by , @xmath360 this together with and implies that @xmath361 combining , and gives @xmath362 reformulating the above estimates completes the proof of the estimate for the sparse riesz property of @xmath0 . [ almostlinear.thm ] let @xmath363 be as in theorem [ sparsereconstruction.optimization.thm ] with additional assumption that @xmath364 . then @xmath0 has the almost linear property on @xmath4 , @xmath365 take @xmath366 , and let @xmath367 and @xmath368 , be as in the greedy algorithm to approximate @xmath2 and @xmath369 respectively . write @xmath370 by , , , and the continuity of @xmath0 and @xmath10 on @xmath67 , we get @xmath371 and similarly @xmath372 for the term @xmath373 , we obtain from and that @xmath374 combining estimates in gives @xmath375 write @xmath376 following the arguments used to establish , we have @xmath377 combining and proves the estimate for the almost linear property of the map @xmath0 . combining theorems [ optimization.thm1 ] , [ sparsereconstruction.optimization.thm ] and [ almostlinear.thm ] leads to the following result on the stable reconstruction of sparse signals @xmath2 from their nonlinear measurements @xmath88 when @xmath0 is not far away from a measurement matrix @xmath10 with the restricted isometry property . [ optimization.thm2 ] let @xmath378 be as in theorem [ sparsereconstruction.optimization.thm ] with @xmath379 then for any given @xmath295 and @xmath296 , the solution @xmath117 of the minimization problem has the following error estimates : @xmath380 and @xmath381 where @xmath130 and @xmath131 are absolute constants independent on @xmath382 and @xmath383 . applying theorem [ optimization.thm2 ] to linear maps , we have the following corollary . [ optimization.thm2.cor ] let @xmath384 and @xmath10 be as in theorem [ optimization.thm2 ] . if @xmath385 then for any given @xmath295 and @xmath296 , the solution @xmath117 of the minimization problem with @xmath386 has the error estimates and . for classical sparse recovery problems , the conclusions in corollary [ optimization.thm2.cor ] have been established under weaker assumptions on the restricted isometry constant @xmath387 , see @xcite and references therein . the first author thanks professor yuesheng xu for his invitation to visit guangdong province key laboratory of computational science at sun yat - sen university , china , where part of this work was done . in this appendix , we provide some sufficient conditions , mostly optimal , for a differentiable map to have the bi - lipschitz property , see theorems [ frame.sufficientcondition ] and [ frame.sufficientcondition.linear ] in banach space setting , and theorems [ frame.sufficientcondition.hilbert ] and [ frame.sufficientcondition.hilberts2 ] in hilbert space setting . for a differentiable map @xmath0 from one banach space @xmath13 to another banach space @xmath14 that has the bi - lipschitz property , we have @xmath388 where @xmath389 are the constants in the bi - lipschitz property . then taking limit as @xmath390 leads to a necessary condition for a differentiable bi - lipschitz map . [ bilipschitz.thm ] let @xmath13 and @xmath14 be banach spaces . if @xmath66 is a differentiable map that has the bi - lipschitz property , then its derivative @xmath391 , has the uniform stability property . for @xmath392 , a differentiable map @xmath0 with the uniform stability property for its derivative has the bi - lipschitz property , but it is not true in general banach space setting . maps @xmath393 , from @xmath394 to @xmath395 in the example below are such examples . [ epepsilon.example ] for @xmath396 and @xmath397 , define @xmath398 by @xmath399,\\ ( \cos \epsilon , \sin \epsilon)+ ( -\sin \epsilon , \cos \epsilon)(t-\pi/2-\epsilon)\\ \hfill { \rm if } \ t\in ( \pi/2+\epsilon , \infty ) , \end{array}\right.\ ] ] see figure [ functionepepsilon.fig ] . [ cols="^,^ " , ] the maps @xmath400 just defined do not have the bi - lipschitz property , but their derivatives @xmath401 have the uniform stability property , @xmath402 where @xmath403 , is the @xmath1-norm on the euclidean space @xmath395 . given a differentiable bi - lipschitz map @xmath0 from one banach space @xmath13 to another banach space @xmath14 such that its derivative @xmath21 is uniformly stable , define @xmath404 the quantity @xmath405 is the minimal radius such that for any @xmath27 , the set @xmath28 of unit vectors @xmath406 , is contained in a ball of radius @xmath407 centered at a unit vector . our next theorem shows that a differentiable bi - lipschitz map @xmath0 with its derivative @xmath21 being uniformly stable and continuous and with @xmath405 in satisfying @xmath407 has the bi - lipschitz property . [ frame.sufficientcondition ] let @xmath13 and @xmath14 be banach spaces , and @xmath0 be a continuously differentiable map from @xmath13 to @xmath14 with the property that its derivative has the uniform stability property . if @xmath405 in satisfies @xmath408 then @xmath0 is a bi - lipschitz map . given @xmath409 with @xmath410 , @xmath411 where @xmath412 with @xmath413 . thus @xmath414 and @xmath415 where @xmath416 are lower and upper stability bounds in the uniform stability property . combining the above two estimates completes the proof . [ alphaf.rem ] the u - shaped map @xmath400 in example [ epepsilon.example ] with @xmath417 and @xmath418 is not a bi - lipschitz map and @xmath419 this indicates that the geometric condition about @xmath405 is optimal . for a differentiable map @xmath0 not far away from a bounded below linear operator @xmath10 , we suggest using @xmath420 as the center of the ball containing the set of unit vectors @xmath421 , and define the minimal radius of that ball by @xmath33 in . then obviously @xmath422 this together with theorem [ frame.sufficientcondition ] implies that a differentiable map @xmath0 satisfying @xmath423 is a bi - lipschitz map . [ frame.sufficientcondition.linear ] let @xmath13 and @xmath14 be banach spaces , and @xmath0 be a continuously differentiable map from @xmath13 to @xmath14 with its derivative having the uniform stability property . if @xmath25 is bounded below and satisfies , then @xmath0 is a bi - lipschitz map . we may use the following quantity to measure the distance between differentiable map @xmath0 and bounded below linear operator @xmath10 , @xmath424 by direct computation , @xmath425 thus the geometric condition in theorem [ frame.sufficientcondition.linear ] can be replaced by the condition @xmath426 . [ linearapproximationcondition.cor ] let @xmath427 and @xmath10 be as in theorem [ frame.sufficientcondition.linear ] . if @xmath426 , then @xmath0 is a bi - lipschitz map . the geometric condition to guarantee the bi - lipschitz property for the map @xmath0 is optimal in general banach space setting , as @xmath428 for the u - shaped map @xmath429 in example [ epepsilon.example ] and the linear operator @xmath430 . but in hilbert space setting , as shown in the next theorem , the geometric condition could be relaxed to @xmath431 . [ frame.sufficientcondition.hilbert ] let @xmath67 and @xmath68 be hilbert spaces , and let @xmath432 be a continuously differentiable map with its derivative having the uniform stability property . if there exists a linear operator @xmath139 satisfying and , then @xmath0 is a bi - lipschitz map . take @xmath433 with @xmath434 . then @xmath435 where @xmath16 is the upper stability bound in . observe that @xmath436 then @xmath437 which implies that @xmath438 where @xmath15 is the lower stability bound in . hence @xmath439 combining and proves the bi - lipschitz property for @xmath0 . the geometric condition is optimal as for the u - shaped map @xmath400 in example [ epepsilon.example ] with @xmath440 and @xmath418 , @xmath441 where @xmath442 . define @xmath443 the maximal angle between vectors @xmath444 and @xmath445 in the hilbert space @xmath68 . then @xmath446 so the geometric condition can be interpreted as that the angles between @xmath444 and @xmath445 are less than or equal to @xmath447 for all @xmath433 . the above equivalence between the geometric condition and the angle condition @xmath448 , together with and , implies the existence of positive constants @xmath449 such that @xmath450 the converse can be proved to be true too . thus @xmath451 if and only if @xmath179 is strictly monotonic . here a bounded map @xmath452 on a hilbert space @xmath60 is said to be _ strictly monotonic _ @xcite if there exist positive constants @xmath453 and @xmath454 such that @xmath455 as an application of the above equivalence , theorem [ frame.sufficientcondition.hilbert ] can be reformulated as follows . [ frame.sufficientcondition.hilberts2 ] let @xmath67 and @xmath68 be hilbert spaces , and let @xmath432 be a continuously differentiable map with its derivative having the uniform stability property . if there exists a linear operator @xmath139 satisfying and , then @xmath0 is a bi - lipschitz map . from theorem [ frame.sufficientcondition.hilberts2 ] we obtain the following result similar to the one in corollary [ linearapproximationcondition.cor ] . [ linearapproximationcondition.cor2 ] let @xmath456 and @xmath0 be as in theorem [ frame.sufficientcondition.hilberts2 ] . if there exists a bounded below linear operator @xmath139 with @xmath457 , then @xmath0 is a bi - lipschitz map . given a differentiable map @xmath0 , it is quite technical in general to construct linear operator @xmath10 satisfying and in banach space setting ( respectively and in hilbert space setting ) . a conventional selection is that @xmath458 for some @xmath47 , but such a selection is not always favorable . let @xmath459 be impulse response vector with its entry @xmath460 being the impulse response of the signal generating device at the innovation position @xmath461 , and @xmath462 be sampling functional vector with entry @xmath463 reflecting the characteristics of the acquisition device at the sampling position @xmath464 . in order to consider bi - lipschitz property of the nonlinear sampling map @xmath465 related to instantaneous companding @xmath466 , a linear operator @xmath467 satisfying and is implicitly introduced in @xcite , where @xmath468 is the inter - correction matrix between @xmath469 and @xmath470 . in this appendix , we prove propositions [ greedy.prop ] and [ aa.prop ] , and conclude it with a remark on the greedy algorithm . the convergence of @xmath342 , follows from @xmath471 by the norm splitting property . denote by @xmath472 the limit of @xmath473 . then the limit @xmath474 satisfies the following consistency condition : @xmath475 for all @xmath476 . the above consistency condition holds as @xmath477 , which together with the norm - splitting property in @xmath67 implies that the projection of @xmath478 onto @xmath479 are zero for all @xmath93 . from the consistency condition , we conclude that hold for all @xmath480 , and hence for all @xmath481 in the closure of @xmath482 . this together with the sparse density property of the sparse approximation triple @xmath105 proves the convergence of @xmath342 , to @xmath95 . take @xmath483 and let @xmath342 , be as in the greedy algorithm . write @xmath345 . thus @xmath484 and @xmath485 by proposition [ greedy.prop ] . this together with and implies that @xmath486 this completes the proof . given a sparse approximation triple @xmath105 , we say that @xmath95 is _ compressible _ ( @xcite ) if @xmath487 having rapid decay , such as @xmath488 where @xmath489 is the best approximation error of @xmath2 from @xmath490 , @xmath491 for the sequence @xmath342 , in the greedy algorithm , we have @xmath492 as @xmath493 the above inequality becomes an equality in the classical sparse recovery setting . we do not know whether and when the greedy algorithm is suboptimal , i.e. , there exists a positive constant @xmath45 such that @xmath494 even for compressible signals . the reader may refer to @xcite for the study of various greedy algorithms . , wiener s lemma : theme and variations , an introduction to spectral invariance and its applications , in _ four short courses on harmonic analysis : wavelets , frames , time - frequency methods , and applications to signal and image analysis _ edited by p. massopust and b. forster , birkhuser , 2010 .	in the first part of this paper , we consider nonlinear extension of frame theory by introducing bi - lipschitz maps @xmath0 between banach spaces . our linear model of bi - lipschitz maps is the analysis operator associated with hilbert frames , @xmath1-frames , banach frames , g - frames and fusion frames . in general banach space setting , stable algorithm to reconstruct a signal @xmath2 from its noisy measurement @xmath3 may not exist . in this paper , we establish exponential convergence of two iterative reconstruction algorithms when @xmath0 is not too far from some bounded below linear operator with bounded pseudo - inverse , and when @xmath0 is a well - localized map between two banach spaces with dense hilbert subspaces . the crucial step to prove the later conclusion is a novel fixed point theorem for a well - localized map on a banach space . in the second part of this paper , we consider stable reconstruction of sparse signals in a union @xmath4 of closed linear subspaces of a hilbert space @xmath5 from their nonlinear measurements . we create an optimization framework called sparse approximation triple @xmath6 , and show that the minimizer @xmath7 provides a suboptimal approximation to the original sparse signal @xmath8 when the measurement map @xmath0 has the sparse riesz property and almost linear property on @xmath9 . the above two new properties is also discussed in this paper when @xmath0 is not far away from a linear measurement operator @xmath10 having the restricted isometry property .
in the current paradigm of structure formation within the @xmath3 cold dark matter ( @xmath3cdm ) framework , initial small fluctuations in the cosmic density field are amplified by gravitational instability , eventually forming highly nonlinear structures called dark matter halos ( see @xcite for a review ) . galaxies then form at the centers of the gravitational potential wells of the dark matter halos by radiative cooling and condensations of baryonic gas ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? in order to reproduce the observed stellar mass function of galaxies in the cdm scenario , however , star formation in dark matter halos has to be inefficient ( e.g. * ? ? ? * ) , and various feedback processes have been proposed to suppress the star formation efficiency in dark matter halos . in this framework , therefore , galaxy formation and evolution are governed by a number of physical processes which , in turn , are characterized by a number of characteristic scales . first , cosmological @xmath4-body simulations have shown that the assembly histories of dark matter halos in general consist of two distinctive phases : an earlier phase of fast mass acquisition during which the potential well of a halo deepens rapidly with time , and a later phase of slow accretion , with a time scale longer than the hubble time ( e.g. * ? ? ? @xcite found that the two phases are separated at a time when a halo obtains about @xmath5 of its final mass ( see also * ? ? ? * ) . second , hydrodynamical simulations have demonstrated that radiative cooling is effective in halos with masses smaller than @xmath6 , so that the accretion rate of cold gas into galaxies is determined by the halo mass accretion rate , independent of radiative cooing ( e.g. * ? ? ? * ; * ? ? ? above this mass scale , on the other hand , radiative cooling is ineffective , so that the cold gas accretion is delayed by the cooling time scale . for massive halos with masses above @xmath7 , the cooling time scale may become longer than the hubble time , and so most of the baryonic gas is expected to be in the hot halo . third , supernova feedback from star formation is believed to be effective for halos with masses below @xmath8 ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? finally agn feedback from accreting super - massive black holes has been proposed as a mechanism to suppress star formation in massive halos , with masses above @xmath9 ( e.g. * ? ? ? * ; * ? ? ? a number of approaches have been adopted to explore the physical processes that govern galaxy formation and evolution , and to facilitate comparisons between theory and observation . the first is hydrodynamical simulation that includes both dark matter and baryonic components ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? however , due to limited resolutions and subgrid implementations of some key processes , the results obtained from such simulations are still questionable , even though they can match some observational data ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? furthermore , high resolution hydrodynamical simulations are computationally expensive , which prohibits the explorations of a large parameter space . because of this , an alternative approach , the semi - analytic model ( sam ) of galaxy formation , has been developed ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? the sam approach combines halo merger histories , obtained either from dark - matter only simulations or from analytical models , with gas and star formation processes using parametrized functions that describe the underlying physical processes . this approach is computationally inexpensive , allowing one to investigate a large set of different models . however , since all the physical processes are approximated with simple empirical functions , the reliability and accuracy of this approach need to be checked . more recently , a third approach has been adopted to understand how galaxies form and evolve in the cosmic density field . the goal of this approach is to establish the connections between galaxies and dark matter halos through an empirical approach , using observational data as constraints . models developed along this line include the halo occupation distribution ( hod ; e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) , the conditional luminosity function ( clf ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) , the halo abundance matching model ( ham ; * ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) , and the halo - based empirical model @xcite . to a certain degree , both the sam and empirical approaches are methods to summarize observational data in terms of model parameters characterizing the galaxy - halo connections . much progress has been made recently in this area . using the clf model and constraints of the observed luminosity function and correlation function of galaxies , @xcite found a characteristic halo mass scale , @xmath10 , in the relationship between galaxy luminosity / stellar mass and halo mass relation , suggesting that star formation efficiency declines rapidly toward both the higher and lower mass ends . with the use of galaxy groups selected from the 2df @xcite and sdss @xcite , @xcite found a similar mass scale from the observed galaxy luminosity / stellar mass - halo mass relations obtained directly from galaxy groups . in particular , @xcite suggested the existence of another characteristic mass scale , @xmath11 , where the galaxy luminosity - halo mass relation may change its behavior . similar results have since been obtained at higher @xmath2 with the use of the observed luminosity / stellar mass functions of galaxies . in particular , the presence of the mass scale at @xmath12 seems to extend to higher @xmath2 without showing strong evolutions @xcite . more recently , @xcite developed a halo - based empirical model to follow the star formation and stellar mass assembly histories of galaxies in dark matter halos . in particular , they used the observed conditional luminosity functions of cluster galaxies obtained by @xcite as an constraint in addition to the field stellar mass functions at different redshifts . they found that the observational data require two additional characteristic scales , a characteristic redshift , @xmath13 , and a corresponding mass scale at @xmath14 , below which star formation changes behavior at the characteristic redshift . these results clearly demonstrate that the observed conditional luminosity / stellar mass function of galaxies in clusters can provide important information about galaxy formation and evolution at high redshift . however , since clusters of galaxies only contain a small fraction of the total galaxy population , the results may be affected by some environmental effects that are specific only to clusters of galaxies . using galaxy groups of @xcite combined with galaxies in the sdss photometric catalogue , @xcite have recently measured the conditional luminosity / stellar mass functions ( hereafter csmfs ) that cover four orders of magnitude in galaxy luminosity , and three orders of magnitude in halo mass , from @xmath15 to @xmath16 . they found a characteristic luminosity scale , @xmath17 , below which the slope of the csmf becomes systematically steeper , and that this trend is present for all halo masses . this ubiquitous faint - end upturn suggests that it is formation , rather than cluster - specific environmental effect , that plays the dominating role in regulating the stellar masses of faint satellites . clearly , these observational results will provide new constraints on models . this paper consists of two parts . first , we compare model predictions from recent empirical models and numerical simulations to the csmfs of lan et al . to test how well different models accommodate the new data . we will show that , among all the models considered , only the @xcite model can match the observational data reasonably well . second , we use the new csmfs to update the empirical model of @xcite . the constrained model is then used to make predictions for other statistical properties of the galaxy population . the organization of this paper is as follows . section [ sec_model ] describes the empirical models to be tested and two recent numerical simulations , illustris @xcite and eagle @xcite , to be compared . in section [ sec_data ] , we describe the observational data that are used in our analysis , and present comparisons of the empirical models and the simulations with them . in section [ sec_update ] , we use the new observational data as constraints to update the parameters of the model of @xcite and examine whether a more extended model is needed by the data . in section [ sec_impl ] , we present a more detailed comparison of the model predictions in star formation rate , stellar mass - halo mass relation , mass assembly history , and stellar mass function for high redshifts . finally , we summarize and discuss our results in section [ sec_sum ] . in this paper we select a number of popular empirical models and two recent hydrodynamical simulations to test against observational data . here we describe these models and simulations briefly . table 1 lists the models and the simulations that we test . readers are referred to the original papers for details . ccc model & reference + / simulation & + y12 & @xcite + m13 & @xcite + b13 & @xcite + l15 & @xcite + l15-u & this work + illustris & @xcite + eagle & @xcite + + [ tab_model ] one of the simplest way to link galaxies to their dark matter halo / subhalo population is to use halo abundance matching ( e.g. * ? ? ? this approach assumes a monotonic relation between halo mass and galaxy stellar mass . satellite galaxies observed at a given redshift were all once central galaxies before they were accreted onto larger halos . since satellites are expected to evolve differently due to environmental effects such as tidal stripping and ram pressure stripping , the many abundance matching models seek for a monotonic relation between galaxy stellar mass and halo mass at the time when a halo first became a subhalo , instead of at the time of observation . most of previous investigations make the assumption that the halo mass - galaxy mass relation is independent of when a sub - halo is accreted into its host ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) . with this assumption , at a given redshift , halos of a given mass are therefore always linked to galaxies of the same stellar mass . however , it was found that applying this method to different redshifts actually leads to different stellar mass - halo mass relation , suggesting that the method implemented in this way is not self - consistent . as an improvement , models have been developed in which the galaxy - halo relation is allowed to depend on both halo mass ( defined e.g. at the time when a halo first becomes a sub - halo ) and the time when a halo becomes a sub - halo . we test four models in this category , by @xcite , @xcite , @xcite , and @xcite , respectively . the model of @xcite ( y12 , hereafter ) takes the same functional form as that proposed in @xcite for the halo mass - galaxy luminosity / stellar mass relation : @xmath18^{-1}. \end{aligned}\ ] ] this is basically a double power law specified by two asymptotic slopes , @xmath19 and @xmath20 , describing the low- and high - mass end behaviors , respectively , and by a characteristic mass scale @xmath21 where the transition between the two power laws occurs , and with @xmath4 being an overall amplitude . the four free parameters were assumed to be redshift dependent and the dependencies were modeled by simple functions . the above relation was used to assign stellar masses to halos at different redshifts . they adopted the halo mass function of @xcite to model the halo population . for sub - halos , the model of @xcite was used to follow both the mass function and the distribution in the accretion time ( the time when a halo first becomes a subhalo ) . a stellar mass is assigned to a sub - halo at the time of accretion according to its mass at that time using equation ( [ eq_mhvsms ] ) . the subsequent evolution of the satellite associated with a sub - halo was followed according to its orbit determined through the dynamical friction model . the model parameters were then obtained by fitting the model predictions to the observed stellar mass functions ( smfs ) of galaxies from @xmath22 to @xmath23 , and the correlation function of @xmath24 galaxies as a function of galaxy luminosity / stellar mass . @xcite ( m13 ) adopted a similar double power - law for the stellar mass - halo mass relation as described by equation ( [ eq_mhvsms ] ) , and simple functional forms to describe the redshift dependencies of the model parameters . they applied the relation to halos and sub - halos obtained from @xmath4-body simulations . individual halos and sub - halos are matched and traced across different snapshots ( i.e. different redshifts ) , so that merger trees are generated to track their evolutions . galaxies hosted at the centers of halos and sub - halos were referred to as centrals and satellites , respectively . for centrals , the stellar masses were given by the stellar mass - halo mass relation using the redshift and halo mass at the snapshot in question . for satellites , the stellar masses were obtained by applying the stellar mass - halo mass relation at the redshift when their halos first became sub - halos using their halo masses at this redshift , as in y12 . the stellar mass of a satellite was assumed to remain unchanged in the subsequent evolution . some uncertainties in the stellar mass - halo mass relation were taken into account . model parameters characterizing the stellar mass - halo mass relation were then tuned to match a set of observed smfs from @xmath22 to @xmath23 . the approach adopted by @xcite ( b13 ) was similar to those of y12 and m13 , but the stellar mass - halo mass relation assumed was more complicated and was designed in part to reproduce the observed smfs at the faint ends . here again , halo merger trees extracted from @xmath4-body simulations were used to trace the formation of dark matter halos . as in y12 and m13 , they applied their stellar mass - halo mass relation to ` infall ' mass at the time of accretion to assign stellar masses to subhalos . subsequent stellar mass loss of satellites after their accretion into their host halos was also taken into account . finally , they used the observed smfs at @xmath22 - @xmath25 , as well as the cosmic star formation rates and specific star formation rates , to constrain their model parameters . the @xcite model ( hereafter l15 ) was based on the star formation rate ( sfr ) - halo mass relation as a function of redshift : @xmath26 where @xmath27 , @xmath28^{-1}$ ] approximates the dynamical time of halos , @xmath29 , with @xmath30 being a characteristic mass scale and @xmath31 is parameter of @xmath32 . thus , @xmath33 for \{@xmath34 , @xmath35 , and @xmath36 } , respectively . this relation is applied only to central galaxies . after a galaxy becomes a satellite , lu et al . assumed that it moves on an orbit determined by its initial energy and orbital angular momentum together with dynamical friction . a satellite galaxy is assumed to merge with the central galaxy once it sinks to the center of the halo . at this time , it adds a fraction ( treated as a free parameter , @xmath37 ) of its mass to the central galaxy , and the rest is assumed to become halo stars . the sfrs in satellites were modeled with a simple exponential model , @xmath38\end{aligned}\ ] ] where @xmath39 is the time when the galaxy becomes a satellite , and @xmath40 $ ] is adopted to reflect halo mass dependence of the time scale , with @xmath41 and @xmath42 being free parameters . the stellar mass in a galaxy is then obtained by integrating the sfr over time , taking into account mass loss due to stellar evolution . lu et al . used halo merger trees generated with the algorithm developed by @xcite , which is based on the extended press - schechter formalism calibrated with @xmath4-body simulations . lu et al . adjusted both their functional forms and free parameters to match the smfs at @xmath43 and the csmfs of galaxies in clusters of galaxies as given by @xcite . they found that the model assuming all the parameters to be independent of redshift is not able to match the observed smfs at high redshift . they therefore extended their model by allowing @xmath44 to change with redshift as @xmath45 . this model was referred to as model ii in lu et al .. model ii was found to be able to describe all the stellar mass functions ( smfs ) at both low and high redshifts , but fails to match the faint - end upturn in the csmf of cluster galaxies . because of this , lu et al . extended their model once more by allowing the parameter @xmath20 , which dictates the sfr in low - mass halos , to depend on redshift : @xmath46 in this model , referred to as model iii by lu et al . , @xmath47 at @xmath48 , and the free parameter , @xmath49 , controls how rapidly the transition to @xmath50 occurs above the characteristic redshift @xmath51 . this model iii was found to be able to fit both the smfs at different redshifts and the csmf of cluster galaxies simultaneously . in this paper , whenever we present results and predictions of lu et al . , we mean those given by model iii with the model parameters from @xcite , except for section [ sec_impl ] where we update the model parameters using the new csmfs of lan et al . ( 2016 ) as constraints . we implement the empirical models described above to the dark matter halo population . we use the algorithm developed by @xcite to generate halo merger trees and to follow the build - up of dark matter halos . as mentioned above , this algorithm is based on the extended press - schechter formalism calibrated with results from @xmath4-body simulations . as shown in @xcite , the predictions of this algorithm match accurately many properties of halo merger trees obtained directly from simulations , including halo mass assembly history , halo merger rate , and sub - halo mass functions . the empirical models described above also take into account some uncertainties in the observational data and in the model assumptions , such as the intrinsic scatter in the stellar mass - halo mass relation , uncertainties in the stellar population synthesis and dust models , eddington bias , and errors in redshift measurements . unfortunately , how these uncertainties change as a function of redshift is poorly established . they are treated differently in different models . m13 adopted constant scatter in the stellar mass - halo mass relation and in the stellar mass estimate , while b13 parametrized the uncertainties as functions of redshift and treated them as a new set of free parameters to be determined in their model fitting . the treatment by y12 lies in between . in our implementations , we follow each individual model as close as possible . we use wmap7 cosmology to obtain the halo mass function , to construct halo merger trees , and to estimate distances from redshifts . we adopt the @xcite imf , the stellar population synthesis model of @xcite to account for stellar mass loss and to obtain stellar mass function from observations . these assumptions are the same as adopted in the original models , except for y12 where a @xcite imf was adopted . we correct the stellar masses of y12 model by a factor of @xmath52 to match the imf we adopt . we also test the predictions from two recent high - resolution , cosmological hydrodynamical simulations . the first is the illustris simulation @xcite , which follows @xmath53 particles for each of the gas and dark matter components in a total volume of @xmath54 , assuming wmap9 cosmology ( @xmath55 ) . the other components the simulation traces are stars , stellar wind particles , and super - massive black holes . the simulation starts from @xmath56 and includes physical processes such as radiative cooling , star formation , and various feedback processes . the free parameters in their model were constrained by using the star formation efficiency obtained from separate simulations that are more accurate in resolving small - scale structures . in our analysis , we use illustris-1 , their flagship simulation that has the highest mass resolution ( @xmath57 and @xmath58 for baryon and dark matter , respectively ) . to match the set of observations adopted here for model testing , we use the snapshot at @xmath59 , which contains a total of @xmath60 groups identified by the fof algorithm . in the simulation , galaxies are defined according to the spatial distribution of stars and stellar wind particles , and the brightness profile fit to them . the simulation assumes the @xcite imf and the stellar population synthesis model of @xcite . as the cosmological parameters of wmap9 are similar to those of wmap7 , the difference in cosmology is ignored in our analysis . we bin their stellar masses to obtain the stellar mass function ( smf ) . another simulation we use is the evolution and assembly of galaxies and their environments ( eagle ; * ? ? ? the eagle traces the evolutions of gas , stars , dark matter , and massive black holes , and implements physically motivated models for gas cooling , star formation law , stellar and agn feedback . the free parameters of the feedback models were tuned to match the smf and black hole mass - stellar mass relation at @xmath24 . the simulation starts from @xmath56 and adopts cosmological parameters from _ planck _ : @xmath61 @xcite . we use their simulation of the largest volume of @xmath62 for our analysis . it contains @xmath63 galaxies with stellar masses similar to or above that of the milky way . unfortunately , recalibrating their result to account for different cosmology is not trivial , since the impact of changing the parameters to the mass function is highly non - linear in principle . however , the other uncertainties that enter the models or the data must overpower the change in cosmology . we thus do not attempt any recalibration of the simulation results to account for the difference in cosmology . the @xcite imf and the spectral synthesis model of @xcite were assumed . we bin their stellar masses to get the stellar mass function . in this section , we describe all observational data used for our analyses . these include the stellar mass function ( smf ) of the general galaxy population ( 3.1 ) , the smfs separately for central and satellite galaxies ( 3.2 ) , and the conditional stellar mass functions ( csmfs ) of galaxies ( 3.3 ) . we use the local ( @xmath64 ) smf obtained from the combination of the results obtained by @xcite and @xcite . the data for stellar masses below @xmath65 is from baldry et al . , while the data at larger stellar masses is from moustakas et al .. here we briefly summarize the methodologies with which the smfs were computed , and refer the readers to their original papers for details . baldry et al . used the new york university value - added galaxy catalogue ( nyu - vagc ; * ? ? ? * ) , which includes @xmath66 galaxies at @xmath67 , to construct the local smf . they adopted the stellar mass estimates from @xcite , @xcite , and @xcite . in the data set we use , the stellar mass estimates are corrected to a @xcite imf . moustakas et al . estimated the local smf using galaxies cross - identified between the sloan digital sky survey data release 7 ( sdss dr7 ; * ? ? ? * ) and the galaxy evolution explorer ( galex ; * ? ? ? * ) deep imaging survey . this results in @xmath68 galaxies with a total sky coverage of @xmath69 . near - infrared photometry of these galaxies was obtained from the two micron all sky survey ( 2mass ; * ? ? ? * ) and the wide - field infrared survey explorer ( wise ; * ? ? ? the photometry in a total of 12 bands ( near - uv and far - uv of the galex , _ ugriz _ bands from sdss model magnitudes , @xmath70 magnitudes from the 2mass , and the integrated photometry at @xmath71 and @xmath72 from the wise all - sky data release ) was used to infer the galaxy stellar masses from spectral energy distribution ( sed ) modelling . more specifically , moustakas et al . used the flexible stellar population synthesis model of @xcite , a @xcite initial mass function ( imf ) , exponentially declining star formation histories ( sfhs ) , and the dust attenuation curve of @xcite , to model the seds of individual galaxies . the smf obtained by @xcite is in good agreement with some previous measurements , such as those of @xcite , @xcite , and @xcite . see their appendix b for detailed analyses how variations in the imf , sfh , spectral synthesis model , and dust attenuation can affect the smf obtained . the left panel of figure [ fig_lf_local_model ] compares the predictions of the empirical models with the observed local smf described above . as one can see , the prediction of the y12 model is too flat in the low - mass end to match the upturn seen in the observation . this discrepancy owes partly to the simple functional form ( a double power - law ) they adopted for the stellar mass - halo mass relation , and partly to the smfs that they used as observational constraints . in fact , y12 found that the two sets of smfs at high redshifts that they adopted led to significant differences in the inferred values of model parameters . the results used here are the predictions of ` smf2 ' referred in the original paper . the model of m13 also predicts a shallower faint - end slope than the observational data . as y12 , m13 also adopted a simple double power - law form for the stellar mass - halo mass relation , and used a local smf that has shallower faint - end slope than the one adopted here to constrain their parameters . in contrast , the prediction of b13 matches well the observed smf , even in the faint end . b13 adopted a rather flexible functional form for the stellar mass - halo mass relation , which is probably required to match the faint - end upturn in the smf . in addition b13 adopted the combined smf of @xcite and @xcite as one of their observational constraints , and so the good match between the model prediction and the data is not surprising . the prediction of the l15 model also matches well the observational data . note that l15 used the smf of @xcite as an observational constraint . their smf extends only to @xmath73 and so the faint - end upturn is not well represented . the faint - end upturn predicted by l15 is largely due to the csmf of galaxies in rich clusters , as given by @xcite , they adopted to constrain their model . the right panel of figure [ fig_lf_local_model ] compares the numerical simulation results with the observational data . the illustris simulation produces too many galaxies in the intermediate mass range as well as in the massive end , but too few low - mass galaxies . the overall shape of the predicted smf is very different from that of the observed smf . on the other hand , the prediction of the eagle simulation matches the observational data reasonably well above the resolution limit . this may not be very surprising , because the free parameters in the eagle simulation were tuned to match local observations . unfortunately , the relatively poor mass resolution does not allow us to investigate whether a faint - end upturn is predicted in the simulation . using the group memberships provided by @xcite group catalog ( see next subsection for more details ) , we can separate galaxies into two populations , centrals and satellites . a central galaxy is defined to be the most massive member in a group , while all other members in a group are called satellites . the csmfs can then be estimated separately for the centrals and satellites . formerly the total smf can be expressed in terms of these conditional functions as @xmath74 where @xmath75 and @xmath76 are the csmfs of the centrals and satellites , respectively , in halos of mass @xmath77 , while @xmath78 is the halo mass function , which is the number density of halos of masses between @xmath77 and @xmath79 . in @xcite , the csmfs are given only for satellites in groups with halo masses above @xmath80 ( see the next subsection for details ) . the satellite smf used here is obtained directly from their measurements by summing up the csmfs of such halos . for central galaxies , we use the results obtained by @xcite from their group catalog . since the group catalog is based on the sdss spectroscopic data , the central smf was measured only for galaxies above @xmath81 ( see table 6 in their paper ) . the data points in figure [ fig_lf_local_cs ] show the smfs for central and satellite galaxies , respectively . separating galaxies into centrals and satellites provides more information about the galaxy population than the total smf alone , and figure [ fig_lf_local_model ] and figure [ fig_lf_local_cs ] demonstrate this point clearly . for instance , although the empirical model by b13 ( see [ sec_model ] ) matches well the faint - end upturn in the observed total smf , this match is now revealed as due to an excess in the smf of central galaxies combined with a deficit in the smf of satellite galaxies . the m13 model has similar problems ; it under - estimates the number of satellite galaxies at the low - mass end even more strongly than b13 . the y12 model matches the central smf reasonably well , but it fails to reproduce the strong upturn in the low - mass end seen in the observed smf of satellite galaxies . overall , the l15 model can match both the observed central and satellite smfs , although some discrepancies in details can still be seen . this match is not trivial , because these observations were not used as constraints in l15 . the comparisons of the two gas simulations with the observational results are shown in the right two panels of figure [ fig_lf_local_cs ] . here we see that the eagle simulation matches the observational data reasonably well above its mass resolution limit . the illustris simulation matches the smf of satellites only in the intermediate mass range ; it over - predicts the central smf over almost the entire mass range , except at the knee of the smf . we use the csmfs obtained by @xcite as our main data set to compare with models . here we summarize briefly their methodology and results . lan et al . used galaxy samples from the nyu - vagc , which is based on the sloan digital sky survey data release 7 ( sdss dr7 ; * ? ? ? a @xmath82-correction was applied using the model of @xcite . in order to associate galaxies with clusters / groups of galaxies , they adopted the group catalog of @xcite , which was constructed by applying the halo - based group finder developed by @xcite to the sdss dr7 . the group finder assigns galaxies into halos using certain criteria in phase space , and galaxies residing in a common halo are considered to be members of the same group . more specifically , a tentative halo mass is assigned to a tentative group based on the galaxies that have already been assigned to the group , assuming a monotonic relation between the total stellar mass of all assigned members with @xmath83 and halo mass . the tentative mass is then used to estimate the virial radius and velocity dispersion of the halo , which in turn are used to update the group membership . the procedure is iterated until both group memberships and halo masses converge for all groups . @xcite used mock catalogs constructed from @xmath4-body simulations to show that the dark matter halo masses estimated in this way are consistent with those directly obtained from the simulations , with scatter of @xmath84 dex over three orders of magnitude in halo masses that cover the mass range relevant to our analyses . when we compare the csmfs obtained from models with the observational results , an uncertainty of @xmath84 dex is included in the model predictions . the halo masses used here are @xmath85 , the total mass enclosed by a radius , @xmath86 , within which the average density is @xmath87 times the mean density of the universe . lan et al . used only groups at @xmath67 , where halos with masses of @xmath88 are complete . to limit the uncertainty in redshifts due to peculiar velocities , they also eliminated groups at @xmath89 . with the groups and their positions identified in the sdss dr7 survey area , lan et al . estimated the excess of galaxy number in each luminosity bin within a projected distance of @xmath86 of each group . the conditional luminosity function ( clf ) of galaxies is then obtained by averaging galaxy counts within all groups of a given halo mass , with subtractions of the background and projection effects due to clustering on large scales ( see * ? ? ? * for the detail ) . lan et al . applied this method to the photometric sample of sdss dr7 , down to a @xmath90-band model magnitude of @xmath91 . this corresponds to @xmath92 ( or @xmath93 ) and @xmath94 at @xmath95 and @xmath96 , respectively . since the number density of more massive halos is smaller , lan et al . was able to estimate the clf down to @xmath97 for low - mass halos ( @xmath98 ) , but only to @xmath99 for massive halos . to convert their clfs into the corresponding csmfs , we use a mass - to - light relation based on galaxy colors and luminosities ( e.g. * ? ? ? * ) to obtain the stellar masses of galaxies . however , the uncertainty in the observed galaxy colors , especially for faint galaxies , may bias the stellar mass estimates and , therefore , the stellar mass functions . to reduce such bias , we first separate galaxies into blue and red populations by using the @xmath100-@xmath90 color separation suggested by @xcite [ see their equation ( 11 ) ] . we then use the observed luminosity of a galaxy and the mean @xmath100-@xmath90 color for the galaxy population at that luminosity , instead of the observed color of the galaxy , to estimate the stellar mass . the mean @xmath100-@xmath90 color - luminosity relations for the blue and red populations are derived in @xcite [ their equations ( c2 ) and ( c3 ) ] based on the same data set . we have made tests either by using the observed galaxy color or by artificially introducing some uncertainties in the galaxy color , and found that all these do not lead to any qualitative change of our results . note again , as described in [ sec_model ] , our model predictions for the stellar masses of individual galaxies also include some uncertainties in the stellar mass estimates to mimic the uncertainties in the observational stellar masses . lan et al . adopted a @xcite imf for the csmfs . with the estimated stellar masses of individual galaxies , we measure the csmfs using the same method lan et al . did for the clfs . the stellar mass functions are measured down to the limiting stellar masses at which both the stellar masses of blue and red galaxies derived from the flux limit photometric sample ( @xmath101 ) are complete . in addition , the limiting stellar mass bins are selected to ensure that they each contain at least five groups . we bootstrap the group catalog 200 times to estimate the errors in the derived csmfs . figure [ fig_clf_model ] compares the csmfs to those predicted by the empirical models . the predictions by the m13 and b13 models are qualitatively similar , with b13 predicting more low - mass galaxies . both models under - predict the csmfs at the low stellar mass ends , and the under - prediction is more significant for groups of lower halo masses . only for massive clusters are the predictions consistent with the observational data . the predictions of y12 are too shallow in the low mass end ; the model systematically under - predicts the csmf at the low mass end and over - predicts that in the intermediate mass range . in particular , y12 does not predict any upturn seen in the data . the l15 model matches the overall behaviors of the csmfs over the entire halo mass range . it also matches the csmfs in detail for most of the halo mass bins . however , the low - mass upturn it predicts for more massive halos may be too steep , especially for the two most massive samples . as mentioned above , the l15 model used the composite clf of galaxies in rich clusters given by @xcite as one of the constraints on their model . the faint - end upturn in this composite clf is significantly steeper than that of lan et al . usded here . the over - prediction is therefore due to the observational data which the model was tuned to match with . the model seems to under - predicts the csmf at the low - mass end in two mass bins : the lowest mass bin of @xmath102 $ ] , and the intermediate mass bin of @xmath103 $ ] . it is unclear if these discrepancies are due to random fluctuations in the data , or indicate that the l15 model has to be modified to accommodate the data . we will come back to this in the following section . figure [ fig_clf_simul ] compares the observational data with the two gas simulations . the illustris simulation mismatches the observation over a wide range of stellar masses for almost all the halo mass bins . overall , the simulation significantly under - predicts the csmfs at the faint ends , and over - predicts them in both the massive and intermediate mass ranges . the eagle simulation appears to be in better agreement with the observation , except that it does not reproduce sufficient number of massive galaxies in low - mass halos . unfortunately , its mass resolution prevents us from probing its behaviors at the faint end . in this section , we update the model parameters of l15 using only the observed csmfs as constraints . we will show whether the l15 model is able to accommodate the new data . here we use only the csmfs as observational constraints to determine the parameters in the l15 model . we use the multinest method developed by @xcite , which makes use of the nested sampling algorithm of @xcite , to compute the posterior distribution of the model parameters . the multinest is found to yield practically the same results as the traditional mcmc method but with @xmath104 times smaller number of likelihood calculations for the problem concerned here . the readers are referred to the original papers for details . the updated parameters and the updated model predictions are shown in table 2 and figure [ fig_clf_zeta ] , respectively . the csmfs predicted with the updated parameters are very similar to those obtained by l15 , except that the updated model ( hereafter l15-u ) predicts slightly flatter slopes at the faint - ends for massive haloes . as mentioned above , this is owing to the fact that the lan et al . clfs have shallower faint - end slopes for massive halos than the cluster galaxy luminosity function used by l15 . the model parameters do not change much either as shown in table 2 . we also show the prediction of l15-u for the smf of the total population in figure [ fig_lf_zeta ] . here l15-u is slightly worse than the original l15 in matching the data at the knee of the smf . it should be noted , however , that the smf was used as a constraint in l15 but not in l15-u . ccc parameters & l15 & l15-u + @xmath105 & @xmath106 & @xmath107 + @xmath108 & @xmath109 & @xmath110 + @xmath19 & @xmath111 & @xmath112 + @xmath113 & @xmath114 & @xmath115 + @xmath50 & @xmath116 & @xmath117 + @xmath49 & @xmath118 & @xmath119 + @xmath51 & @xmath120 & @xmath121 + @xmath122 & @xmath123 & @xmath124 + @xmath125 & @xmath126 & @xmath127 + @xmath128 & @xmath129 & @xmath130 + @xmath131 & @xmath132 & @xmath133 + @xmath134 & @xmath135 & @xmath136 + @xmath37 & @xmath137 & @xmath138 + + [ tab_param ] figure [ fig_sfr_l15 ] shows the sfr as a function of halo mass and redshift predicted by l15 and l15-u , respectively . again , the change is not large . this demonstrates that the low-@xmath2 csmfs alone can constrain models in a similar way as the field smfs at different redshifts . furthermore , as we will see in [ sec_impl ] ) , they also contain information about the low - mass end of the smf at high @xmath2 , where direct observations are still uncertain . as mentioned in section [ sec_model ] , l15 assumed the characteristic redshift , @xmath51 , the redshift at which the sfr in low mass progenitors changes behavior , to be independent of the host halo mass . however , it is plausible that @xmath51 depends on the host halo mass , because structure formation , and presumably star formation , are expected to occur earlier in regions that correspond to higher mass halos at the present day . motivated by this , we test a more extended model family in which the characteristic redshift changes with host halo mass at @xmath22 , @xmath139 : @xmath140 where @xmath141 controls the halo mass dependence of @xmath51 , and @xmath142 is @xmath51 for halos of @xmath143 . we use the same csmfs as used in the earlier subsection to constrain model parameters . to test if such an extension is necessary , we use the bayes factor , @xmath144 where @xmath145 is a given data set , @xmath21 and @xmath146 are two different models , and @xmath147 and @xmath148 are the parameter space of the models . this factor quantifies the preference of a given data set for one model family over the other . as it integrates over all parameter space of each of the model families , it naturally penalizes over - fitting . when all the data points of the csmfs are used as constraints , the bayes factor between the extended model ( the one including @xmath141 ) and the original l15 parametrization is given by @xmath149 , which indicates a strong need for having @xmath141 statistically . the median value of @xmath150 thus obtained implies that the characteristic redshift @xmath51 is @xmath151 for halos of @xmath152 , in comparison to @xmath153 for halos with @xmath154 . this increase of @xmath51 with host halo mass leads to flatter faint - end slopes for massive halos , giving better matches to the faint - ends of the csmfs for both low - mass and high - mass halos . it is worth noting , however , that the uncertainties in the stellar mass estimates may change the csmfs in both the lowest and highest mass ends , where the slopes of the csmf are steep . as a test , we use only the csmfs in the range @xmath155{\rm m_\odot}$ ] as the observational constraints . in this case , the models with or without @xmath141 are almost equally favored in terms of the bayes factor . given these , we conclude that the original form of the l15 model can still accommodate the new csmfs , and that the current data are still too uncertain to conclude if a more extended model family is required . in this section , we compare all the empirical models in their predictions for the stellar mass - halo mass relation , the star formation rates and stellar masses in halos of different masses at different redshifts , and for the smfs of high - redshift galaxies . figure [ fig_msmh ] shows the stellar mass - halo mass relation predicted by our updated model , in comparison with the predictions of the other three empirical models considered here and the results from the literature . different models made different assumptions for conversions from luminosity to stellar mass , for prescription of scatter in the relation , and cosmological models . they also employed different observations as constraints for the models . given all these differences , it is remarkable that the predictions of most models are consistent with each other within @xmath156 dex at @xmath157 for a large range of halo masses . all models predict a characteristic mass scale , @xmath158 , at which the stellar mass to halo mass ratio peaks . among the more recent results , b13 is an exception in that it predicts a strong upturn at the low - mass end . the earlier result of @xcite was obtained by using their luminosity - halo mass relation together with the assumption of a constant stellar mass to luminosity ratio , @xmath159 ( in the @xmath160 band of 2dfgrs which they used to constrain their model ) . at higher redshift , however , the predictions by different empirical models differ significantly . in particular , the update of l15 , l15-u , predicts a much higher star formation efficiency for low mass halos at high redshift , because of the boost of star formation rate at @xmath161 in low mass halos to match the upturns in the csmfs . figure [ fig_sfr_model ] compares the empirical models in terms of their predictions for the star formation rate ( sfr ) of central galaxies in halos at different redshifts . some models predict complicated star formation histories that are clearly due to over - fitting of the observational data . the l15-u model predicts much higher sfrs at @xmath162 in low - mass halos than other models , which is clearly a consequence of the strong upturns at the faint - ends of the csmfs used to constrain the model . the existence of a characteristic redshift , @xmath163 , is clearly seen in halos with @xmath164 , and its physical implications will be discussed later . figure [ fig_sfr_mh0 ] shows the model predictions for the star formation histories of central galaxies in halos of different present - day masses . the predictions of different models are very different . in particular , for present - day dwarf galaxies that reside in halos of @xmath165 , l15-u predicts a very active star formation episode at @xmath166 . in contrast , most of the stars in such halos are formed at @xmath167 in all other models . this difference has other observational consequences . indeed , as discussed in @xcite and @xcite , the early starburst in low - mass halos predicted by l15 is consistent with the observations that a significant fraction of old stellar population exists in local dwarf galaxies ( e.g. * ? ? ? * ) and that the star formation rate function at the low - rate end is very steep at @xmath168 ( e.g. * ? ? ? for milky - way sized halos , the star formation history predicted by l15-u is broader than those predicted by the other three models . for massive halos with @xmath169 , l15-u predicts a decline of the sfr with decreasing redshift starting from relatively high redshifts , in contrast to the predictions of b13 and y12 that the star formation rates remain relatively high all the way to the present time , and to the prediction of m13 that a rapid decline only occurs at @xmath170 . all these results demonstrate that different empirical models can make vastly different predictions for the star formation histories for present - day galaxies , even though all the models are tuned to match the observed smfs . figure [ fig_assembly ] shows the net stellar mass assembly histories for the central galaxies in halos of different present - day masses predicted by different models . the model predictions take into account _ in situ _ star formation , accretion of satellites , and stellar mass loss due to stellar evolution . again , for low - mass halos , where the increase of stellar mass is dominated by _ in situ _ star formation @xcite , l15-u is distinct from the other models in that about half of their stellar mass at the present was already in place by @xmath171 via star formation ( see figure [ fig_sfr_mh0 ] ) . for milky - way sized halos , however , the differences between the model predictions are milder . all the models predict that about half of stellar mass was in place by @xmath172 . there is a significant difference between l15-u and other models at high @xmath2 . for example , l15-u predicts that about 15% of the final stellar mass was assembled by @xmath163 , while less than 10% was predicted by the other models . for central galaxies in present - day massive halos with @xmath173 , the predictions of different models again become very different . m13 predicts a much later assembly for these galaxies than any other models . the predictions of b13 look similar to l15-u , but the increase in stellar mass with time is due to different reasons . while l15-u predicts that the increase at @xmath167 is dominated by accretion of stars from satellites , b13 predicts that a significant fraction of the increase at @xmath167 is actually due to _ in situ _ star formation ( see figure [ fig_sfr_mh0 ] ) . this difference is again due to the boost of star formation in low - mass halos at high @xmath2 in the l15-u model . the increased amount of stars formed in progenitors at high @xmath2 makes the accretion of stars more important in the growth of stellar mass in a massive galaxy , and the fraction of stars formed in situ has to be decreased proportionally in order to match the final stellar mass of the galaxy . the results demonstrate the importance of properly modeling the star formation in low - mass progenitors at high @xmath2 in order to understand the star formation and stellar mass assembly histories of massive galaxies at the present day . figure [ fig_highz ] shows the predictions of the empirical models for the stellar mass functions of galaxies at a number of redshifts . the predictions of b13 and m13 are similar in both slopes and amplitudes at the low mass ends , but b13 predicts many more massive galaxies than m13 , particularly at high redshifts . y12 predicts significantly flatter slopes at the low - mass ends , and more galaxies in the intermediate mass range , than the other three models . the stellar mass functions predicted by the l15-u model match the predictions of b13 at @xmath174 , but are significantly steeper at the low - mass ends . we select some observational smfs of high redshift galaxies from the literature to compare with the model predictions . specifically , we use the smfs at @xmath175 given by @xcite and @xcite . prez - gonzlez et al . used a sample combining data in three different fields with a total area of @xmath176 that have a total @xmath177 systems selected with the @xmath178 photometry of _ spitzer space telescope _ the sample is complete down to @xmath179 . marchesini et al . combined data from the deep nir musyc , the ultra - deep fires , and the goods - cdfs surveys to derive the smfs from the optical to mir broad bands photometry . prez - gonzlez et al . assumed a @xcite imf while marchesini et al . adopted a pseudo-@xcite imf it is known that the stellar mass estimated using a salpeter imf is roughly a factor of 1.4 higher than that given by a pseudo - kroupa or chabrier imf , and we correct all the stellar masses to the imf we adopt here . for the smfs at even higher redshifts , @xmath180 , we use the results by @xcite and @xcite . both b13 and l15-u match the observational data well at @xmath181 , while the other two models match the high-@xmath2 data poorly at the high - mass end . the prediction of the l15-u is significantly steeper than the observational results given by song et al . at @xmath168 . if the high-@xmath2 smfs are as shallow as those given by song et al . , then there may be a tension between the observed csmfs at low @xmath2 and the observed smfs at high @xmath2 , at least within the model family represented by the halo - based empirical model of l15 . galaxy formation and evolution within the current cosmological frame are controlled by a number of physical processes , many of which are still poorly understood from first principles . in the absence of a proper understanding of these processes , halo - based empirical models provide a useful way to establish the link between galaxies and cdm halos purely on the basis of observations and the current cosmology . in this paper we use a variety of galaxy stellar mass functions to test a number of popular empirical models . in particular , we focus on using the conditional stellar mass functions ( csmfs ) of galaxies in galaxy groups as obtained by @xcite to test the models . we find that the csmfs predicted by different models can be very different , even though they are all tuned to match the observed stellar mass function of the total galaxy population . this clearly demonstrates the power of the csmfs in constraining models . since the csmfs are measured from observations in the nearby universe , the samples that can be used are larger , and the stellar mass functions can be measured to the low - mass ends . as the galaxies that reside in present - day galaxy systems , such as clusters and groups of galaxies , are expected to have formed at various redshifts , the csmfs in groups / halos of different masses carry important information about galaxy formation in dark matter halos at different redshifts . in this paper , we use the csmfs of galaxies in groups of different masses to compare with the predictions of a number of popular halo - based empirical models and two numerical simulations of galaxy formation . we find that the two numerical simulations fail to match the observational data one way or another . the empirical models by @xcite and @xcite fail to reproduce the faint - end upturn of the field smfs from observations . the model by @xcite reproduces the faint - end upturn , but it is a combined result of over - prediction for central galaxies and under - prediction for satellites at the faint - end . in contrast , the model by @xcite matches reasonably well the csmfs in halos of different masses . the lu et al . model predicts a much higher star formation efficiency than the other models for low - mass halos at redshifts higher than a characteristic redshift after which the star formation is suppressed . the csmfs are then used as constraints to update the original model of l15 . the model parameters obtained here are very similar to those obtained in the original paper which uses a completely different set of observational constraints , demonstrating that the different data sets are consistent with each other . the observational constraints clearly prefer a model in which star formation in low - mass halos changes behavior at a characteristic redshift @xmath0 . there is also a tentative evidence that this characteristic redshift depends on environments , becoming @xmath1 in regions that eventually evolve into rich clusters of galaxies . however , given the uncertainties of the current observed csmfs in the low - mass ends , this environmental dependence of @xmath51 needs to be confirmed with better data . we use our constrained model to make predictions for a number of statistical properties of the galaxy population . these include the stellar mass functions of galaxies at high @xmath2 , the stellar mass - halo mass relations at different redshifts , and the star formation and stellar mass assembly histories of galaxies in dark matter halos of different masses . a comparison of our model predictions with those of other empirical models shows that different models can make vastly different predictions for these properties , even though all of them are tuned to match the observed stellar mass functions of galaxies . in particular , our constrained model predicts a much higher _ in situ _ star formation rate at @xmath182 for present - day dwarf galaxies than the other models . as a result , such galaxies have about @xmath183 of their current - day stellar mass already in place by @xmath171 . because of this boosted star formation in low - mass halos at high @xmath2 , the role of accretion of stars from satellite galaxies , relative to _ in situ _ star formation , in the build up of massive galaxies is more important in our model than in the other models . one of the main predictions of our constrained model is the existence of a characteristic redshift that separates an active star formation phase from a subdued star formation phase in low - mass halos . this change in star formation mode is likely related to the feedback processes that regulate star formation . as discussed in @xcite , energy feedback from stars and agns associated with active star formation and super - massive black hole accretion at high redshift may preheat the gas media around dark matter halos and suppress gas accretion and star formation at lower redshift @xcite . based on plausible assumptions about the star formation histories of the universe and the density of the intergalactic medium , the pre - heating is expected to occur around @xmath184 , and the specific entropy of the preheated gas is @xmath185 , which is important in affecting star formation in low - mass halos , because of their relatively shallow gravitational potential wells , but has no significant effects on halos with masses above @xmath186 ( e.g. * ? ? ? this preheating may also explain why the cold gas mass function at @xmath24 is shallow @xcite . in such a scenario , the pre - heating is expected to occur earlier in regions occupied by present - day massive halos , because intensive star formation and agn activities are expected to occur earlier in higher density regions where gravitational collapse is more accelerated . our tentative finding of the positive dependence of the characteristic redshift on halo mass is in agreement with such an expectation , but better observational data are needed in order to examine such dependence in more detail . hjm acknowledges the support from nsf ast-1517528 .	we compare predictions of a number of empirical models and numerical simulations of galaxy formation to the conditional stellar mass functions ( csmf ) of galaxies in groups of different masses obtained recently by lan et al . to test how well different models accommodate the data . among all the models considered , only the model of lu et al . can match the observational data ; all other models fail to reproduce the faint - end upturn seen in the observation . the csmfs are used to update the halo - based empirical model of lu et al . , and the model parameters obtained are very similar to those inferred by lu et al . from a completely different set of observational constraints . the observational data clearly prefer a model in which star formation in low - mass halos changes behavior at a characteristic redshift @xmath0 . there is also tentative evidence that this characteristic redshift depends on environments , becoming @xmath1 in regions that eventually evolve into rich clusters of galaxies . the constrained model is used to understand how galaxies form and evolve in dark matter halos , and to make predictions for other statistical properties of the galaxy population , such as the stellar mass functions of galaxies at high @xmath2 , the star formation and stellar mass assembly histories in dark matter halos . a comparison of our model predictions with those of other empirical models shows that different models can make vastly different predictions , even though all of them are tuned to match the observed stellar mass functions of galaxies . [ firstpage ] methods : statistical galaxies : formation galaxies : evolution galaxies : haloes .
understanding the origin and evolution of structure in the universe remains one of the most fundamental and active challenges of current astrophysical research . as the evidence in favor of a cosmological origin for the narrow , displaced absorption lines in qso spectra became overwhelming , it also became clear that both the metal - line systems and the lyman @xmath0 systems are invaluable tools for the study of some aspects of this problem . since ground - based lyman @xmath0 studies refer only to redshifts @xmath61.6 , they complement studies of galaxy clustering properties , the majority of which involve redshifts much less than this . however , precisely because the redshift regimes have been so different and because it has not been at all clear what relation exists between the typical low - column - density lyman @xmath0 absorbers and galaxies , these two approaches have remained disjoint . it was somewhat unexpected , but pleasing , that low - redshift lyman @xmath0 absorbers were found in sufficient numbers to enable meaningful studies of the evolution of the lyman @xmath0 absorbers and their relation to galaxies ( @xcite ) . this has presented astronomers with the opportunity to join these two lines of investigation . there are two levels at which such such attempts can be carried out : 1 ) purely statistical investigations aimed at comparing the clustering properties of galaxies and lyman @xmath0 absorbers , and 2 ) investigation of individual cases in which the possibility of establishing the presence ( or absence ) of a clear link between a lyman @xmath0 absorption line and something we could call a `` galaxy '' presents itself . preliminary discussions along these lines may be found in papers by @xcite and by @xcite . the present paper is a first attempt to pursue both these approaches along the sightline to 3c273 . the remainder of this paper is organized as follows : in [ observations ] we describe the different sets of observations we have assembled to investigate the environment of the lyman @xmath0 absorbers along the 3c273 sightline . in [ analysis ] we analyze them for possible associations or lack of associations of individual lyman @xmath0 absorbers with galaxies , and also give some statistical analysis of the clustering properties of the lyman @xmath0 absorbers with galaxies . in [ sumdisc ] we discuss these results in light of current models for the lyman @xmath0 absorbers and provide a brief summary and suggestions for further work . throughout this paper h@xmath7 is taken to be 80 km / s / mpc , the distance to the virgo cluster is taken to be 16.0 mpc ( a distance modulus of ( m - m)@xmath8=31.02 ) ( @xcite ) , and q@xmath7 is taken to be 0 . it has long been realized that imaging of the gas directly responsible for the low - column - density lyman @xmath0 absorbers is well beyond the reach of current technology . specifically , the neutral - hydrogen column densities of order @xmath9 @xmath10 that ghrs detected toward 3c273 are about 4 or 5 orders of magnitude below what can be imaged in 21 cm emission , even without taking into account the powerful radio background contributed by 3c273 itself . the h@xmath0 recombination surface brightness associated with these neutral - hydrogen column - densities is also several orders of magnitude below what is feasible to detect , unless the incident flux of ionizing photons is several orders of magnitude higher than that expected from the integrated background radiation . however , it has frequently be suggested that the lyman @xmath0 absorbers are intimately connected with , or are actual extensions of , entities which _ can _ be imaged either by means of h@xmath0 emission , starlight , or 21 cm emission - e.g. dwarf galaxies ( @xcite ) or shells of expanding gas ( @xcite ) or the outer regions of galactic disks ( @xcite ) . in the case of dwarf irregulars , for example , a very small episode of recent star formation might betray the presence of a dwarf irregular whose outer envelope produces the lyman @xmath0 absorbers . alternatively , expanding shells of gas might produce h@xmath0 emission via collisional ionization at a shock front . in addition , of course , possible association of individual lyman @xmath0 absorbers with specific galaxies , as well as statistical studies of absorber - galaxy correlation can be carried out with a sample of redshifts for galaxies in the field surrounding 3c273 . in this section we describe three such new sets of observations of a region centered on 3c273 . these are : [ obscoron ] coronagraph observations with narrow - band filters of a 5 diameter region , [ obscosmic ] deep broad - band imaging of a 17 diameter region , and [ obsfiber ] fiber spectroscopy of a [email protected] region down to a limiting magnitude of b=19 . we describe the analysis of these three sets of observations in [ analysis ] observations of a 53@xmath1153 region ( radius @xmath212 kpc at virgo ) around 3c273 were obtained during 1992 february 3 - 7 , with the university of hawaii coronagraph ( @xcite ) on the las campanas 2.5 m dupont telescope . a thinned 1024@xmath111024 tektronics ccd was used , binned 2@xmath112 , giving a scale of 1.23/pixel . the coronagraph blocking mask had a diameter of 5 . data were obtained with a gunn r filter , and also 5 specially acquired filters , 3 with width 13.5 , centered at 6586.2 , 6598.0 and 6756.0 ( hereafter referred to as vn1 , vn2 and hn ) , and 2 with width 25 , centered at 6643.2 and 6718.9 ( hereafter referred to as vb and hb ) . ( the above widths and central wavelengths are quoted for an f7.5 beam and a temperature of 15 c ) . the narrow - band wavelengths were chosen to match the redshifted position of h@xmath0 at the velocities of the 2 lyman @xmath0 absorbers listed in @xcite at velocities corresponding to the virgo cluster ( @xcite ) , and one absorber at 1251 . this is the lowest redshift , strong lyman @xmath0 system beyond the virgo cluster . the observing procedure involved cycling through the 6 different filters , with exposure times of 10 mins for the r and 25 filters , and 20 mins for the 13.5 filters . seven such cycles were completed over the 5 night run . during the observing run , it was discovered that the narrow - band filter hn had slightly non - parallel faces , resulting in detectable ` ghost ' images offset from bright stars and also 3c273 . in an attempt to minimise the effect of these , this filter was rotated though 90 between each night . for calibration purposes , observations were also obtained of mkn 49 ( an emission line galaxy in the virgo cluster with radial velocity 1524 km / s , and hence with h@xmath0 line - emission within 2 of the peak of vn2 ) and m87 , and also a number of bright standard stars . the images were reduced using iraf . the reduction steps were : bias subtraction , division by a flat field taken on the same night as the observations , rotation and shifting of the images to match a reference image , sky subtraction and averaging together of images taken with the same filter . it was found after the run that refocusing the coronagraph between taking the flat fields and the data meant that the flat field division left significant features in the data , both at the edges of the coronagraph field and also throughout the data at the location of what are presumed to be dust particles on the coronagraph optics or ccd window . this problem was particularly noticeable for the data taken on the last night of the run . however , the residuals are greatly reduced in the combined data . continuum sources were removed from the narrow - band images by subtracting off a scaled version of the two 25 filter observations . we investigated scaling methods , including measuring stars or galaxies in the images , but found that the scaling derived was consistent with simply subtracting off the average of the 25 filters from each 13.5 observation . that is , there was no evidence for a significant continuum slope or calibration difference across the 150 region of interest , and the factor 2 shorter broad - band exposure fairly accurately balanced the higher throughput of the 25 filters . this apparent consistency may be fortuitous : due to the small field ( small number of galaxies ) , and the errors in measuring the flux of faint galaxies . figure [ fig - coron ] shows the sum of the 25 data , and the continuum subtracted 13.5 data for 3c273 . as can be seen , the psf is a function of angle off - axis , and becomes quite broad and asymmetric at the edge of the field , due to aberrations in the coronagraph optics . a mosaic of gunn r band images of the field around 3c273 was obtained on 1992 february 23 , at the palomar 5 m , with the prime focus cosmic system . this recently commissioned camera contains a thinned 2048@xmath112048 tektronics ccd . due to poor seeing , this was read out with 2@xmath112 binning giving a scale of 0.56/pixel . exposures were taken roughly centered on 4 positions offset from 3c273 by 54 ne , nw , se and sw . four exposures of 5 minutes each were obtained at each position . during the observations , the moon rose , and so the background level of the images varies by almost a factor 2 . the data were reduced using iraf . after bias subtraction , the data were flat fielded using images of the dome . this left significant large scale structure in the images , and so a ` skyflat ' was constructed from the combined data images . this flat was median smoothed to remove any small scale structure . because of a region of very low sensitivity near the center of the ccd , and also the location of a very bright star coincidentally at the same place in one set of images , the flat fielded data still shows a weak negative feature near the center of each image . the sky background was determined for each image separately and subtracted . the offsets between the images were determined by measuring the positions of the qso and 2 bright nearby stars , which were common to all images and the data were then mosaiced together , giving a combined image with diameter 172 centered on 3c273 . the resulting image is show as figure [ fig - cosmic ] . during 1992 february 8 - 10 , spectra of objects in a [email protected] rectangle surrounding 3c273 were obtained with the fiber spectrograph at the las campanas 2.5 m dupont telescope . ( @xcite ) . this system has 128 fibers which are manually plugged into an aluminum plate over a [email protected] field . the fibers have a projected diameter of 35 . they feed the slit of a floor - mounted spectrograph . a 600 l / mm grating was used , with the 2d - frutti photon - counting detector , giving a resolution of 8.6 fwhm . three fields were observed for about 6000 seconds each , offset ew from each other . objects observed were chosen from a database produced by scanning an ukst iiiaj plate of the region with the apm scanning machine at cambridge . the apm produces a catalog of all the objects on a plate , with estimates of isophotal magnitudes , size and a ` sigma ' parameter that measures how much the image parameters differ from those of stars with comparable magnitude . we chose objects to observe from this catalog with ` sigma'@xmath43.0 . this includes many fairly compact objects , and resulted in a rather high contamination by stars , but also means that we found a number of compact galaxies that would otherwise have been missed . for each [email protected] fiber field , a magnitude sorted list of candidate galaxies was produced . due to restrictions on the minimum fiber separation , 330 out of a possible 336 object fibers were used , with 16 fibers per field set on blank sky . the fiber spectra were extracted and reduced using the iraf _ apextract _ package . the spectra were first traced and extracted . then fiber - to - fiber throughput differences were corrected with a flat field image , and the fibers were wavelength calibrated and rebinned to a common linear wavelength scale . the wavelength calibration used an arc spectrum to determine the non - linear relationship between wavelength and pixel number , after which a zero point shift was measured for each fiber using the strong sky lines in the data . finally , the sky emission was subtracted from each fiber using an unscaled template constructed from the 16 sky fibers in each frame . classification and radial velocity measurements for each spectrum were done in two ways . first each spectrum was inspected by eye , and classified as either a star , galaxy or unknown . then a radial velocity was determined by measuring either the position of the 4000 break , or that of the [ oii ] @xmath123727.6 feature ( actually a doublet , but unresolved in our data ) . a subjective assessment was also made of the reliability of the resulting radial velocity , dividing the sample into ` possible ' and ` definite ' . all the spectra were also analysed using the _ fxcor _ routine in iraf . each object was cross - correlated with three different templates : ( a ) a template made up from the 27 best s / n stars in the data , ( b ) a template made up from 3 high s / n late - type stars , and ( c ) a template made up from 16 emission line galaxies ( all shifted to their rest frames ) in the data . for each spectrum , the correlation with the highest peak was then selected . a reassuringly close match was found between the by - eye classification and velocities and the results from _ fxcor_. the resulting histogram of velocities was inspected . apart from 4 low s / n or hot stars ( for which a good template was not constructed ) , the stellar velocities found by _ fxcor _ had a distribution well represented by a gaussian with zero mean and an approximate dispersion ( @xmath13 ) of 85 km / s . we take this to be a reasonable estimate for our radial velocity uncertainties . it was also found that the subjective ` possible ' category in the by - eye radial velocity measurements matched rather well to a cross - correlation peak height less than 0.3 returned by _ fxcor_. apart from one outlier , the difference between the _ fxcor _ velocity and that measured by eye for the galaxy identifications with cross - correlation peak heights above 0.3 also was fairly well fit by a gaussian with @xmath13 of 85 km / s . in the end we obtained 129 definite galaxies , 43 possible galaxies , 86 definite stars , 4 possible stars , 10 fibers that had to be unplugged due to over - illumination ( which were hence almost certainly stars ) , 37 fibers which showed no flux ( either very low surface brightness galaxies or positional errors ) and 21 spectra which showed flux but which were unclassifiable either as stars or galaxies ( based on very low cross - correlation peak heights from fxcor and visual inspection ) . we should reiterate that a much higher ` success ' rate in finding galaxies could have been achieved by raising the cutoff ` sigma ' value used in choosing galaxies from the apm scans , at the cost of missing some compact galaxies . we present in table [ tbl - galcat ] the resulting galaxy redshifts . for each object , the table lists the ra and dec , an approximate b magnitude and the heliocentric radial velocity . the magnitudes were calculated using the apm isophotal magnitudes measured from the plates , crudely calibrated using the b magnitudes listed by @xcite and @xcite for objects in the scanned region . they could be in error by as much as 0.5 . a number of the brightest galaxies in the field were not included in the fiber survey . those with known redshifts within the survey region were added to the sample ( 4 objects taken from the may 5 1990 version of the cfa redshift catalog , @xcite ) . this gives a total sample of 176 galaxies with redshifts within the survey region which is roughly complete to a b magnitude of 19.0 . we plot the results of the survey in a number of projections . figure [ fig - zhist ] shows a redshift histogram of the galaxies , figure [ fig - fiber - sky ] shows the distribution of galaxies on the sky , while figure [ fig - pie ] shows the pie - diagrams obtained . in this section we go through each of the three observations described in [ observations ] in turn , deriving constraints on the absorbers . in [ analline ] we derive limits on h@xmath0 line emission at the velocities of the narrow - band filters and in [ analdwarf ] we calculate the maximum absolute magnitude a dwarf galaxy could have and remain undetected in our broad - band imaging . a long description of the correlation analysis between the ly@xmath0 absorbers and the galaxies found in the fiber survey is given in [ analcorr ] , in which the various available absorber and galaxy sub - samples are discussed , and two alternative extreme hypotheses for the absorber - galaxy correlation function are tested . finally in [ particular ] we discuss some particular aspects of the absorber - galaxy distribution found in the 3c273 sightline . the final continuum - subtracted coronagraph images have a measured rms dispersion ( away from residuals due to bright stars ) of 0.0018 dn / pixel / s . from calibration observations of hd84937 , this is equivalent to a 1@xmath13 flux limit of 2.10@xmath14 ergs/@xmath15/s/ , equivalent to an emission measure of approximately 2.8 pc @xmath16 . by blinking the images , it can be seen that none of the objects visible in the broad - band images have emission line fluxes @xmath17 . for comparison , the vn2 image of mkn49 showed a peak h@xmath0 line flux of 3.10@xmath18 ergs/@xmath15/s/ , i.e. 5,000 times higher than our 3@xmath13 detection limit . one can also perform the following thought experiment . what would the orion nebula look like if placed at the distance of the virgo cluster ? as discussed in @xcite , the orion nebula is actually a relatively low luminosity h ii region , with an h@xmath0 luminosity of only 10@xmath19 ergs / s . nevertheless , if placed at a distance of 16 mpc , it would still have a flux of 1.3 10@xmath20 ergs/@xmath15/s ( and would be unresolved - the nebula has a diameter of 5 pc , while at virgo , 1 corresponds to approximately 80 pc ) . thus it would be a factor 220 brighter than our 3@xmath13 limit . one can also calculate that the stromgren sphere around a single main sequence star of spectral class @xmath21b1/ would be detectable at the 3@xmath13 level ( @xcite ) . unfortunately , the expected surface brightness of an optically thick slab of hydrogen , simply bathed in the local uv background would not be detectable . taking the limit on the uv background from @xcite , and using the formulae from @xcite , one finds a surface brightness of @xmath22 ergs/@xmath15/s/ , about a factor 20 below our 3@xmath13 limit . higher spectral and spatial resolution data have been taken by t. williams ( private communication ) using the rutgers fabry - perot system at the ctio 4 m telescope . this data has not been fully analysed , but should produce even lower surface brightness limits ( or detection ) than our data . the main motivation for taking the cosmic images was to determine whether there were any dwarf galaxies within a 40 kpc radius from the line of sight to 3c273 which are too faint to be visible on the poss sky survey plates . examples of low surface brightness dwarfs in the virgo cluster can be seen in @xcite . in order to determine how faint a dwarf galaxy , which had morphological properties typical of those found in the virgo cluster , would be detectable in our image , we used the iraf _ artdata _ package to insert artificial dwarf galaxies into the data array . the magnitude scale was derived by assuming that the sky background in the gunn r band ( before the moon rose ) was 21.5 ( @xcite ) . b absolute magnitudes were converted to gunn r assuming a b - r color of 1.0 ( b - v of 0.7 ) , and a conversion from r to gunn r of r = r+0.43 + 0.15(b - v ) ( @xcite ) . thus a dwarf galaxy with m@xmath23=-15.5 was taken to have m@xmath24=-16.0 . the virgo distance modulus was taken to be 31.02 . dwarf galaxy properties were taken from @xcite . in particular , typical exponential scale lengths , with intensity @xmath25 for virgo dwarfs were taken to be 2 - 4 kpc . figure [ fig - artdata ] shows the same data as figure [ fig - cosmic ] , but with three dwarf galaxies added . one to the ne with m@xmath5=-14.5 and scale length 4.4 kpc , one to the nw with m@xmath5=-13.5 and scale length 2.2 kpc , and one to the sw with m@xmath5=-13.5 and scale length 4.4 kpc . these experiments demonstrated that we would be able to detect dwarf galaxies as faint as -13.5 at the distance of virgo , having morphological properties similar to those found near the cluster center . for comparison , the dwarf galaxy illustrated in figure 2 , panel 4 , labelled ` 1547 ' of @xcite has an absolute b magnitude of -14.8 . a complementary approach is to search for h i 21 cm emission from virgo dwarf galaxies . recently , van gorkom and her collaborators have set extremely low limits with the vla over a 40@xmath1140 arcmin field centered on 3c273 and over a velocity range of about 1000 km / s centered on 1300 km / s to a 1 sigma column - density limit of approximately 10@xmath26 @xmath10 ( @xcite ) . we have shown there are no h ii regions , or other strong h@xmath0 line - emitting gas , or dwarf galaxies near the virgo absorbers . having determined the redshift distribution of galaxies near the sightline to 3c273 , ( [ obsfiber ] ) we would now like to address the statistical question of the degree to which the lyman @xmath0 absorbers are correlated with luminous galaxies . if they are correlated , is the lyman @xmath0 absorber - galaxy correlation the same as the galaxy - galaxy correlation ? a cursory inspection of figure [ fig - pie ] is enough to show that there will not be a simple answer to this question . while there do seem to be lyman @xmath0 absorbers associated with clumps of galaxies ( e.g. the virgo absorbers , or the set of absorbers around z=0.02 - 0.03 ) , there are also absorbers in conspicuous ` voids ' ( at z=0.06 - 0.07 ) , and there are no absorbers associated with the prominent excess of galaxies at z=0.078 . we consider statistical tests for various assumptions about the correlation between the lyman @xmath0 absorbers and the galaxies near the sightline towards 3c273 for which we have redshift information . ultimately , the goal should be a quantitative and complete statistical description of the clustering properties of the lyman @xmath0 absorbers themselves and their correlation with various types of galaxies , clusters , voids etc . given both our rather meagre understanding of this problem and the small data set , we shall concentrate on testing the following two extreme null hypotheses about the lyman @xmath0 absorber clustering properties : 1 . the lyman @xmath0 absorbers are uncorrelated with galaxies and are randomly distributed . ( the second part of this assumption necessarily implies the first , but the converse is not necessarily true : the absorbers could be correlated among themselves but be uncorrelated with galaxies ) 2 . the lyman @xmath0 absorbers are correlated with galaxies in the same way that galaxies are correlated . more precise formulations of this hypothesis depend upon the particular test applied , as described below . in carrying out most of the tests described below , it is necessary to compute the 3-dimensional distance between every absorber - galaxy or every galaxy - galaxy pair . the question arises as to how to compute the component along the line of sight , since departures from a perfectly smooth hubble flow distort the mapping of redshift onto radial distance . with no information other than the angular coordinates and redshifts for the objects , we can not uniquely determine the separation along the line of sight for any individual pair of objects . for purposes of statistical tests we therefore make two different assumptions about this component ; the degree to which we do or do not derive similar results will provide some indication of the sensitivity of the test to the uncertainty in the estimation of this component : 1 . we simply ignore any departures from hubble flow 2 . we adopt the formalism of @xcite to estimate this component . these authors show how , knowing the two - point correlation function for the _ projected _ distance between pairs , in principal one can invert the integral equation relating the projected distance correlation function to the three - dimensional spatial correlation function . however , for our limited and rather noisy sample , this is not a very satisfactory procedure . moreover , we would like statistical estimates for the 3-dimensional separation for each pair for the purposes of carrying out other types of tests ( e.g. nearest neighbor tests ) . we therefore use the davis - peebles formalism as follows : adopting the assumptions described in their paper , the integrand in their equation ( 22 ) represents the probability that a pair with projected separation r@xmath27 mpc and velocity difference @xmath28 km / s has a separation in the radial direction of _ y _ mpc . we further assume their functional form for _ h(r ) _ in their equation ( 23 ) , with the parameter @xmath29 , and adopt their functional form for _ f(v ) _ , the probability of the relative velocity difference , as well as their expression for @xmath13 , the dispersion in _ f(v ) _ , i.e. their equation ( 32 ) . for @xmath30 or @xmath31 , where @xmath32 is the characteristic correlation length , the probability is strongly peaked at @xmath33 , i.e. , the pair separation is very likely to be that given by assuming a pure hubble flow . however , as both of these inequalities fail to be satisfied , a second maximum in the probability distribution arises at @xmath34 whose height depends upon the strength of the correlation - i.e. , a pair with a moderate velocity separation and small projected separation may be separated by their hubble flow distance , but if the correlation between such pairs is strong , this moderate velocity separation is more likely to arise from a pair at about the same distance from us , with strong gravitational interaction between them . as noted at the outset of this discussion there is no way to unambiguously determine the relative separation of any pair along the line of sight , but since we are interested in statistical applications _ we adopt the expectation value of this probability distribution as our second alternative algorithm_. a problem with this second approach is that evaluating the probability distribution for the radial separation of the pair requires that we know in advance the two - point correlation function between the pair . ideally , we could have dealt with this problem by using an iterative approach : for both the absorber - galaxy and galaxy - galaxy pairs separately , we start with some `` fiducial estimates '' for the two - point correlation ( e.g. @xmath35 mpc and @xmath36 ; c.f . equation 19 in davis and peebles ) to compute the expectation value of the radial separation for each pair , and thus compute the two - point correlation functions . with these new , separate best - fit values of @xmath37 and @xmath32 for the correlation functions for the galaxy - absorber pairs and the galaxy - galaxy pairs , we could then repeat the process until the parameters for the two - point correlation functions have converged . in fact , since the results of our statistical tests do not appear to be very sensitive to departures from a pure hubble flow , we have not carried out this iteration , but have simply used the single set of parameters ( @xmath35 mpc and @xmath36 ) in calculating the expectation values for both sets of pairs . in the following , we refer to these two algorithms for estimating the radial separations as the `` pure hubble flow '' and `` perturbed hubble flow '' cases . when listing object separations in the following sections we will give the separations found from the expectation value of the perturbed hubble flow model in brackets following the value for the pure hubble flow model . before carrying out any statistical tests we define the two samples and discuss appropriate corrections for completeness . a carefully defined list of lyman @xmath0 absorption lines is essential to a proper statistical discussion of the correlation properties of lyman @xmath0 absorbers with galaxies . the preferred list would obviously be drawn from a homogenous set of observations with the smallest detectable equivalent width covering all or most of the relevant redshift range . several line lists for the 3c273 sight line have been published ( @xcite ) . these line lists are compared in table [ tbl - lines ] . however , not only are these lists based upon 3 different hst spectrograph configurations ( ghrs g160 m , ghrs g140l , fos g130 ) but they have also been produced by different reduction procedures and line - finding and measuring algorithms and with differing acceptance criteria for what constitutes a `` real '' line . to investigate the importance of this latter source of inhomogeneity we have run the same continuum fitting , absorption line - finding and line - measuring software , `` jason '' , used for the fos line list , which is described in detail by @xcite on the data sets of @xcite and @xcite . the jason software was designed to run on fos data with an approximately gaussian point spread function ( psf ) . unfortunately this is not a good representation for the ghrs large aperture psf , and also in general the lines were resolved - meaning that the observed line profiles had neither the instrumental psf nor a gaussian shape . the current version of jason does not perform such convolutions , and so we have run the search routines assuming a fixed psf with the correct ( non - gaussian ) shape for the ghrs , but with no account taken for resolved lines . this means that the ews output by the jason software are not accurate , but the detection significance levels are approximately correct ( see @xcite ) . the data set used is that described by @xcite . we tabulate the significance levels from jason for the lyman @xmath0 lines in column 5 of table [ tbl - lines ] , after the positions and ews published in @xcite . it can be seen that all of the ` reliable ' lines listed by @xcite are confirmed by the jason software , with the notable exception of the line at @xmath121276.54 , which was also not found by @xcite . we checked this line by running the jason software on the original data used by @xcite , obtaining a significance level of 4.5 for the line . the line finding and fitting software used in @xcite was developed by r. carswell and j. webb , and is described in that paper . it used the ghrs psf convolved with a voit profile with variable width . two other entries in table [ tbl - lines ] require special comment : the line at @xmath12=1317.08 is identified as lyman @xmath0 in the list of @xcite whereas in the list of @xcite it is identified as ni ii . this issue has been discussed in detail in @xcite who give reasons for preferring the ni ii identification which we adopt . the second case involves the line at @xmath12=1393.86 . as discussed in detail by @xcite this line appears to be a blend of one of the members of the galactic si iv doublet and another strong line , whose only plausible indentification is lyman @xmath0 . the procedure used by @xcite in estimating the strength of this line ( which involves a detailed comparison of the line profiles of the si iv doublet ) is not incorporated into the jason formalism . for this reason we can not assign a formal uncertainty in the line strength . however the residuals from an unblended fit to the si iv doublet are highly significant . as a result of the above considerations , we have decided to adopt the following samples of absorbers for our statistical tests : ( 1 ) for the lyman @xmath0 absorber `` _ full _ sample '' we adopt the list of 16 lyman @xmath0 absorbers ( and their redshifts ) given in @xcite along with the additional low - redshift line ( @xmath121224.52 ) given by @xcite ; as noted in table [ tbl - lines ] , this last line is visible in the ghrs g140l spectra but was below the significance threshold of @xcite . this `` full sample '' is inhomogenous and/or biased in three senses : i ) the high resolution ghrs data covers only redshifts above [email protected] ; below this , only the fos data of @xcite and the ghrs g140l low resolution data are available . ( observations with the ghrs g160 m grating in the redshift regime from 0.0 to 0.016 are scheduled for hst cycle 3 ) . thus , the full sample may be biased against weak low - redshift lines in this redshift range that may be detected by these cycle 3 observations . ii ) the line of sight toward 3c273 may be somewhat atypical in that it passes through the southern extension of the virgo cluster . iii ) some of the weakest lines listed as `` possible '' in the @xcite list may not be real . accordingly , along with the full sample of 17 lines , we shall also consider two subsamples : ( 2 ) a `` _ homogeneous _ sample '' made of the set of 14 lines found only with the ghrs g160 m observations using the original carswell and webb software ( i.e. all but the first 3 lines in column 1 of table [ tbl - lines ] ) . ( as it happens , this is also equivalent to deleting the 3 low - redshift ( [email protected] ) lines possibly associated with the virgo cluster ) . ( 3 ) a `` _ strong _ sample '' composed of the set of 10 lines from @xcite with -log(p)@xmath47.5 ( but including the line at z=0.14658 for which a formal probability estimate was not possible due to blending - i.e. the above sample with the lines marked ` d ' in the comments column of table [ tbl - lines ] removed ) . we have listed the sample membership in the final column of table [ tbl - lines ] . note that in contrast to the galaxy sample discussed below there is no intrinsic observational selection against the higher redshift absorbers . one could , of course , define further samples . in particular , at the request of the referee , we have also run our statistical tests of the complete set of 5 lines listed in @xcite ( i.e. the line list in column 8 of table [ tbl - lines ] ) . unfortunately , the number of lines in this list is so small that neither of the two null hypotheses considered below in connection with cloud - galaxy association can be rejected with any significance . for this reason , and in order to keep the various combinations of absorption line and galaxy samples to managable proportions , we limit our tables of statistical results to consideration of the three samples defined above . an appropriate sample of galaxies with which to carry out the correlation analysis would be one which is complete to some limiting absolute magnitude throughout a cylindrical volume centered on the 3c273 sightline ( i.e. out to a constant impact parameter ) with a radius large enough to sample most of the expected power in the correlation function and length over ( and beyond ) the full redshift range covered by the lyman @xmath0 line sample . the observed sample described in [ obsfiber ] fails this requirement in two obvious respects : i ) it detects only the more luminous galaxies at the higher redshifts , ii ) it contains no galaxies with large impact parameters at low - redshifts . for some tests these deficiencies are probably not important but for others they are . accordingly , we will consider two galaxy samples . the first is the sample described in [ obsfiber ] , which we will refer to as the ` cone ' sample . the second is the cone sample together with all galaxies from the may 5 1990 version of the cfa redshift catalog , ( @xcite ) within 10 mpc of the 3c273 line - of - sight . this gives a heterogeneous sample with an unknown selection function , but is closer to the ideal ` filled cylinder ' than the cone sample . it contains 1498 galaxies , the vast majority being at virgo distances , and will be referred to as the ` cylinder ' sample . thus for each hypothesis tested below there are 6 galaxy - absorber sample combinations , and 2 possible estimates for the distance between every pair of objects giving a total of 12 data sets for each statistical test . it is obvious that this hypothesis can not literally be true - every sightline that passes close to any galaxy , except those utterly devoid of gas , will surely produce a detectable lyman @xmath0 line , and indeed examples of this are already known ( @xcite ) . nevertheless , in light of the fact that the high - redshift lyman @xmath0 absorbers show almost no power in their two - point correlation function ( c.f . @xcite and references therein ) it is of interest to see if the present data set does or does not exclude this hypothesis , and if it does , how strongly . having formulated this null hypothesis we consider two statistics as measures of correlation ( or lack of it ) : 1 . the average over all the lyman @xmath0 absorbers of the distance between a given lyman @xmath0 absorber and the n nearest galaxies in the sample , with n=1,3 and 5 . 2 . the total number of absorber - galaxy pairs within a fixed radius r , with r=500 kpc and 10 mpc . in order to see whether the values of these observed statistics are such that the null hypothesis can be rejected , we must find the distribution of these same statistics for many realizations of the null hypothesis . to do this , we carry out 1000 monte carlo simulations in which the same number of absorbers as that of the particular lyman @xmath0 sample under consideration are laid down randomly ( but follow the `` global '' distribution @xmath39 as determined by @xcite ) . redshift limits for the random absorbers were 0.03@[email protected] for the total sample , and 0.016@[email protected] for the homogeneous and strong samples . the results for the nearest - neighbour tests are summarised in table [ tbl - nearest ] . we show results for the single nearest neighbour galaxy , the mean of the 3 nearest and the mean of the 5 nearest . for each case , the three columns list the observed mean nearest - neighbour(s ) distance , the average of the mean nearest - neighbour(s ) distances produced by the monte carlo simulations , and the number of the 1000 monte carlo simulations that had a mean nearest - neighbour distance less than the observed one . thus this last column , divided by 1000 , can be taken as the probability that the observed value could arise from a sample of absorbers distributed at random with respect to the galaxies . the most striking result is that there is a less than 0.1% probability that the average nearest - neighbour distance to the single closest galaxy could arise from a randomly distributed set of absorbers . this is true for all sample combinations , and for either pure or perturbed hubble flow . for all samples , as one includes more galaxies in the nearest - neighbour average , the significance drops . after seeing this result , we wanted to test whether all the significance came from the nearest galaxy , and so ran tests on the second nearest galaxy only ( also given in table [ tbl - nearest ] ) . one can see that the observed mean distance is still significantly lower than that expected for a randomly distributed set of absorbers , but this could be explained by a combination of the highly significant correlation with the nearest galaxy and the strong galaxy - galaxy two - point correlation . this point is discussed in some detail by @xcite in the context of bright galaxies found near quasar mgii absorbers . we also give in table [ tbl - neargal ] the ra , dec and redshift of the nearest galaxy in the ` cylinder ' sample to each ly@xmath0 absorbers of the ` total ' sample ( assuming pure hubble flow ) . the final column in this table is the minimum absolute magnitude that could have been detected in the fiber survey . this shows that for one absorber there is no known galaxy with absolute magnitude above -17.8 within nearly 10 mpc , ( the same distance for the perturbed hubble flow model ) and that even the nearest absorber - galaxy pair in our sample are separated by 350 kpc ( 240 kpc ) . we will return to this in [ particular ] the results for the number of galaxies within a fixed radius of each absorber are given in table [ tbl - ranrad ] . these tests are essentially a comparison of the integrated two - point correlation function out to the given radius ( see @xcite for a discussion of this point ) . the numbers given in the table are the observed number of absorber - galaxy pairs within the given radius , the average of the number of pairs found in 1000 monte carlo simulations with randomly distributed absorbers , and the number of the monte carlo simulations with a larger number of pairs than that observed . thus the final column divided by 1000 is the probability that the observed numbers of pairs or more would arise from a random distribution of absorbers . table [ tbl - ranrad ] shows that there is no significant excess of galaxies within volumes of radius 10 mpc centered on the absorbers compared to a random distribution , apart from the cylinder / total subsample . for this combination there are a large number of pairs between the virgo absorbers and the many virgo galaxies in the cfa catalog . this result may be interpreted as saying that it is surprising to find 3 out of 17 absorbers below z=0.008 ( although see [ tgs ] ) . there is a marginally significant excess of absorber - galaxy pairs within 500 kpc , over that expected for a random distribution of absorbers , although the inclusion of the virgo velocity range removes the significance of this result . in summary , these tests seem to be consistent with the nearest - neighbour distance results , showing that there is an excess of close pairs of absorbers and galaxies , but that this result vanishes if the averaging is done over several galaxies or large radii . in order to test this hypothesis , one would like to use the same tests as were used in [ uncorr ] a difficulty arises though in generating a large number of monte carlo samples . we are loathe to compare the observed distributions with simulations involving anything other than the actual observed _ galaxy _ distribution since differences between the observations and simulations ( based , for example on n - body or other galaxy clustering models ) may result simply from inadequacies of such models , and it is not clear how to create simulations of ( fake ) absorber-(real ) galaxy distributions having cross correlation properties which are the same as the observed galaxy - galaxy correlation properties . one way to deal with this difficulty is to use a test which does not require the generation of monte carlo samples : if the absorbers are distributed in the same way as the galaxies , then , given the pencil - beam nature of the galaxy sample , the redshift distributions of the absorbers and the galaxies should be identical , after correction for differing selection effects in the two samples . table [ tbl - ks ] shows the kolmogorov - smirnov d - values and probabilities that the absorber redshift distribution is the same as that of the galaxies , after the galaxy distribution is corrected using the selection function shown in figure [ fig - zhist ] . this selection function was derived by assuming a schechter luminosity function with m@xmath41=-19.5 and @xmath0=-0.97 ( @xcite ) . because this test requires a known selection function , it can only be run for the ` cone ' galaxy sample . also , as it directly compares the redshifts , it does not require any assumptions about pure or perturbed hubble flow . it can be seen from table [ tbl - ks ] that when all the absorption lines are included there is a highly significant difference in the redshift distributions . this significance level becomes marginal when the virgo absorbers are removed , and vanishes when only strong absorbers beyond virgo are considered . our other test of the hypothesis that the absorbers and galaxies have identical correlation functions uses the observed galaxy sample to generate our monte carlo ` absorber ' sample . for each realization a number of the actual galaxy redshifts were selected at random , and were treated as absorbers on the 3c273 sightline . the actual algorithm involved selecting at random a number of the observed galaxy redshifts equal to the number in the absorber sample ( making no correction for the galaxy selection function , and with no restriction on how close together the chosen galaxies were ) , and treating these redshifts as if they were measured absorber redshifts on the 3c273 line of sight . the galaxies which provided these redshifts were removed from the galaxy sample for each test , to avoid an excess of ` spurious ' pairs . the number of absorber - galaxy pairs within a given radius were determined for both the real and 1000 monte carlo samples , in an identical manner to the second test in [ uncorr ] above . this procedure should be valid as long as the radius within which the pair counts are being made is significantly larger than the typical distance between a galaxy and the 3c273 line - of - sight . because of this , tests were not run for a 500 kpc radius . the results from these tests are given in table [ tbl - galrad ] . the columns are the same as for table [ tbl - ranrad ] , except that the final column lists the number of monte carlo runs with _ fewer _ absorber - galaxy pairs than the real sample . thus this number divided by 1000 is the probability that the absorbers could have a correlation function as strong as that between galaxies . as can be seen , the absorber - galaxy correlation function averaged over 10 mpc is significantly weaker than that between galaxies . in summary , the two tests above seem to show that the absorber - galaxy correlation function is significantly weaker than the galaxy - galaxy correlation function over large scales ( 10 mpc ) . even though there is clear evidence for galaxy - absorber clustering , _ there is a significant difference between the strength with which the galaxies are clustered with respect to each other and the strength with which the lyman @xmath0 absorbers are associated with galaxies . _ some of the results of the two preceding sections can be inferred directly by inspection of the actual two - point correlation functions themselves . a logarithmically binned version of the correlation functions for pure and perturbed hubble flows are shown in figure [ fig-2pt ] . the absorber - galaxy correlation function was generated using the total lyman @xmath0 sample and the cone galaxy sample . the correlation functions were normalised in the usual way using random samples with the selection function shown in figure [ fig - zhist ] . the error bars were estimated using the formulae in @xcite . as expected , the pure and perturbed hubble flow models agree fairly well for separations larger than one or two mpc ; the perturbed hubble flow model produces more very close pairs , since even small velocity differences wipe out small separations for the pure hubble flow model . in both cases , the absorber - galaxy correlation is clearly weaker than the galaxy - galaxy correlation on scales from about 1 - 10 mpc . however , although visual inspection of the absorber - galaxy correlation function may suggest that the correlation is significant out to about 10 mpc , in fact the pair tests summarized in table [ tbl - ranrad ] and the nearest neighbor tests summarized in table [ tbl - nearest ] both indicate that a statistically significant absorber - galaxy correlations can be detected in our data sets only over volumes which are smaller than this . actual associations between individual observed galaxies and absorbers in the 3c273 sight line are difficult to prove for a number of reasons : ( 1 ) the smallest projected distance to the 3c273 line - of - sight of all the galaxies in our sample is still 160 kpc . ( 2 ) galaxy rotation or velocity dispersions could produce velocity differences as large as 200 km / s between the mean galaxy velocity and an actually associated absorber ( also comparable to the 3@xmath13 error in our fiber data velocity measurements ) ( 3 ) in regions of high galaxy density along the line - of - sight , peculiar motions of the galaxies and the absorbers in cluster potential wells may make the velocity - distance relationship complex . this is especially true of the virgo region , and for the z=0.0034 and 0.0053 absorbers ( which also lie on the steep portion of the damping wings of the galactic lyman @xmath0 , and have only been observed at low resolution ) . with these caveats , it can be seen from table [ tbl - neargal ] that the closest absorber - galaxy pair has a separation of 350 kpc ( 240 kpc ) . outside of the virgo region ( where the velocity - distance relationship may be less complex , but also where our galaxy sample goes to much less faint absolute magnitudes ) , the smallest separation is 410 kpc ( 350 kpc ) . the best published example of an association between a lyman @xmath0 absorber and a galaxy is given in @xcite , where they find a galaxy within 90 kpc of the sightline to h1821 + 643 which has a strong lyman @xmath0 absorber ( ew 950 m ) within 400 km / s . there are no galaxies in our sample this close to the line of sight ( or indeed any absorption systems this strong ) . one can ( somewhat arbitrarily ) break the absorbers in our sample into two groups : ( 1 ) those with a galaxy within 1 mpc of the line - of sight , and with a velocity difference of less than 400 km / s ( all of the virgo systems and 6 of the higher - redshift systems . despite its entry in table [ tbl - neargal ] , the absorber at z=0.02622 actually has 2 galaxies within 400 km / s and 1 mpc projected separation . they do not appear in the table , as their large velocity differences ( @xmath4250 km / s ) make their separation large , assuming pure hubble flow ) , and ( 2 ) the rest ( 8 systems ) . there is no significant difference in the ew distribution of these two samples . our data can also be used to consider the question : what is the average galaxy diameter within which one would see a neutral - hydrogen column density of at least 10@xmath42 @xmath10 ? in practice this is rather a naive question , as the cross - section almost certainly depends on galaxy luminosity and probably also morphology . one might also expect a patchy distribution of neutral - hydrogen in the outer parts of galaxies leading to a covering factor not equal to one . nevertheless , ignoring these complications , and excluding the virgo velocity range due to ( a ) the possibility of large peculiar velocities ( b ) a wish to avoid the large range of intrinsic galaxy luminosities in the sample and ( c ) because our absorber ew limit is higher in this region , we find that of the 12 galaxies with projected separation to the line - of - sight less than 1 mpc , 8 show lyman absorption systems within 400 km / s . a velocity difference of up to about 400 km / s could possibly be attributed to internal motions within a large galaxy , coupled with our measuring error , or alternatively to a small group . having 8 or more such matches would occur 0.6% of the time if the absorbers were randomly distributed in velocity space . however , of these 8 galaxies , two are associated with the absorber at z=0.02622 , and three with the absorber at z=0.2933 . clearly several distinct galaxies can not be producing the same absorber , and so in fact only 5 of the 12 systems within 1 mpc of the line - of - sight to the quasar can be legitimately associated with individual absorbers . reducing the projected separation to 500 kpc , one finds 3 galaxies , with velocity separations of 20 , 120 and 0 km / s ( see table [ tbl - nearest ] ) . this may be interpreted as saying that the cross - section of galaxies with luminosities greater than 1/10 of l@xmath3 for log(nh)@xmath43 is between 0.5 and 1 mpc , although by `` cross - section '' we do not mean that the absorbers are neccesarily associated with the actual galaxy in question . because of the 3 virgo absorbers , all of the virgo galaxies have an absorber within 400 km / s . however , it is clearly unreasonable to suggest that more than one galaxy is associated with a given absorber . thus of the 19 virgo galaxies within 500 kpc of the line - of - sight , 16 must not be producing absorption . of these 19 , 2 have absolute magnitudes brighter than 1/10 l@xmath3 , and so are consistent with the above statement , despite the higher absorber ew limit , as long as only one of the galaxies with luminosity @xmath401/10 l@xmath3 is producing observed absorption . one of the stronger lines in our absorber sample is also the most isolated . any galaxy brighter than -18 would be in our sample at the redshift of the isolated z=0.06655 absorber . however , the nearest such galaxy has a projected separation of 4.75 mpc and a velocity difference of 710 km / s , which corresponds to a spatial separation of 10 mpc for both pure and perturbed hubble flow models . indeed studying the galaxy distribution in figure [ fig - pie ] , the three absorbers with 0.060@[email protected] seem to lie in a ` void ' in the galaxy distribution . while it is certainly possible that these absorbers are associated with galaxies below our absolute magnitude limit , _ these isolated absorbers are a clear demonstration that one can not associate all low - redshift lyman @xmath0 absorbers with luminous ( l@xmath3 ) galaxies . _ the most striking feature of the observed galaxy redshift distribution is a concentration of galaxies centered at a redshift of z @xmath21 0.078 . this excess is almost certainly associated with a structure which includes the galaxy cluster abell 1564 . this cluster has a tabulated center at 12:32:25 + 02:07:11 ( 1950 ) , giving it an offset of 89 sw of the 3c273 sightline . this angle corresponds to a separation of 7.3 mpc at the cluster distance . abell 1564 is the closest abell cluster to the 3c273 sightline , but is only richness class 0 . redshifts for two member galaxies are given in @xcite , giving an average cluster redshift of 0.0793 . selecting all galaxies in our fiber sample with redshifts between 0.07 and 0.085 ( 54 galaxies ) , one derives a mean redshift of 0.0781 with rms of 0.0021 ( 630 km / s ) . for comparison , selecting galaxies between redshifts of 0.15 and 0.17 ( the cluster around 3c273 , see figure [ fig - zhist ] ) , one gets 24 galaxies with mean redshift 0.1581 and rms 0.0024 ( 720 km / s ) . despite the location of abell 1564 to the sw , the galaxies in the redshift slice 0.07 - 0.085 show a weak concentration to the ne of the fiber field , although there are galaxies in all parts of the fiber area surveyed . if they are really part of a structure including a1564 , the structure must have a size of at least 8 mpc . however , the nearest lyman @xmath0 absorber is 0.0095 from the peak in redshift space ( or 2850 km / s , which is 4.5 times the rms from the mean of the concentration ) and is at a distance of 36 mpc from the center of the concentration . it thus seems unlikely that there are any observed lyman @xmath0 absorbers in our line of sight which are physically associated with this concentration . we can test a slightly different form of the hypothesis considered in [ ident - corr ] , ( though we grant its application to the z @xmath21 0.078 galaxy concentration involves _ post facto _ statistics ) , namely : what is the probability that , if the lyman @xmath0 absorbers are distributed in space `` in the same way '' as the galaxies in our sample , we should find the observed number of lyman @xmath0 absorbers ( none , in our case ) . by `` in the same way '' we mean that : @xmath44 and by `` in the vicinity of '' we shall mean within @xmath45 of the redshift peak of the concentration . we consider that our selection function can be meaningfully applied over the redshift range from about z= 0.016 out to our adopted `` proximity cutoff '' at z= 0.151 . carrying out the integration of our sample ( i.e. the galaxy numbers weighted by the selection function ) between @xmath45 of the peak of the galaxy concentration and over the redshift range from 0.016 to 0.151 we find that _ after correction for the selection function _ , about 0.33 of an absolute - magnitude - limited cylindrical sample of galaxies should be found within 2.5@xmath13 of the velocity peak of this structure . since the lyman @xmath0 absorbers are not subject to such a selection function , we can apply this equation to all the volume elements along the line of sight and , assuming further that the effective cross - section of the lyman @xmath0 absorbers ( @xmath13 ) is constant over the redshift range 0.016 - 0.151 , one can write : @xmath46\ ] ] @xmath47\ ] ] this equation represents the number of lyman @xmath0 absorbers in the vicinity of the peak that we should expect to see if our hypothesis is correct . over the redshift interval 0.016 - 0.151 we have observed up to 14 lyman @xmath0 absorbers so that we expect @xmath48 absorbers , and we observe none . the probability of this occuring is thus @xmath49 . this is not a strong result due to the rather small number of absorbers expected , and the _ post facto _ nature of the test as we acknowledged above , but is additional suggestive evidence that the lyman @xmath0 absorbers do not follow the galaxy distribution and in particular that they may avoid strong concentrations of galaxies . it could be strengthened , of course , by observing additional lines of sight through dense concentrations . it would be nice to repeat the above calculation for the case of the three absorbers found in the virgo velocity region . however , there are a number of reasons why any estimate of the probability of finding three absorbers in this region with an apparent overdensity of galaxies is highly uncertain . firstly , our absorber sample is incomplete in this velocity range , as described in [ lyasamp ] secondly , the area of sky surveyed is so small that the selection function correction for this velocity range is very large . taking our selection function at face value , the 8 galaxies actually in our survey at virgo velocities imply that 60% of an absolute - magnitude - limited cylindrical galaxy sample from z=0 to 0.151 would be found in that velocity range . our hypothesis above would then predict that ( neglecting the incompleteness of the absorber sample ) one would expect 0.6@xmath1117@xmath210 absorbers , when we only see 3 . unfortunately , it is not possible to place a meaningful uncertainty on this apparent under - density of absorbers in the virgo region , for the reasons listed above . in summary , there is marginal evidence , both from a clump of galaxies at z=0.078 and also from the virgo velocity range , that lyman @xmath0 absorbers are less common in regions of high galaxy density . we have assembled several types of observations in an attempt to find objects with which the lyman @xmath0 absorbers along the line of sight to 3c273 might be associated , and in order to carry out statistical tests of galaxy - lyman @xmath0 absorber association . in particular , we obtained narrow - band images centered on and off the expected position of any h@xmath0 emission which might be associated with three of the low redshift lyman @xmath0 absorbers , and obtained deep broad - band images of a 17@xmath1117 field centered on 3c273 . both these searches were negative . our failure to identify any broad - band or h@xmath0 emission from plausible `` galaxy - like '' objects a few tens of kpc from the 3c273 sight line at the approximate distance of the virgo cluster will be checked by more sensitive and extensive searches for h@xmath0 by t. williams ( private communication ) and 21 cm emission by @xcite and @xcite . we have also obtained redshifts for a large number of galaxies in the vicinity of the 3c273 sightline . again , we find no unambiguous instance of association of any of the lyman @xmath0 absorbers with individual galaxies . we define a number of samples for both the lyman @xmath0 absorbers and the galaxies and estimate the 3-dimensional separation between each galaxy - galaxy pair and each lyman @xmath0 absorber - galaxy pair based upon two models for converting the observed redshift difference between any pair into a radial separation , viz . i ) the assumption of a pure hubble flow and ii ) a statistical model of `` perturbed '' hubble flow based upon work of @xcite . the resulting data base is used to carry out statistical tests to confirm or reject two null hypotheses about the association of galaxies and lyman @xmath0 absorbers , namely : i ) the lyman @xmath0 absorbers show no tendency to cluster around galaxies ii ) the lyman @xmath0 absorbers cluster around galaxies exactly as the galaxies cluster about each other . while neither of these two hypotheses can be unambiguously rejected in the sense that every combination of samples and flow hypotheses reject both of them at significant levels , the evidence from these tests , and from the galaxy - galaxy and lyman @xmath0 absorber - galaxy two - point correlations themselves , points quite strongly to the conclusion that both hypotheses are false . in particular , over length scales from about 1 to 10 mpc there seems little doubt that the lyman @xmath0 absorbers cluster around galaxies less strongly than the galaxies themselves cluster . this is born out by an examination of a redshift interval centered at about z=0.078 at which a strong concentration of galaxies occurs but in the neighborhood of which there are no lyman @xmath0 absorbers . additionally , we find at least one lyman @xmath0 absorber for which no galaxy with absolute magnitude brighter than about -18 can be found closer than about 5 mpc . taken together , all this evidence suggests that the most significant conclusion we have reached is that _ the majority of low - redshift lyman @xmath0 absorbers are not intimately associated with normal luminous galaxies . _ in view of the fact that it has long been realized that at high redshifts there is an absence of power in the lyman @xmath0 absorber two - point correlation function in redshift space , except possibly at the very smallest velocity separations , this conclusion is not too surprising . on the other hand , it is also fairly clear that there is _ some _ tendency for the lyman @xmath0 absorbers to cluster around galaxies , and even weak evidence that this clustering becomes strong at very small separations . also @xcite have investigated the auto - correlation function of the low - redshift absorbers seen in the line - of - sight to h1821 + 643 , and show that there is only a 4% probability that the observed ` clumping ' arose from a randomly distributed sample . none of the foregoing points unambiguously , in our estimation , to a particular model for the formation and evolution of the lyman @xmath0 absorbers . as have others , we simply offer the following speculations which appear to be compatible with the facts as they are presently understood : at high redshifts , the lyman @xmath0 absorbers consist primarily of entities which are only very loosely associated with larger mass objects ( e.g. proto galaxies ) , and which are evolving fairly rapidly . possibly this dominant population consists of absorbers in which gravitational binding ( eg by dark matter ) plays no significant role . in addition to this group , there is a smaller population of absorbers which are evolving less rapidly , possibly stabilized by dark matter and which are clustered more strongly about galaxies . at very low redshifts , this latter population is beginning to constitute a large enough fraction of the absorbers that power in the two - point correlation function is detectable . thus , the present mix of lyman @xmath0 absorbers appears to have clustering properties intermediate between present - epoch normal galaxies and a random non - clustered population . this property also appears to be shared by the low - luminosity moderate redshift `` blue galaxies '' ( @xcite ) , leading to the plausible conjecture that the lyman @xmath0 absorbers are more closely related to low mass , low luminosity galaxies than they are to l@xmath3 galaxies , though the relation is clearly not one - to - one . we have no good way at present of estimating the characteristic scale or masses of the low - redshift lyman @xmath0 absorbers . a guess at a diameter of 30 kpc is as plausible as any . in particular , consider a pancake whose diameter is 30 kpc and whose thickness is 10 kpc . in this case , at the present epoch , a hydrogen column - density of 10@xmath42 - 10@xmath50 @xmath10 normal to the face of the absorber , coupled with estimates for the present - epoch energy - density of ionizating radiation leads to a total gas mass of order 10@xmath51 m@xmath52 , but a mass of only a few hundred solar masses of _ neutral _ hydrogen . it is of interest that the mass function of neutral - hydrogen gas clouds appears to be truncated below about 10@xmath53 m@xmath52 ( @xcite ) . as maloney ( 1992,1993 ) has shown , the gas in a flaring galaxy with decreasing column - density will undergo a rather sudden transition along its face from being mostly neutral to mostly ionized , with the consequence that few if any contours with neutral - hydrogen column - density of order 10@xmath54 @xmath10 are known . similarly , it is conceivable that , given the appropriate run of length scale with mass , a sequence of masses would have the property of making a sudden transition in the mass function of neutral - hydrogen starting at about 10@xmath55 m@xmath52 , leading to a dearth of objects with total hi masses for several orders of magnitude below this . if these speculations have any connection with reality then one might expect to see some similarity in the clustering properties of the lyman @xmath0 absorbers and low mass galaxies . we are currently attempting to obtain redshifts of galaxies of lower luminosity along the 3c273 sightline in order to investigate this possibility . we would like to thank greg aldring and steve shectman for help and advice in using the lco fiber system , neil reid for giving us access to one of the new palomar sky - survey plates , alan sandage for allowing us to inspect one of his older iiiaj plates , mark davis and margaret geller for advice on generating monte carlo samples with known correlation functions , don schneider for help with the jason software and xavier barcons , donald lynden - bell , mario mateo , houjon mo , michael rauch , john webb , jeff willick and dennis zaritzky for useful conversations . we would also like to thank the referee , j. n. bahcall , for his helpful and constructive comments which improved the content and presentation of the paper . slm and rjw acknowledge support through nasa contract nas5 - 30101 and nsf grant ast-9005117 . allen , c. w. 1973 , `` astrophysical quantities 3rd edition '' , athlone press , london , pg 267 bahcall , j. n. , jannuzi , b. t. , schneider , d. p. , hartig , g. f. , bohlin , r. and junkkarinen , v. , 1991a , in `` the first year of hst observations '' , ed a. l. kinney and j. c. blades , pg 46 bahcall , j. n. , jannuzi , b. t. , schneider , d. p. , hartig , g. f. , bohlin , r. and junkkarinen , v. , 1991b , , 377 , l5 bahcall , j. n. , jannuzi , b. t. , schneider , d. p. , hartig , g. f. and green , r. f. , 1992a , , 397 , 68 bahcall , j. n. , jannuzi , b. t. , schneider , d. p. , hartig , g. f. and jenkins , e. b. , 1992b , , 398 , 495 bahcall , j. n. , bergeron , j. , boksenberg , a. , hartig , g. f. , jannuzi , b. t. , kirhakos , s. , sargent , w. l. w. , savage , b. d. , schneider , d. p. , turnshek , d. a. , weymann , r. j. and wolfe , a. m. , 1993 , , in press binggeli , b. , sandage , a. and tammann , g. a. , 1985 , , 90 , 1681 brandt , j. c. _ et al . _ , 1993 , , 105 , 831 chernomordic , v. v. and ozernoy , l. l. , 1983 , nature , 303 , 153 davis , m. and peebles , p. j. e. 1983 , , 267 , 465 fransson , c. and epstein , r. , 1982 , , 198 , 1127 haynes , m.p . and giovanelli , r. 1984 , , 89 , 758 huchra , j. p. 1990 may 5 1990 version of the cfa redshift catalog , obtained through the astronomical data center . jacoby , g. h. , branch , d. , ciardullo , r. , davies , r. l. , harris , w. e. , pierce , m. j. , pritchet , c. t. , tonry , j. l. and welch , d. l. , 1992 , , 104 , 599 kennicutt , r. c. , 1984 , , 287 , 116 kent , s. m. , 1985 , , 97 , 165 loveday , j. , peterson , b. a. , efstathiou , g. and maddox , s. j. , 1992 , , 390 , 338 maloney , p. , 1992 , , 389 , l89 maloney , p. , 1993 , , in press massey , p. , 1990 , noao newsletter # 21 , march 1990 mo , h. j. , jing , y. p. and borner , g. , 1992 , , 392 , 452 mo , h. j. , einasto , m. , xia , x. y. and deng , z. g. , 1992 , , 255 , 382 mo , h. j. , morris , s. l. and weymann , r. j. , in preparation metcalfe , n. , fong , r. , shanks , t. and kilkenny , d. 1989 , , 236 , 207 morris , s. l. , weymann , r. j. , savage , b. d. and gilliland , r. l. , 1991 , , 377 , l21 osterbrock , d. e. , `` astrophysics of gaseous nebulae and active galactic nuclei '' , 1989 , university science books , ca phillips , s. , disney , m. j. and davies , j. i. , 1993 , , 260 , 453 pritchet , c. j. and infante , l. , 1992 , , 399 , l35 rauch , m , carswell , r. f. , chaffee , f. h. , foltz , c. b. , webb , j. k. , weymann , r. j. , bechtold , j. and green , r. f. , 1992 , , 390 , 387 salzer , j. j. 1992 , , 103 , 385 sandage , a. and bingelli , b. , 1984 , , 89 , 919 savage , b. d. , lu , l. , weymann , r. j , morris , s. l. and gilliland , r. l. , 1993 , , 404 , 124 schneider , d. p. , hartig , g. f. , jannuzi , b. t. , kirhakos , s. , saxe , d. h. , weymann , r. j. , bahcall , j. n. , bergeron , j. , boksenberg , a. , sargent , w. l. w. , savage , b. d. , turnshek , d. a. and wolfe , a. m. , 1993 , , in press shectman , s. a. , 1992 , in `` fiber optics in astronomy ii '' , ed p. m. gray , asp volume 37 , pg 26 songaila , a. , bryant , w. and cowie , l. l. , 1989 , , 345 , l71 stockton , a. 1980 , in iau symposium 92 , `` objects of high redshift '' , ed g. o. abell and p. j. e. peebles , pg 87 van gorkom , j. , h. , 1993 , teton conference proceedings , in press , ed m. shull and h. thronson van gorkom , j. , h. , bahcall , j.n . , jannuzi , b. , and schneider.d . 1993 , , in press vilas , f. and smith , b. a. , 1987 , , 26 , 4 weinberg , d. h. , szomoru , a. , guhathakurta , p. and van gorkom , j. h. , 1991 , , 372 , l13	we present new ground - based data following up on the hst discovery of low - redshift lyman @xmath0 absorption in the sight - line to the quasar 3c273 . our goal is to investigate the relationship between the low - column - density absorbers and higher column - density objects such as galaxies or h ii regions . narrow - band filter observations with a coronograph show that there are no h ii regions or other strong h@xmath0 line - emitting gas within a 12 kpc radius of the line - of - sight to the quasar , at the velocities of three of the absorbers . broad - band imaging in gunn r shows that there are no dwarf galaxies at virgo distances with absolute magnitude above m@xmath1 - 13.5 and within a radius of 40 kpc from the line - of - sight to the quasar . finally , we present fiber spectroscopy of a complete sample of galaxies within a radius of 1 , down to an apparent magnitude of b@xmath219 . analysis of this sample , combined with galaxies within 10 mpc of the quasar line - of - sight taken from the literature , shows that the absorbers are definitely not distributed at random with respect to the galaxies , but also that the absorber - galaxy correlation function is not as strong as the galaxy - galaxy correlation function on large scales . we show that our data are consistent with the hypothesis that all galaxies more luminous than 1/10 l@xmath3 have effective cross - sections ( for association with absorbers whose neutral - hydrogen column - density ( log(nh ) ) is @xmath413.0 ) , of between 0.5 and 1 mpc . we also show a clear case of a lyman @xmath0 absorber which has no galaxy brighter than m@xmath5=-18 within a projected distance of 4.8 mpc , and discuss the possibility that lyman @xmath0 absorbers are destroyed in a rich galaxy environment .
in some extensions of the standard model , there are new heavy particles which belong to an su(2)@xmath2u(1 ) multiplet @xmath3 and have no , or very small , mixing with other particles . the masses of these particles are almost degenerate to a value @xmath1 by the gauge symmetry . although the spontaneous breaking of the su(2)@xmath2u(1 ) symmetry by higgs bosons may generate mass splitting @xmath4 among them , the tree - level mass splitting generally behaves as @xmath5 and becomes very small for @xmath6 . this is especially the case for a very heavy fermion multiplet where tree - level renormalizable couplings to the higgs bosons @xmath7 are forbidden by symmetry . some of the examples are the almost pure winos or higgsinos @xcite in special parameter regions of the minimal supersymmetric standard model @xcite , su(2)@xmath0 triplet fermions in type iii seesaw model for neutrino masses @xcite , and also models @xcite where vector - like heavy fermion multiplets are added to the standard model by hand . in such cases , it has been known @xcite that the dominant part of the gauge - symmetry - breaking mass splitting within the multiplet @xmath3 comes from the radiative correction . although the form of the mass correction strongly depends on models , the contributions involving electroweak gauge bosons @xmath8 , shown in fig . 1(a ) for the one - loop , are common in a wide class of extended models . since gauge symmetry breaking in this diagram comes from the squared masses ( @xmath9 , @xmath10 ) in the loops , one naively expect the @xmath11 contribution to the mass splitting . however , due to the singularity of the diagram near the threshold , at @xmath12 , @xmath13 contribution to the mass splitting appears , which does not vanish in the @xmath6 limit : roughly speaking , it is `` nondecoupling '' . this mass splitting is phenomenologically interesting , especially in the case where the neutral component @xmath14 of @xmath3 , either fermion or boson , is stable or has very long lifetime , and may be a candidate for the cosmological dark matter . in such a case , the loop - generated mass splitting between charged components @xmath15 of @xmath3 and @xmath14 is crucial for estimating the rates of the @xmath16 decays expected at colliders , and also for possible resonant annihilation @xmath17 for indirect detection of @xmath14 @xcite . to evaluate the mass splitting within @xmath3 to the next - to - leading order , we need two - loop calculation of the mass correction for the members of @xmath3 . in this paper , we perform such calculation for the loop corrections by the standard model particles , generated by the electroweak gauge interactions of @xmath3 . for simplicity , we concentrate on the su(2)@xmath0-breaking and `` nondecoupling '' part of the mass correction , which should be relevant for the mass splitting in the @xmath6 case . since the electroweak contributions to the mass correction should be determined by the su(2)@xmath2u(1 ) representation of @xmath3 , we work in the framework of the minimal dark matter model @xcite , which has been proposed as a minimalist approach to the dark matter problem , for the fermion case . in this case , dirac or majorana fermions in an su(2)@xmath0 multiplet @xmath3 with su(2)@xmath0 isospin @xmath18 and u(1 ) hypercharge @xmath19 ( and having no su(3 ) color ) are added to the standard model . the lagrangian is @xmath20f , \ ] ] where @xmath21 for dirac(majorana ) fermions , respectively . note that the mass corrections presented in this paper are common to both types of fermions . @xmath22 denotes su(2)@xmath2u(1 ) gauge covariant derivative for @xmath3 . since @xmath3 has no direct couplings to the higgs boson , the members of @xmath3 , @xmath23 ( with charge @xmath24 , @xmath25 ) have a common mass @xmath1 at the tree - level . we assume that @xmath1 is sufficiently larger than the masses of standard model particles ( @xmath26 , @xmath27 , top quark @xmath28 , higgs boson @xmath29 ) , typically @xmath30(tev ) which is cosmologically favored in the minimal dark matter model @xcite . we also use approximation that all other particles in the standard model are massless . the pole mass @xmath31 of @xmath23 at the two - loop order is given in terms of the self energy of @xmath23 @xmath32 as @xmath33 - [ m\sigma_k^{(2)}(m^2)+\sigma_m^{(2)}(m^2 ) ] \nonumber \\ & & + [ m\sigma_k^{(1)}(m^2)+\sigma_m^{(1)}(m^2 ) ] [ \sigma_k^{(1)}(m^2 ) + 2m^2\dot{\sigma}_k^{(1)}(m^2 ) + 2m\dot{\sigma}_m^{(1)}(m^2 ) ] \nonumber \\ & \equiv & m+ \delta m^{(1)}+ \delta m^{(2 ) } . \label{eq : masscor}\end{aligned}\ ] ] here @xmath34 and @xmath35 are the one - loop and two - loop parts , respectively . the dot in eq . ( [ eq : masscor ] ) denotes the derivative with respect to the external momentum squared . the absorptive part of the self energy is @xmath36 and need not be considered here . loop integrals are regularized by the dimensional regularization ( @xmath37 ) with the subtraction scheme . the form of the one - loop mass correction @xmath38 is well known @xcite . abbreviating the factor @xmath39 , it is expressed as @xmath40 where @xmath41 , @xmath42 , @xmath43 , @xmath44 , and @xmath45 \nonumber \\ & = & m \left [ \frac{3}{\epsilon}-3\log m^2 + 4 -f\left ( \frac{m_v}{m}\right ) \right ] , \nonumber \\ f(x ) & \equiv & 2x(2+x^2)\sqrt{4-x^2}\tan^{-1}\frac{\sqrt{2-x}}{\sqrt{2+x } } -x^2+x^4\log x \nonumber \\ & = & 2\pi x -3x^2+\frac{3}{4}\pi x^3 + o(x^4 ) . \label{eq : x1}\end{aligned}\ ] ] we use the passarino - veltman one - loop functions @xcite defined as @xmath46^{-\epsilon } , % \int\frac{d^dk}{(2\pi)^d } \frac{1}{(k^2-m_1 ^ 2 ) [ ( k+p)^2-m_2 ^ 2 ] } , \nonumber \\ b_{22}(p^2 , m_1 , m_2 ) & = & \frac{1}{2\epsilon(1-\epsilon)}\int_0 ^ 1 dz \ , [ ( 1-z)m_1 ^ 2+zm_2 ^ 2 - z(1-z)p^2 -i\delta]^{1-\epsilon } , % \left . % \int\frac{d^dk}{(2\pi)^d } \frac{k^{\mu}k^{\nu } } % { ( k^2-m_1 ^ 2 ) [ ( k+p)^2-m_2 ^ 2 ] } \right\|_{g_{\mu\nu } { \rm part } } . \end{aligned}\ ] ] and @xmath47.\ ] ] the @xmath48 term of eq . ( [ eq : x1 ] ) gives the nondecoupling mass splitting within the multiplet . for example , for @xmath49 , the one - loop mass splitting between @xmath23 and the neutral component @xmath14 of @xmath3 is written as @xcite , independent of @xmath18 , @xmath50 where , in the @xmath6 limit , @xmath51 the numerical value in eq . ( [ eq10 ] ) is obtained by using the pole masses @xmath52 gev , @xmath53 gev , @xmath54 , and the qed running coupling in the scheme @xmath55 , cited from ref . @xcite , as input parameters . note that the value ( [ eq10 ] ) should change by @xmath56 mev depending on choices of the renormalization scheme for the input parameters . we now calculate the two - loop mass correction @xmath57 coming from diagrams shown in fig . 1(b - e ) . we use feynman gauge fixing for simplicity , although the final result should not depend on the gauge fixing method . in the multiplet @xmath3 . the solid thick line and wavy line represent @xmath3 and electroweak gauge bosons @xmath58 , respectively . the black circle in ( b ) represents the one - loop self energy of the gauge bosons @xmath59 , by the standard model particles . , width=566 ] the contribution of the diagram fig . 1(b ) with the insertion of the one - loop self energy of the electroweak gauge boson , @xmath60 , is written as @xmath61 where @xmath62[k^2-m_{v_2}^2 ] [ ( k+p)^2-m^2 ] } \right\|_{\psla = m } \ , . \label{eq : sigma2b}\end{aligned}\ ] ] here we list the analytic forms of @xmath63 in the standard model for completeness @xcite . the contributions from the @xmath64 quark loops are , up to the overall factor @xmath65 ( @xmath66 is the color number of quarks ) , @xmath67 , \label{eq13 } \\ % % % % % % % % % % % % % % % % % % % \pi_{tb}^{\gamma\gamma}(k^2 ) & = & \sum_{q = t , b } s_w^2q_q^2\pi^{vv}_q(k^2 ) , \\ % % % % % % % % % % % % % % % % % % % \pi_{tb}^{\gamma z}(k^2 ) & = & \sum_{q = t , b } \frac{s_w}{c_w } q_q\left ( \frac{1}{2}i_{3q}-q_q s_w^2 \right ) \pi^{vv}_q(k^2 ) , \\ % % % % % % % % % % % % % % % % % % % \pi_{tb}^{zz}(k^2 ) & = & \sum_{q = t , b } \frac{1}{c_w^2 } \left ( \frac{1}{2}(i_{3q})^2 -i_{3q}s_w^2q_q + s_w^4 q_q^2 \right ) \pi^{vv}_q(k^2 ) \nonumber\\ & & + \frac{m_t^2 } { 2c_w^2 } b_0(k^2,m_t , m_t ) , \end{aligned}\ ] ] where @xmath68 the contributions of other quarks and leptons are obtained by appropriate changes of @xmath69 , @xmath70 , and @xmath71 . for the gauge and higgs boson loops , we have , abbreviating the overall factor @xmath39 , @xmath72 \nonumber\\ & & + ( 1 + 8(1-\epsilon)c_w^2 ) \widetilde{b}_{22}(k^2,m_w , m_z ) \nonumber\\ % & & + [ ( 4k^2+m_z^2+m_w^2)c_w^2-m_z^2s_w^4]b_0(k^2,m_w , m_z ) & & + [ 4c_w^2 k^2 -m_z^2 + 3m_w^2]b_0(k^2,m_w , m_z ) \nonumber\\ & & + \widetilde{b}_{22}(k^2,m_w , m_h ) -m_w^2 b_0(k^2,m_w , m_h ) , \label{eq18 } \\ % % % % % % % % % % % % % % % % % % % \pi_{vh}^{\gamma\gamma}(k^2 ) & = & s_w^2 [ 4(3 - 2\epsilon)\widetilde{b}_{22}(k^2,m_w , m_w ) + 4k^2 b_0(k^2,m_w , m_w ) ] , \\ % % % % % % % % % % % % % % % % % % % \pi_{vh}^{\gamma z}(k^2 ) & = & 4s_wc_w \left ( 3 - 2\epsilon-\frac{1}{2c_w^2 } \right ) \widetilde{b}_{22}(k^2,m_w , m_w ) \nonumber\\ & & + s_wc_w ( 4k^2 + 2m_z^2 ) b_0(k^2,m_w , m_w ) , \\ % % % % % % % % % % % % % % % % % % % \pi_{vh}^{zz}(k^2 ) & = & \left [ 4c_w^2(3 - 2\epsilon)-4+\frac{1}{c_w^2 } \right ] \widetilde{b}_{22}(k^2,m_w , m_w ) \nonumber\\ % & & + \left [ 4c_w^2 k^2 + \left ( 4-\frac{2}{c_w^2}\right ) m_w^2 \right ] & & + [ 4c_w^2 k^2 + 4m_w^2 - 2 m_z^2 ] b_0(k^2,m_w , m_w ) \nonumber \\ & & + \frac{1}{c_w^2 } [ \widetilde{b}_{22}(k^2,m_z , m_h ) -m_z^2 b_0(k^2,m_z , m_h ) ] . \label{eq21}\end{aligned}\ ] ] in addition , there are also the contributions of @xmath3 to @xmath59 . however , it is shown that the resulting @xmath73 contributions to @xmath57 are completely cancelled by the renormalization of the parameters in @xmath38 . we may calculate the integrals ( [ eq : sigma2b ] ) by extending the general formulas for the two - loop mass corrections @xcite , by including finite masses for @xmath74 . however , since we are interested in the su(2)@xmath0-breaking and nondecoupling part of eq . ( [ eq : sigma2b ] ) , it is prefered to expand the integrals ( [ eq : sigma2b ] ) in @xmath75 and then separate the @xmath73 terms from the dominant and gauge - symmetric @xmath76 terms , before numerical evaluation . this is achieved by applying the asymptotic expansion of the feynman integrals near the threshold @xmath77 , as described in ref . the @xmath73 part of the integral ( [ eq : sigma2b ] ) is then obtained as @xmath78[k^2-m_{v_2}^2 ] ( 2k\cdot p ) } \pi^{v_1v_2}(k^2 ) . \label{eq : sigma2b2 } \end{aligned}\ ] ] in the following , we show only the @xmath73 part ( [ eq : sigma2b2 ] ) of the corrections @xmath79 . by substituting the self energies ( [ eq13][eq21 ] ) , the integrals ( [ eq : sigma2b2 ] ) are expressed in terms of the two - loop functions ( @xmath80 ) @xmath81^a ( 2k\cdot p ) } \left [ b_0(k^2 , m_1 , m_2 ) -\frac{1}{\epsilon } \right ] , \nonumber \\ \frac{i}{(4\pi)^2 } x_{22-a}(m_v , m_1 , m_2 ) & \equiv & \int\frac{d^dk}{(2\pi)^d } \frac{m}{[k^2-m_v^2]^a ( 2k\cdot p ) } \nonumber \\ & & \times \left [ b_{22}(k^2 , m_1 , m_2 ) -\frac{1}{\epsilon } \left ( \frac{m_1 ^ 2+m_2 ^ 2}{4}- \frac{k^2}{12 } \right ) \right ] , % \nonumber \\ % \frac{i}{(4\pi)^2 m } x_{a}(m_v ) & \equiv & % \int\frac{d^dk_1}{(2\pi)^d } % \frac{1}{[k_1 ^ 2-m_v^2]^a ( 2k_1\cdot p ) } , \label{xinteg}\end{aligned}\ ] ] and products of the one - loop functions . note that the functions in eq . ( [ xinteg ] ) are independent of @xmath1 and have no overall divergences . we calculate these functions by numerical integration of the feynman parameter integrals shown below , @xmath82 , \nonumber \\ \\ x_{0 - 2}(m_v , m_1 , m_2 ) & = & \frac{\pi}{2m_v } \left [ \log\frac{m_v^2}{\mu^2 } + 2\int_0 ^ 1 dz \ , \left ( \frac{\sqrt{z(1-z)}}{\sqrt{r_1(1-z)+r_2z}+\sqrt{z(1-z ) } } \right . \nonumber \\ & & \left . + \log(\sqrt{r_1(1-z)+r_2z}+\sqrt{z(1-z ) } ) \right ) \right ] , \\ x_{22 - 1}(m_v , m_1 , m_2 ) & = & -\frac{\pi}{3}m_v^3 \left [ \frac{1}{4}\ { 1 - 3(r_1+r_2 ) \ } \log\frac{m_v^2}{\mu^2 } -\frac{2}{3 } + \frac{9}{4}(r_1+r_2 ) \right . \nonumber \\ & & \left . + \int_0 ^ 1 dz \ , \left ( \ { -3z(1-z)+2(1-z)r_1 + 2zr_2 \}\frac{\sqrt{(1-z)r_1+zr_2 } } { \sqrt{z(1-z ) } } \right . \nonumber\\ & & + 3\ { z(1-z)-(1-z)r_1-zr_2 \ } \left\ { \log(\sqrt{r_1(1-z)+r_2z}+\sqrt{z(1-z ) } ) \right . \nonumber \\ & & \left . \left . \left . -\frac{1}{2}\log{z(1-z ) } \right\ } \right ) \right ] , \\ x_{22 - 2}(m_v , m_1 , m_2 ) & = & -\frac{\pi}{2}m_v \left [ \frac{1}{4}(1-r_1-r_2)\log\frac{m_v^2}{\mu^2 } -\frac{1}{2 } + \frac{3}{4}(r_1+r_2 ) \right . \nonumber \\ & & + \int_0 ^ 1 dz \ , \left ( -3\sqrt{z(1-z)}\sqrt{(1-z)r_1+zr_2 } + \ { 3z(1-z)-(1-z)r_1-zr_2 \ } \right . \nonumber \\ & & \left . \left . \times \left\ { \log(\sqrt{(1-z)r_1+zr_2}+\sqrt{z(1-z ) } ) - \frac{1}{2}\log(z(1-z ) ) \right\ } \right ) \right ] , \end{aligned}\ ] ] where @xmath83 . @xmath84 is the renormalization scale . here we show the explicit forms of the integrals ( [ eq : sigma2b2 ] ) , after subtracting @xmath85 divergences from @xmath86 by the scheme , and separating the mass corrections to the gauge bosons @xmath87 from @xmath88 . the @xmath64 contributions coming from eqs . ( [ eq13][eq17 ] ) are , up to the overall factor @xmath65 , @xmath89 x_{zz}(m_t ) -m_t^2 g_0(m_z , m_t , m_t ) \nonumber \\ & & + [ ( i_{3b})^2 - 2i_{3b}q_bs_w^2 + 2q_b^2s_w^4 ] x_{zz}(0 ) -c_w^2\frac{\pi}{m_z}\delta m_z^{2(tb ) } , \label{eq30}\end{aligned}\ ] ] with @xmath90 here we used the notations @xmath91 , \nonumber \\ g_0 ( m_v , m_1,m_2 ) & \equiv & x_{0 - 2}(m_v , m_1,m_2 ) + \frac{\pi}{2m_v } [ { \rm re}b_0(m_v , m_1,m_2 ) - 1/\epsilon ] , \end{aligned}\ ] ] and substituted analytic forms of the two - loop integrals ( [ xinteg ] ) at @xmath92 and at @xmath93 . analytic forms of other integrals involving @xmath69 are shown in appendix . contributions of other quarks and leptons are obtained by taking @xmath94 , where @xmath95 and , for leptons , changing ( @xmath70 , @xmath71 ) . similarly , the gauge and higgs boson contributions coming from eqs . ( [ eq18][eq21 ] ) are , up to the factor @xmath39 , @xmath96 \nonumber \\ & & + \frac{8}{3}s_w^2c_w^2 \pi m_z -8s_w^2c_w^2[3x_{0 - 1}(m_z , m_w , m_w)+2\pi^2 m_w ] , \\ % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % c_w^2\delta\sigma_{zz}^{vh } & = & -2(12c_w^4 - 4c_w^2 + 1 ) g_{22}(m_z , m_w , m_w ) + \frac{4}{3 } c_w^4 \pi m_z \nonumber\\ & & -8c_w^4 x_{0 - 1}(m_z , m_w , m_w ) -4(4c_w^2 - 1)m_w^2 [ g_0(m_z , m_w , m_w ) ] \nonumber\\ & & -2g_{22}(m_z , m_z , m_h ) + 2m_z^2 g_0(m_z , m_z , m_h ) -c_w^2\frac{\pi}{m_z}\delta m_z^{2(vh ) } . \label{eq41}\end{aligned}\ ] ] other diagrams shown in fig . e ) are also evaluated by using the threshold expansion @xcite , keeping only the @xmath48 parts . their sum , with subtracting subdivergences by the scheme and after the ( one - loop)@xmath97(one - loop ) term in eq . ( [ eq : masscor ] ) is added , is given as @xmath98 \nonumber\\ & & + 8\pi c_w^2 m_z i_3(i_3 - t_w^2 y)\log \frac{m_w^2}{\mu^2 } - 4\pi c_w^2 ( c_f - i_3 ^ 2 ) f_{zw } . \label{eq42}\end{aligned}\ ] ] here @xmath99 \nonumber\\ & & -\frac{1}{3 } ( 2+c_w^{-2})m_z \int_0 ^ 1 dz \ , z^{-3/2 } ( 1-z)^{-1/2 } [ ( c_w^2 z + 1-z)^{3/2}-1 ] % 8m^3 j_{201}(m_z , m_w ) \equiv 4m_z \left ( % \frac{1}{\epsilon}+2 - 2\log 2 - \log m_z^2 -\log m_w^2 \right ) % + 4\pi f_{zw}. \nonumber \\ % & \sim & -0.0266 m_w , & \sim & -0.027 m_w , \end{aligned}\ ] ] is the two - loop function appearing in fig . 1(c , d ) with both @xmath26 and @xmath27 bosons . we then need to add the counterterms coming from the renormalization of the parameters in the one - loop contributions ( [ eq : dm1 ] , [ eq : x1 ] ) ; @xmath100 in @xmath101 and ( @xmath102 , @xmath103 , ) in the coupling constants . we adopt the scheme where the pole masses ( @xmath104 , @xmath105 ) and the running coupling of qed @xmath106 , which are used in eq . ( [ eq10 ] ) , are chosen as the input parameters . in this scheme , the renormalization is achieved by removing the last @xmath107 terms from @xmath108 ( [ eq27 ] , [ eq38 ] ) and @xmath109 ( [ eq30 ] , [ eq41 ] ) , and adding the counterterms for ( @xmath102 , @xmath103 , ) expressed as tree - level functions of ( @xmath105 , @xmath104 , @xmath106 ) . it is checked that the final form of the two - loop @xmath73 mass correction to @xmath23 is finite and independent of the renormalization scale @xmath84 . here we comment on the mass splitting of a new heavy scalar su(2)@xmath0 multiplet @xmath110 . in contrast to the case of the fermion multiplet , direct couplings of @xmath110 to the higgs bosons , such as @xmath111 , should always exist @xcite . nevertheless , assuming that the effect of these direct couplings is negligible , we have verified that the nondecoupling @xmath73 parts of the one - loop @xcite and two - loop mass corrections @xmath4 are identical to those for the fermions in the same gauge representation . this result is quite natural in the view that the @xmath73 mass correction could be understood as the energy of the electroweak gauge fields around a static point source , and should be insensitive to the spin of the source particle @xcite . we show the numerical results of the two - loop contributions to the mass splitting within the @xmath49 fermion multiplet . as seen in eqs . ( [ eq11 ] , [ eq42 ] ) , the one - loop relation ( [ eq9 ] ) still holds with the change @xmath112 , where @xmath113 . the contribution @xmath114 of the quark - lepton subloop diagrams ( including corresponding counterterms ) is shown in fig . 2 as a function of @xmath69 . at @xmath115 gev , there is cancellation between the @xmath64 subloop contribution , shown in the dashed line , and remaining contribution with subloops of other quarks or leptons , @xmath116 mev , giving the total shift @xmath117 mev at @xmath115 gev . between fermions in a heavy su(2)@xmath0 multiplet with @xmath49 , from diagrams in fig . 1(b ) with quark and lepton subloops . solid and dashed lines denote full and @xmath64 subloop contributions , respectively . , width=529 ] the remaining contribution @xmath118 from diagrams without quarks or leptons ( again including corresponding counterterms ) is shown in fig . 3 as a function of @xmath119 . at @xmath120 gev , the shift is @xmath121 mev , smaller than the quark - lepton loops . between fermions in a heavy su(2)@xmath0 multiplet with @xmath49 , from diagrams in fig . 1(b - e ) with gauge and higgs bosons . , width=529 ] these two - loop contributions are much smaller than the @xmath73 part of the leading one - loop contribution ( [ eq : dm1 ] ) , as expected . however , for @xmath49 , it may compete with the @xmath1-dependence of the one - loop contribution ( [ eq9 ] ) which behaves like @xmath122 mev for large @xmath1 due to the accidental cancellation of the @xmath123 term in eq . ( [ eq9 ] ) . in comparison , in the case of the higgsino - like doublet @xmath124 with ( @xmath125 , @xmath126 ) , the two - loop corrections to the mass splitting @xmath127 , which is @xmath128 mev at the one - loop , is @xmath129 mev from quark and lepton loops at @xmath115 gev , and @xmath130 mev from gauge and higgs boson loops at @xmath120 gev , respectively . we have calculated the two - loop electroweak contribution to the @xmath73 correction to the masses of new heavy fermions in an su(2)@xmath0 multiplet @xmath3 , which causes gauge - symmetry - breaking and `` nondecoupling '' mass splitting within @xmath3 . analytic formula of the @xmath73 mass corrections have been presented for @xmath3 in general su(2)@xmath2u(1 ) representation . the two - loop contribution has turned out to be typically o(mev ) , which is of similar order to the @xmath1 dependence of the one - loop contribution for the @xmath49 case . in the case of the @xmath64 contributions ( [ eq27][eq30 ] ) , feynman parameter integrals for the functions ( [ xinteg ] ) can be analytically performed . for @xmath131 with @xmath132 , we have @xmath133 , \\ x_{0 - 2 } & = & \frac{\pi}{2m_v } \left [ \log\frac{m_v^2}{\mu^2 } + \log r + \frac{2}{\sqrt{4r-1}}\tan^{-1}\sqrt{4r-1 } \right ] , \\ x_{22 - 1 } & = & -\frac{\pi}{3}m_v^3 \left [ \frac{1}{4 } ( 1 - 6r ) \log\frac{m_v^2}{\mu^2 } -\frac{2}{3 } + \frac{7}{2}r \right . \nonumber \\ & & \left . + \frac{1 - 6r}{4}\log r + \frac{1}{2}(4r-1)^{3/2 } \tan^{-1}\sqrt{4r-1 } \right ] , \\ x_{22 - 2 } & = & -\frac{\pi}{2}m_v \left [ \frac{1}{4}(1 - 2r)\log\frac{m_v^2}{\mu^2 } -\frac{1}{2 } + \frac{1}{2}r \right . \nonumber \\ & & \left . + \frac{1 - 2r}{4}\log r -\frac{1}{2}\sqrt{4r-1}\tan^{-1}\sqrt{4r-1 } \right ] .\end{aligned}\ ] ] for @xmath134 and @xmath135 , @xmath136 , \\ x_{0 - 2 } & = & \frac{\pi}{2m_v } \left [ \log\frac{m_v^2}{\mu^2 } -2\sqrt{r } -r\log r + 2(r+1)\log(1+\sqrt{r } ) \right ] , \\ x_{22 - 1 } & = & -\frac{\pi}{3}m_v^3 \left [ \frac{1}{4 } ( 1 - 3r ) \log\frac{m_v^2}{\mu^2 } + \frac{1}{12}(-8 - 6\sqrt{r } + 21r + 16r^{3/2 } -3r^2 + 6r^{5/2 } ) \right . \nonumber \\ & & \left . + \frac{(r-3)r^2}{4}\log r -\frac{1}{2}(r-1)^3\log(1+\sqrt{r } ) \right ] , \\ x_{22 - 2 } & = & -\frac{\pi}{2}m_v \left [ \frac{1}{4}(1-r)\log\frac{m_v^2}{\mu^2 } + \frac{1}{12}(-6 -6\sqrt{r } + 3r + 4r^{3/2 } + 3r^2 -6r^{5/2 } ) \right . \nonumber \\ & & \left . -\frac{r^2(r-1)}{4}\log r + \frac{1}{2}(r-1)(r^2 - 1)\log(1+\sqrt{r } ) \right ] .\end{aligned}\ ] ] s. mizuta and m. yamaguchi , phys . b * 298 * ( 1993 ) 120 [ arxiv : hep - ph/9208251 ] . s. mizuta , d. ng and m. yamaguchi , phys . b * 300 * ( 1993 ) 96 [ arxiv : hep - ph/9210241 ] . c. h. chen , m. drees and j. f. gunion , phys . * 76 * ( 1996 ) 2002 [ erratum - ibid . * 82 * ( 1999 ) 3192 ] [ arxiv : hep - ph/9512230 ] ; c. h. chen , m. drees and j. f. gunion , phys . d * 55 * ( 1997 ) 330 [ erratum - ibid . d * 60 * ( 1999 ) 039901 ] [ arxiv : hep - ph/9607421 ] . g. f. giudice and a. pomarol , phys . b * 372 * ( 1996 ) 253 [ arxiv : hep - ph/9512337 ] . l. randall and r. sundrum , nucl . b * 557 * ( 1999 ) 79 [ arxiv : hep - th/9810155 ] ; g. f. giudice , m. a. luty , h. murayama and r. rattazzi , jhep * 9812 * ( 1998 ) 027 [ arxiv : hep - ph/9810442 ] . j. l. feng , t. moroi , l. randall , m. strassler and s. f. su , phys . * 83 * ( 1999 ) 1731 [ arxiv : hep - ph/9904250 ] ; t. gherghetta , g. f. giudice and j. d. wells , nucl . b * 559 * ( 1999 ) 27 [ arxiv : hep - ph/9904378 ] . j. f. gunion and s. mrenna , phys . d * 62 * ( 2000 ) 015002 [ arxiv : hep - ph/9906270 ] . j. hisano , s. matsumoto and m. m. nojiri , phys . d * 67 * ( 2003 ) 075014 [ arxiv : hep - ph/0212022 ] ; j. hisano , s. matsumoto and m. m. nojiri , phys . * 92 * ( 2004 ) 031303 [ arxiv : hep - ph/0307216 ] ; j. hisano , s. matsumoto , m. m. nojiri and o. saito , phys . d * 71 * ( 2005 ) 015007 [ arxiv : hep - ph/0407168 ] ; j. hisano , s. matsumoto , m. m. nojiri and o. saito , phys . d * 71 * ( 2005 ) 063528 [ arxiv : hep - ph/0412403 ] . h. p. nilles , phys . rept . * 110 * ( 1984 ) 1 ; h. e. haber and g. l. kane , phys . * 117 * ( 1985 ) 75 . r. foot , h. lew , x. g. he and g. c. joshi , z. phys . c * 44 * ( 1989 ) 441 . e. ma and d. suematsu , mod . lett . a * 24 * ( 2009 ) 583 [ arxiv:0809.0942 [ hep - ph ] ] . s. dimopoulos , n. tetradis , r. esmailzadeh and l. j. hall , nucl . b * 349 * ( 1991 ) 714 [ erratum - ibid . b * 357 * ( 1991 ) 308 ] ; m. sher , phys . d * 52 * ( 1995 ) 3136 [ arxiv : hep - ph/9504257 ] ; s. d. thomas and j. d. wells , phys . * 81 * ( 1998 ) 34 [ arxiv : hep - ph/9804359 ] . m. cirelli , n. fornengo and a. strumia , nucl . b * 753 * ( 2006 ) 178 [ arxiv : hep - ph/0512090 ] ; m. cirelli , a. strumia and m. tamburini , nucl . b * 787 * ( 2007 ) 152 [ arxiv:0706.4071 [ hep - ph ] ] ; m. cirelli , r. franceschini and a. strumia , nucl . phys . b * 800 * ( 2008 ) 204 [ arxiv:0802.3378 [ hep - ph ] ] . h. c. cheng , b. a. dobrescu and k. t. matchev , nucl . b * 543 * ( 1999 ) 47 [ arxiv : hep - ph/9811316 ] . d. pierce and a. papadopoulos , nucl . b * 430 * ( 1994 ) 278 [ arxiv : hep - ph/9403240 ] . g. passarino and m. veltman , nucl . b * 160 * ( 1979 ) 151 . c. amsler _ et al . _ [ particle data group ] , phys . b * 667 * ( 2008 ) 1 . w. j. marciano and a. sirlin , phys . d * 22 * ( 1980 ) 2695 [ erratum - ibid . d * 31 * ( 1985 ) 213 ] ; d. c. kennedy and b. w. lynn , nucl . b * 322 * ( 1989 ) 1 . s. p. martin , phys . d * 71 * ( 2005 ) 116004 [ arxiv : hep - ph/0502168 ] ; s. p. martin , phys . d * 72 * ( 2005 ) 096008 [ arxiv : hep - ph/0509115 ] . a. czarnecki and v. a. smirnov , phys . b * 394 * ( 1997 ) 211 [ arxiv : hep - ph/9608407 ] ; v. a. smirnov , _ `` applied asymptotic expansions in momenta and masses , '' _ springer tracts in modern physics * 177 * ( springer , berlin , 2002 ) .	new heavy particles in an su(2)@xmath0 multiplet , sometimes introduced in extensions of the standard model , have highly degenerate tree - level mass @xmath1 if their couplings to the higgs bosons are very small or forbidden . however , loop corrections may generate the gauge - symmetry - breaking mass splitting within the multiplet , which does not vanish in the large @xmath1 limit due to the threshold singularity . we calculate the electroweak contribution to the mass splitting for a heavy fermion multiplet , to the two - loop order . numerically , two - loop electroweak contributions are typically o(mev ) . tu-848 * electroweak two - loop contribution to the mass splitting within a new heavy su(2)@xmath0 fermion multiplet * + youichi yamada [ cols= " < " , ]
density functional theory(dft ) have success to predict almost all of structure of molecular and crystal , but band theory of dft explains conductivity partly . the classical theory on conductivity is based on band structure of bloch s theorem , in which charge carriers are electrons excited from valence bands to conduction bands or occupied in partially filled band . but it fails to explain the poor conductivity of mott insulators@xcite such as coo , mno , cuo , which have half - filled band structure , while cuo is the principle component of cuprate high temperature superconductor . so , a unified theory of conductivity on metals and mott insulators is one of keys for high temperature superconductivity . it became practicable after we found the relations between delocalization of electrons and types of atomic orbitals when projecting bands of crystal to atomic orbitals , because delocalized electrons was just electron mobility@xcite with respect to their original ions in lattice . then , charge carriers will be not excited electrons , but the delocalized ones near fermi surface . certainly , this way is still based on the band structure of bloch s theorem . this paper includes five parts : first we elaborated that slater type orbital(sto ) was the best function to project out the atomic components from plane wave function , in which @xmath4 was the best parameter to represent the degree of delocalization in three dimensions . secondly , the localization - delocalization duality of @xmath0 electrons and localization of @xmath1 and @xmath2 ones are found , from which conductivity and hubbard model is explained . thirdly , the mechanism of superconductivity is given based on the barrier between localization and delocalization and the phase diagram@xcite of cuprate superconductors is explained qualitatively . fourthly , we will use percolation theory@xcite to connect the microscopic and macroscopic mechanism of ( super)conductivity . at last , we will end this paper by a conclusion . to give band structure of crystal , there are two widely used methods , plane waves and atomic orbitals . what we used is the former in addition to _ ab initio _ pseudopotential@xcite , and these provide a very successful scheme to calculate the ground state properties of crystal . but only atomic basis is adequate to analysis atomic properties of crystal . we have to project out the corresponding pseudo atomic orbitals@xcite from the bands of plane wave calculations using pseudopotential . for posterior analysis of atomic properties , the best pseudo atomic orbital is single slater type orbital @xmath5@xcite , where @xmath6 is the normalization constant , @xmath7 is the spherical harmonics , @xmath8 is the radical function and @xmath4 is a parameter of @xmath9 and @xmath10 . there are four reasons for sto : ( 1 ) single @xmath4-optimised stos can almost be equivalent to plane wave basis when atoms are far enough each other , ( 2 ) single stos have the well - defined physical mean , where the @xmath9 is main quantum number , @xmath10 is angular momentum quantum number and @xmath11 is magnetic quantum number , ( 3 ) there is a single variable @xmath4 to indicate the transformation of atomic orbitals between isolated atom and atom in crystal after optimised , ( 4 ) @xmath4 is also a parameter to indicate the degree of delocalization of electrons with respect to the original ions , because the classical radius of atomic orbital is @xmath12 so that the smaller @xmath4 , the larger radius , the more delocalization . criterion should be given to optimise the variable @xmath4 . in projecting , the quality of an atomic basis is quantified by its ability to represent those eigenstates of plane wave basis , i.e. , the completeness of crystal orbitals in subspace spanned by the atomic basis , which is always less than 100% . from the rayleigh - ritz minimal principle , energy of an electronic ground state @xmath13 , where @xmath14 is hamiltonian and @xmath15 is a normalized trial function for the given number of electrons @xmath16 . after projecting to a single sto of one variable @xmath4 , the above formula will become @xmath17 $ ] so that the energy of system is a function of @xmath4 and the less incompleteness of atomic basis , the lower trial energy of system . then we could analyse electron state by the relation of incompleteness or energy and @xmath4 of sto . the electronic structures of supercell are calculated using _ ab initio _ plane - wave code cpmd3.9.1@xcite of density functional theory , where sizes of supercell are larger than 1nm@xmath181nm@xmath181 nm . the exchange - correlation functional was perdew - burke - ernzerhof@xcite ones of the generalised gradient approximation . the interaction of ion - electron was described by goedecker pseudopotential@xcite with cutoff energy of plane - wave 40 or 100 rydberg for @xmath19 or @xmath20 and @xmath21 shell respectively and shell of valence electrons is as few as possible . the data of normal crystal structures come from experiment . fig . 1a and 1b gave some typical graphs of incompleteness of atomic basis versus @xmath4 of sto in different crystal structure : @xmath0 type orbitals in body - centred cubic ( k ) , face - centred cubic ( cu ) and hexagonal close - packed ( zn ) lattice , @xmath2 type ones in body - centred cubic ( cr ) and face - centred cubic ( cu ) lattice , @xmath1 type one in diamond ( c ) . from these figures we could find that , there must be two minimums at large and small @xmath4 separated by a group of barriers for electrons in @xmath0 orbital , and only one minimum for ones in @xmath1 and @xmath2 orbitals at large @xmath4 , so that electrons of @xmath0 orbitals have localization - delocalization duality , while ones of @xmath1 and @xmath2 orbitals only is localized in three dimensional crystal lattices . localized electrons always appear in lattice and bond atoms one by one so that solids is solid , while delocalized electrons appear in metals to form long - range bonds so that metals have extensibility . by the way , the mean of `` duality '' here is analogous with the wave - particle duality in quantum mechanics for @xmath0 type sto of @xmath22 is just plane wave and the one of @xmath23 just point particle . some typical graphs of incompleteness of atomic orbital versus @xmath4 of sto . @xmath4 is in atomic unit and the @xmath24 in ( c ) is lattice parameter of supercell . ] because localization - delocalization duality of atomic orbitals is a phenomena of electronic structure of crystal on _ ab initio _ calculation , we will investigate the origin of minimums and barriers . fig . 2 is graph of incompleteness of @xmath19 orbital of cu in four atoms cell , where the insets above the curve are corresponding distribution of orbitals in two dimensional section . in these insets , the right and left are localized and delocalized @xmath19 orbitals respectively , which give sketch of electron distribution in crystal , and the middle is @xmath19 orbitals in barrier , in where electron density is the lowest . from the view of wave , the electronic density in crystal is variable periodically in three dimensions so that the @xmath0 wave , an isotropic wave in three dimensions could spread from near to far away original ion , but it will undergo a scatter just when leaving the original ion . to @xmath1 , @xmath2 and @xmath25 wave , there are two types of symmetry : @xmath1 , @xmath26 , @xmath27 waves are isotropic only in two dimensions so that the periods of electron density are not commensurable between these two dimensions and the other dimension ; the others is anisotropic , so that the periods of electron density is not commensurable among all of three dimensions . then @xmath1 , @xmath2 and @xmath25 waves would undergo scatter continuously not to spread away from the localized orbitals . this localization of @xmath1 , @xmath2 and @xmath25 wave in three dimensions gives one instance for anderson localization of disordering@xcite . now we conclude that only waves whose symmetry is fit with medium could spread , that is to say , s wave could spread in three dimensional lattice and @xmath1 , @xmath26 and @xmath27 wave could spread in two - dimensional sheet . graph of incompleteness of @xmath19 orbital of cu versus @xmath4 and radius of sto and corresponding distribution of orbitals on two dimensional section . the @xmath28 for radius is bohr radius @xmath29 nm ] this relation between delocalization of electrons and types of wave has a far - reaching consequences that charge carriers is @xmath0 wave in three dimensional conductor such as metal conductors and superconductors , and @xmath1 , @xmath2 and/or @xmath25 wave in two dimensional conductors such as graphite , organic conductors and cuprate superconductors . in mott insulators such as coo , mno , cuo , electrons near fermi surface belong to @xmath20 of transition metals and @xmath21 of o so that there are not @xmath0 but @xmath2 and/or @xmath1 type charge carriers . then mott insulators are insulators in three dimension . the above also explain experimental facts that charge carriers are @xmath0 waves in metal superconductors and @xmath2 waves in cuprate superconductors@xcite , but the @xmath2 wave of cu in our theory is not @xmath30 but @xmath26 , and @xmath21 of o could also be delocalized to contribute to charge carriers in cuo@xmath3 plane and direction perpendicular to this plane . we conjectured that @xmath30 symmetry of charge carrier was just some linear combination of @xmath31 orbitals of o in cuo@xmath3 plane . when do the delocalized electrons occur to transit mott insulators to conductors or superconductors ? that is to say , when the localized orbital , the fundamental state in bond need to be delocalized ? we knew that metallic bond is unsaturated bond of @xmath0 orbitals , and @xmath32 bond in organic conductors is also unsaturated bond of @xmath1 orbitals relative to @xmath33 bond , so that electron deficient of bond is the common characteristic of delocalization of bond as well as of electrons . so the transition of mott insulator - metal comes from electron deficient of bond , in which the doped cuprate superconductors are just some instances . to clarify the change of localization - delocalization duality in mott metal - insulator transition , we investigate an ideal mott transition : the delocalization of @xmath19 of cu for different distance of atoms in four atoms cell(fig . 1c ) , in which more energy for delocalization of electrons is needed as lattice parameter become large . and this difference of energy between delocalization and localization is just as hubbard energy in hubbard model@xcite . the lattice parameter @xmath34 nm means the corresponding `` metal '' has poor conductivity for the minimum distance of atoms is large to @xmath35 nm . and crystal of @xmath36 nm is just true copper metal . from above we concluded that the distance larger , the more incompleteness of delocalization , the less conductivity so that metal transited to insulator . as transforming the incompleteness of projecting to energy of system , graphs of incompleteness versus @xmath4 become two electronic states separated by a energy barrier , and these two states can transition each other by electron tunneling . when delocalized and localized electrons form bonds by electron pairing of different spin , there are three possible kinds of electron pair : hybrid @xmath37 and @xmath38 , localized @xmath39 and delocalized @xmath40 pair , where @xmath41 and @xmath42 are spin of electrons and subscripts @xmath43 and @xmath44 represent localized and delocalized state respectively . in high temperature , almost all of the electrons have enough energy across the barrier so that only hybrid pairs can exist . as the temperature falls , some of electrons in lowest energy of states are not able to transit and are delocalized or localized , then delocalized and localized pairs will occur because of bonding , which had been supported by two - component electrons in cuprate superconductors@xcite . it shows that the barrier between localization and delocalization determines the temperature of pairing of charge carriers , even transition temperature of superconductivity in metals . the energy barriers in metals are very low because their electron density is highly homogeneous in real space so that the temperature for prohibiting transition of some electron pairs should be very low , while in cuprate superconductors the barriers are high because their electronic structure is much more prone to inhomogeneity@xcite . one of keys of cuprate high temperature superconductivity is its `` phase diagram''(fig . 3c ) . to explain this , we should first determine the delocalized states of electrons near fermi surface as mott insulators are doped to transit to metals ( fig . 3b ) , which could be deduced from fig . the pseudogap in normal state of cuprate superconductors would come from the imbalance of electrons of localized and delocalized state . in high temperature , the difference of energy between localized and delocalized states are negligible so that partition of electrons in two states is duality and fermi surface fill in , but as temperature falls below temperature of pseudogap , this difference of energy would dominate partition , then there are more localized electrons than delocalized ones and no enough electrons for fermi surface so that this surface will be broken ( fig . so , pseudogap does correspond with a phase transition with broken fermi surface . on the other hand , more doping produces more itinerant electrons and reduces inhomogeneity so that the barrier will decreases as doping , so @xmath45 will rise first and then drop . in underdoped side , there is no enough carriers and transition temperature @xmath45 is dominated by hole concentration , and in overdoped side , the energy barrier is low so that @xmath45 is determined by energy barrier for delocalized state . schematic phase diagram of cuprate superconductors ( c ) and corresponding schematic diagrams for density of state(dos ) ( a ) and delocalization of orbitals ( b ) near fermi surface . @xmath46 is the neel temperature for three dimensional antiferromagnetic state , @xmath45 the transition temperature of superconductivity and @xmath47 the temperature below which the normal - state pseudogap opens , @xmath4-like is the corresponding parameter of delocalization in two dimensional cases . ] for superconductivity of cuprate superconductors , there is a similar critical fraction @xmath48 of hole concentration , which could be deduced from the delocalized radius of electron and percolation theory . percolation theory is the bridge between microscopic and macroscopic mechanism of ( super)conductivity , in which the delocalized orbitals were thought of ( super)conductor spheres and localized ones insulator spheres . because the radius of delocalized orbital is much larger than of localized one , we should use the ( super)conductor sphere as unit to calculate thresholds for percolation . once the fraction of carriers exceeds the corresponding threshold , insulators will transit to metal , or electron pairs will cohere to form superconductors . for example , in two - dimensional sheet of cuprate superconductivity , it is less than @xmath49 , where @xmath50 is threshold of square in site percolation , @xmath51 is theoretical minimum for radius of delocalization divided by one of localization , and @xmath52 comes from dimensionality . this threshold is compare with our critical fraction @xmath48 in phase diagram of cuprate conductors . in three dimensional metals , for example in face - centred cubic(fcc ) copper , the threshold of ( super)conductivity is less than @xmath53 , where @xmath54 is threshold of site percolation for fcc lattice , @xmath55 is minimum for radius of delocalization divided by one of localization in fig . 1b , and @xmath51 comes from dimensionality . it is compare with @xmath56@xcite , fraction of charge carriers in metal superconductivity . from above instances we knew that thresholds of percolation of ( super)conductivity were very sensitive to the dimensionality of ( super)conductors and were very small in three dimensional case so that electron pairing and its condensation of coherence occur almost simultaneously , while the critical fraction in two dimensional sheet of cuprate superconductor is rather large so that electrons pairing should occur before its condensation of coherence distinctly . so we predict that superconductors at room temperature should be ionic compound of delocalized @xmath0 electrons because they should have high barrier between delocalization and localization and small thresholds of percolation in three dimension , which may find from the other end of @xmath20 transition metal oxide such as sc@xmath3o@xmath57 , tio@xmath58 or v@xmath3o@xmath59 , because their @xmath19 and @xmath20 orbitals are both near fermi surface to form @xmath0 type conductors when insulators transit to conductors . we found the localization - delocalization duality of @xmath0 electrons in metals , from which we unified the theory of conductivity for metals , mott insulators and organic conductors . the phase diagram of cuprate superconductors also be explained by ideal mott insulator - metal transition qualitatively . we also predicted that higher temperature superconductor might be ionic compound with delocalized @xmath0 electrons . mott , _ metal - insulator transitions _ ( taylor and francis , 1990 ) . anderson , j.k . burdett , p.t . _ , * 116 * , 8808(1994 ) . j.l . tallon and j.w . loram , _ physica c _ , * 349 * , 53(2001 ) . d. stauffer and a. aharony , _ introduction to percolation theory _ , 2nd ( taylor and francis , 1992 ) . d.r.hamann , m.schlter , and c. chiang , _ phys . _ , * 43 * , 1494(1979 ) . d. sanchez - portal , e. artacho and j.m . soler , _ solid state commun . _ , * 95 * , 685(1995 ) . slater , _ phys . _ , * 36 * , 57(1930 ) . cpmd ( copyright ibm corp . 1990 - 2004 , copyright mpi fr festkrperforschung stuttgart 1997 - 2001 ) . perdew , k. burke and m. ernzerhof , _ phys . _ , * 77 * , 3865(1996 ) . c. hartwigsen , s. goedecker and j. hutter , _ phys . _ , * b58 * , 3641(1998 ) . anderson , _ phys . _ , * 109 * , 1492(1958 ) . v.p . mineev and k.v . somokhin , _ introduction to unconventional superconductivity _ , ( gordon and breach , 1999 ) . stevens , et al . _ , * 78 * , 2212(1997 ) . v.j . emery and s.a . kivelson , _ physica c _ , * 209 * , 597(1993 ) . norman , et al . _ nature _ , * 392 * , 157(1998 ) . a. mourachkine , _ room temperature superconductivity _ , ( cambridge international science publishing , 2004 ) .	when wavefunction of crystal was projected out in atomic basis , we found that electrons of @xmath0 orbital had localization - delocalization duality , and the ones of @xmath1 and @xmath2 orbital were only localized in three dimensional crystal lattice . the existence of @xmath0 type delocalized electrons is based on three dimensional ordering of electron density in metals . then charge carriers of metals for conductivity are just such delocalized electrons so that the ones for superconductivity are electron pairs of @xmath0 wave . this way could also be generalized to two dimensional case for cuo@xmath3 plane of cuprate superconductor to explain its phase diagram qualitatively . we also gave some _ ab initio _ explain for hubbard model .
matrices with rapidly decaying singular values frequently appear in computational mathematics . such matrices are numerically of low rank and this is exploited in applications such as particle simulations @xcite , model reduction @xcite , boundary element methods @xcite , and matrix completion @xcite . however , it can be theoretically challenging to fully explain why low rank techniques are so effective in practice . in this paper , we derive explicit bounds on the singular values of matrices with displacement structure and in doing so justify many of the low rank techniques that are being employed on such matrices . let @xmath10 with @xmath11 , @xmath12 , and @xmath13 , we say that @xmath14 has an @xmath15-displacement rank of @xmath16 if @xmath14 satisfies the sylvester matrix equation given by @xmath17 for some matrices @xmath18 and @xmath19 . matrices with displacement structure include toeplitz ( @xmath20 ) , hankel ( @xmath20 ) , cauchy ( @xmath21 ) , krylov ( @xmath21 ) , and vandermonde ( @xmath21 ) matrices , as well as pick ( @xmath20 ) , sylvester ( @xmath20 ) , and lwner ( @xmath20 ) matrices . fast algorithms for computing matrix - vector products and for solving systems of linear equations can be derived for many of these matrices by exploiting @xcite . in this paper , we use the displacement structure to derive explicit bounds on the singular values of matrices that satisfy by using an extremal problem for rational functions from complex approximation theory . in particular , we prove that the following inequality holds ( see theorem [ thm : ratio ] ) : @xmath22 where @xmath23 denote the singular values of @xmath14 and @xmath24 is the zolotarev number for complex sets @xmath25 and @xmath26 that depend on @xmath27 and @xmath28 . researchers have previously exploited the connection between the sylvester matrix equation and zolotarev numbers for selecting algorithmic parameters in the alternating direction implicit ( adi ) method @xcite , and others have demonstrated that the singular values of matrices satisfying certain sylvester matrix equations have rapidly decaying singular values @xcite . here , we derive explicit bounds on all the singular values of structured matrices . table [ tab : summarytable ] summarizes our main singular value bounds . cccc matrix class & notation & singular value bound & ref . + ' '' '' pick & @xmath29 & @xmath30 & sec . [ sec : pickmatrices ] + cauchy & @xmath31 & @xmath32 & sec . [ sec : cauchymatrices ] + lwner & @xmath33 & @xmath34 & sec . [ sec : loewnermatrices ] + krylov , herm . @xmath35 & @xmath36 & sec . [ sec : vandermondematrices ] + real vandermonde & @xmath37 & @xmath38 & sec . [ sec : vandermondematrices ] + pos . semidef . hankel & @xmath2 & @xmath39 & sec . [ sec : posdefhankel ] + not every matrix with displacement structure is numerically of low rank . for example , the identity matrix is a full rank toeplitz matrix and the exchange matrix exchange matrix @xmath14 is obtained by reversing the order of the rows of the @xmath1 identity matrix , i.e. , @xmath40 for @xmath41 . ] is a full rank hankel matrix . the properties of @xmath27 and @xmath28 in are crucial . if @xmath27 and @xmath28 are normal matrices , then one expects @xmath14 to be numerically of low rank only if the eigenvalues of @xmath27 and @xmath28 are well - separated ( see theorem [ thm : ratio ] ) . if @xmath27 and @xmath28 are both not normal , then as a general rule spectral sets for @xmath27 and @xmath28 should be well - separated ( see corollary [ cor : ratioextension ] ) . by the eckart young theorem ( * theorem 2.4.8 ) , singular values measure the distance in the spectral norm from @xmath14 to the set of matrices of a given rank , i.e. , @xmath42 for an @xmath8 , we say that the _ @xmath43-rank _ of a matrix @xmath14 is @xmath0 if @xmath0 is the smallest integer such that @xmath44 . that is , @xmath45 thus , we may approximate @xmath14 to a precision of @xmath46 by a rank @xmath47 matrix . an immediate consequence of explicit bounds on the singular values of certain matrices is a bound on the @xmath43-rank . table [ tab : summaryranktable ] summarizes our main upper bounds on the @xmath43-rank of matrices with displacement structure . cccc matrix class & notation & upper bound on @xmath48 & ref . + ' '' '' pick & @xmath29 & @xmath49 & sec . [ sec : pickmatrices ] + cauchy & @xmath31 & @xmath50 & sec . [ sec : cauchymatrices ] + lwner & @xmath33 & @xmath51 & sec . [ sec : loewnermatrices ] + krylov , herm . arg . & @xmath35 & @xmath52 & sec . [ sec : vandermondematrices ] + real vandermonde & @xmath37 & @xmath52 & sec . [ sec : vandermondematrices ] + pos . semidef . hankel & @xmath2 & @xmath53 & sec . [ sec : posdefhankel ] + zolotarev numbers have already proved useful for deriving tight bounds on the condition number of matrices with displacement structure @xcite . for example , the first author proved that a real @xmath1 positive definite hankel matrix , @xmath2 , with @xmath54 , is exponentially ill - conditioned @xcite . that is , @xmath55 and that this bound can not be improved by more than a factor of @xmath56 times a modest constant . the hilbert matrix given by @xmath57 , for @xmath58 , is the classic example of an exponentially ill - conditioned positive definite hankel matrix ( * ? ? ? ( 3.35 ) ) . similar exponential ill - conditioning has been shown for certain krylov matrices and real vandermonde matrices @xcite . this paper extends the application of zolotarev numbers to deriving bounds on the singular values of matrices with displacement structure , not just the condition number . the bounds we derive are particularly tight for @xmath59 , where @xmath60 is small with respect to @xmath56 . improved bounds on @xmath59 when @xmath61 may be possible with the ideas found in @xcite . nevertheless , our interest here is to justify the application of low rank techniques on matrices with displacement structure by proving that such matrices are often well - approximated by low rank matrices . the bounds that we derive are sufficient for this purpose . for an integer @xmath0 , let @xmath62 denote the set of irreducible rational functions of the form @xmath63 , where @xmath64 and @xmath65 are polynomials of degree at most @xmath0 . given two closed disjoint sets @xmath66 , the corresponding zolotarev number , @xmath24 , is defined by @xmath67 where the infinum is attained for some extremal rational function . as a general rule , the number @xmath24 decreases rapidly to zero with @xmath0 if @xmath25 and @xmath26 are sets that are disjoint and well - separated . zolotarev numbers satisfy several immediate properties : for any sets @xmath25 and @xmath26 and integers @xmath0 , @xmath68 , and @xmath69 , one has @xmath70 , @xmath71 , @xmath72 and @xmath73 . they also satisfy @xmath74 if @xmath75 and @xmath76 as well as @xmath77 , where @xmath78 is any mbius transformation @xcite . as @xmath79 the value for @xmath24 is known asymptotically to be @xmath80 where @xmath81 is the logarithmic capacity of a condenser with plates @xmath25 and @xmath26 @xcite . to readers that are not familiar with zolotarev numbers , it may seem that trades a difficult task of directly bounding the singular values of a matrix @xmath14 with a more abstract task of understanding the behavior of @xmath24 . however , zolotarev numbers have been extensively studied in the literature @xcite and for certain sets @xmath25 and @xmath26 the extremal rational function is known explicitly ( * ? ? ? 50 ) ( see section [ sec : zolotarev ] ) . our major challenge for bounding singular values is to carefully select sets @xmath25 and @xmath26 so that one can use complex analysis and mbius transformations to convert the associated extremal rational approximation problem in into one that has an explicit bound . the paper is structured as follows . in section [ sec : zolotarevbounds ] we prove , giving us a bound on the singular values of matrices with displacement structure in terms of zolotarev numbers . in section [ sec : zolotarev ] we derive new sharper bounds on @xmath82,[a , b])$ ] when @xmath83 by correcting an infinite product formula from lebedev ( see theorem [ thm : zproduct ] and corollary [ cor : zbound ] ) . in section [ sec : pickcauchyloewnermatrices ] we derive explicit bounds on the singular values of pick , cauchy , and lwner matrices . in section [ sec : hankelmatrices ] we tackle the challenging task of showing that all real vandermonde and positive definite hankel matrices have rapidly decaying singular values and can be approximated , up to an accuracy of @xmath8 , by a rank @xmath84 matrix . in appendix [ appendix : correction ] we further detail the unfortunate consequences of the erroneous infinite product formula from lebedev and present corrected results . let @xmath14 be an @xmath85 matrix with @xmath11 that satisfies . we show that the singular values of @xmath14 can be bounded from above in terms of zolotarev numbers . first , we assume that @xmath27 and @xmath28 in are normal matrices and later remove this assumption in corollary [ cor : ratioextension ] . in theorem [ thm : ratio ] the spectrum ( set of eigenvalues ) of @xmath27 and @xmath28 is denoted by @xmath86 and @xmath87 , respectively . was presented by the first author at the cortona meeting on structured numerical linear algebra in 2008 @xcite as well as several other locations . the statement has not appeared in a publication by the first ( or second ) author before . similar statements based on the presentation have appeared in ( * ? ? ? * theorem 2.1.1 ) , ( * ? ? ? * theorem 4 ) , and most recently ( * ? ? ? * theorem 4.2 ) . ] let @xmath12 and @xmath13 be normal matrices with @xmath11 and let @xmath25 and @xmath26 be complex sets such that @xmath88 and @xmath89 . suppose that the matrix @xmath10 satisfies @xmath90 where @xmath91 is an integer . then , for @xmath92 the singular values of @xmath14 satisfy @xmath93 where @xmath24 is the zolotarev number in . [ thm : ratio ] let @xmath94 and @xmath95 be polynomials of degree at most @xmath0 . first , we show that @xmath96 suppose that @xmath97 and @xmath98 , where @xmath99 , then @xmath100 in the last sum we have terms of the form @xmath101 , with @xmath102 . by adding together the terms occurring in @xmath103 for general degree @xmath0 polynomials @xmath64 and @xmath65 , we conclude that there exist coefficients @xmath104 such that @xmath105 this shows that the rank of @xmath103 is bounded above by @xmath0 times the number of columns of @xmath106 , proving . now , let @xmath107 , where @xmath64 and @xmath65 are polynomials of degree @xmath0 so that @xmath108 is the extremal rational function for the zolotarev number in . this means that @xmath94 and @xmath95 are not zero on @xmath26 and @xmath25 , respectively , so that @xmath109 and @xmath110 are invertible matrices . from we know that @xmath111 has rank at most @xmath112 and hence , the matrix @xmath113 is of rank at most @xmath112 . let @xmath114 be the best rank @xmath115 approximation to @xmath14 in @xmath116 and let @xmath117 . since @xmath118 is of rank at most @xmath115 , @xmath119 is of rank at most @xmath120 . this implies that @xmath121 where in the last inequality we used the relation @xmath122 . finally , since @xmath27 and @xmath28 are normal we have @xmath123 and @xmath124 . we conclude by the definition of @xmath108 that @xmath125 as required . theorem [ thm : ratio ] shows that if @xmath27 and @xmath28 are normal matrices in , then the singular values decay at least as fast as @xmath126 in . in particular , when @xmath86 and @xmath87 are disjoint and well - separated we expect @xmath126 to decay rapidly to zero and hence , so do the singular values of @xmath14 . for those readers that are familar with the adi method @xcite , an analogous proof of theorem [ thm : ratio ] is to run the adi method for @xmath0 steps with shift parameters given by the zeros and poles of the extremal rational function for @xmath24 . by doing this one constructs a rank @xmath112 approximant @xmath127 for @xmath14 , which shows that @xmath128 . the connection between zolotarev numbers and the optimal parameter selection for the adi method has been previously exploited @xcite . we have presented the above proof here because it does not require knowledge of the adi method . for matrices @xmath27 and @xmath28 that are not normal , theorem [ thm : ratio ] can be extended by using @xmath129-spectral sets @xcite . given a matrix @xmath27 , a complex set @xmath25 is said to be a @xmath129-spectral set for @xmath27 if the spectrum @xmath86 of @xmath27 is contained in @xmath25 and the inequality @xmath130 holds for every bounded rational function on @xmath25 , where @xmath131 . similar extensions have been noted when @xmath132 in and the sets @xmath25 and @xmath26 are taken to be the fields of values for @xmath27 and @xmath28 , respectively @xcite . we have the following extension of theorem [ thm : ratio ] : suppose that the assumptions of theorem [ thm : ratio ] hold , except that the matrices @xmath27 and @xmath28 are not necessarily normal . also suppose that @xmath25 and @xmath26 are @xmath129-spectral sets for @xmath27 and @xmath28 for some fixed constant @xmath133 , respectively . then , we have @xmath134 . [ cor : ratioextension ] it is only the first inequality in of the proof of theorem [ thm : ratio ] that requires @xmath27 and @xmath28 to be normal matrices . when @xmath27 is not a normal matrix , then the inequality @xmath135 may not hold . instead , we replace it by the @xmath129-spectral set bound given by @xmath130 . note that since @xmath94 and @xmath136 are not zero on @xmath26 and @xmath25 , respectively , one can show via the schur decomposition that @xmath109 and @xmath110 are invertible matrices . theorem [ thm : ratio ] and corollary [ cor : ratioextension ] provide bounds on the singular values of @xmath14 in terms of zolotarev numbers . therefore , to derive analytic bounds on the singular values of matrices with displacement structure , we must now calculate explicit bounds on zolotarev numbers a topic that fortunately is extensively studied . in this section , we derive explicit lower and upper bounds for the zolotarev numbers @xmath137,[a , b ] ) , \qquad 0<a < b<\infty,\ ] ] which we use in sections [ sec : pickcauchyloewnermatrices ] and [ sec : hankelmatrices ] . the sharpest bounds that we are aware of in the literature take the form @xmath138 see ( * ? ? ? * theorem 1 ) for the lower bound , and ( * ? ? ? ( 2.3 ) ) for the upper bound . there are also bounds obtained directly from an infinite product formula for @xmath139 ( * ? ? ? * ( 1.11 ) ) ; unfortunately , the original product formula in ( * ? ? ? * ( 1.11 ) ) contains typos and the erroneous formula has been copied elsewhere , for example , ( * ? ? ? * ( 4.1 ) ) , ( * ? ? ? * ( 3.17 ) ) , and ( * ? ? ? * sec . 4 ) . for more details . ) ] we prove a corrected infinite product formula in theorem [ thm : zproduct ] and further discuss the typos in appendix [ appendix : correction ] . the value of @xmath140 in is related to the logarithmic capacity of a condenser with plates @xmath141 $ ] and @xmath142 $ ] : @xmath143,[a , b])}\right ) , \qquad \rho=\exp\left(\frac{\pi^2}{2\mu(a / b)}\right),\ ] ] where @xmath144 is the grtzsch ring function , and @xmath129 is the complete elliptic integral of the first kind ( * ? ? ? * ( 19.2.8 ) ) @xmath145 the bounds in are not asymptotically sharp , and in corollary [ cor : zbound ] we show that the constant of @xmath146 " in the upper bound can be replaced by @xmath147 " . for a proof of this sharper upper bound , we first return to the work of lebedev @xcite and derive a corrected infinite product formula for @xmath148 . [ thm : zproduct ] let @xmath149 be an integer and @xmath83 . then for @xmath150,[a , b])$ ] we have @xmath151 where @xmath152 is the grtzsch ring function . we start by establishing a product formula for the inverse of @xmath153 that is apparently not widely known . for @xmath154 set @xmath155 . since @xmath156 , we have that @xmath157 and from ( * ? ? ? * ( 22.2.2 ) ) we obtain @xmath158 here , @xmath159 and @xmath160 are the classical theta functions ( * ? ? ? * ( 20.2.2 ) & ( 20.2.3 ) ) and the second equality above is derived from the infinite product formula in ( * ? ? ? * ( 20.4.3 ) & ( 20.4.4 ) ) . in order to deduce an explicit product formula for @xmath148 , we first note that the value of @xmath161 is extensively reviewed by akhiezer , . ] see ( * ? ? ? 51 ) , ( * ? ? ? * tab . 1 & 2 , p. 150 7 & 8) , and ( * ? ? ? this value is equal to @xcite , for some @xmath162 , @xmath163 here , there is a unique @xmath162 since the grtzsch ring function @xmath164\rightarrow[0,\infty]$ ] is a strictly decreasing bijection . next , we recall that gauss transformation ( * ? ? ? xxi ) and landen s transformation ( * ? ? ? xx ) are given by @xmath165 from which we conclude that @xmath166 therefore , from we have @xmath167 where @xmath140 is given in . the infinite product formula for @xmath148 follows by setting @xmath168 and @xmath169 in . the infinite product in theorem [ thm : zproduct ] can be estimated by observing that @xmath170 for all @xmath171 . this leads to the following simple upper and lower bounds which are sufficient for the purpose of our paper . let @xmath149 be an integer and @xmath83 . then for @xmath150,[a , b])$ ] we have @xmath172 where @xmath152 is the grtzsch ring function . [ cor : zbound ] corollary [ cor : zbound ] shows that @xmath173 is an asymptotically sharp upper bound in the sense that the geometric decay rate and the constant 4 " can not be improved if one hopes for the bound to hold for all @xmath0 . however , this does not necessarily imply that our derived singular value inequalities are asymptotically sharp . on the contrary , they are usually not . for asymptotically sharp singular value bounds , we expect that one must consider discrete zolotarev numbers , i.e. , @xmath126 in theorem [ thm : ratio ] , which are more subtle to bound and are outside the scope of this paper . we often prefer the following slightly weaker bound that does not contain the grtzsch ring function : @xmath174,[a , b ] ) \leq 4\left[\exp\left(\frac{\pi^2}{2\log(4b / a)}\right)\right]^{-2k},\qquad 0<a < b<\infty,\ ] ] which is obtained by using the bound @xmath175 , see ( * ? ? ? * ( 19.9.5 ) ) . this makes our final bounds on the singular values and @xmath43-rank of matrices with displacement rank more intuitive to those readers that are less familiar with the grtzsch ring function . later , in section [ sec : hankelmatrices ] we will need to use properties of an extremal rational function for @xmath176,[a , b])$ ] and we proof them now . zolotarev @xcite studied the value @xmath148 and gave an explicit expression for the extremal function for @xmath148 ( see ) by showing an equivalence to the problem of best rational approximation of the sign function on @xmath141\cup[a , b]$ ] . we now repeat this to derive the desired properties of the extremal rational function . let @xmath149 be an integer and @xmath83 . there exists an extremal function @xmath177 for @xmath176,[a , b])$ ] such that 1 . for @xmath178 $ ] , we have @xmath179 , 2 . for @xmath180 , we have @xmath181 , and 3 . for @xmath182 , we have @xmath183 . [ thm : zextremalrational ] we give an explicit expression for an extremal function for @xmath148 by deriving it from the best rational approximation of the sign function on @xmath141\cup [ a , b]$ ] . according to ( * ? ? ? 50 & 51 , p. 144 , line 6 ) we have @xmath184\cup[a , b]}\left\|{\rm sgn}(z ) - r(z)\right\| = \frac{2\sqrt{z_k}}{1+z_k},\qquad { \rm sgn}(z ) = \begin{cases}1,&z\in [ a , b],\\ -1,&z\in [ -b ,- a ] , \end{cases}\ ] ] where the infimum is attained by the rational function ( * ? ? ? 51 , tab . 2 , no . 7 & 8) @xmath185 here , @xmath106 is a real constant selected so that @xmath186 equioscillates on @xmath141\cup[a , b]$ ] , @xmath187 , and @xmath188 is the first jacobian elliptic function . zolotarevsign ( 50,0 ) @xmath189 zolotarevbounds ( 50,0 ) @xmath0 ( 25,30 ) ( 52,29 ) ( 67,35 ) figure [ fig : zolotarevbounds ] ( left ) shows the error between the sign function on @xmath190\cup[1,10]$ ] and its best @xmath191 rational approximation , which equioscillates @xmath192 times on @xmath190 $ ] and @xmath193 $ ] to confirm its optimality . in order to construct an extremal function for @xmath148 with the required properties , we observe from that @xmath106 and @xmath194 are real , and thus * @xmath195 is real - valued for @xmath196 , * @xmath197 is purely imaginary for @xmath198 , and * @xmath195 is an odd function on @xmath199 , i.e. , @xmath200 for @xmath198 . as a consequence , the rational function given by @xmath201 is real - valued for @xmath196 with @xmath181 , and of modulus @xmath202 on the imaginary axis . finally , as @xmath203 takes values in the interval @xmath204 = \left[\frac{-(1+\sqrt{z_k})^2}{1+z_k } , \frac{-(1-\sqrt{z_k})^2}{1+z_k}\right]\ ] ] for @xmath178 $ ] , we have @xmath205,\ ] ] implying that @xmath206 for @xmath178 $ ] . hence , using @xmath181 we have @xmath207 showing that @xmath208 is extremal for @xmath82,[a , b])$ ] , as required . figure [ fig : zolotarevbounds ] ( right ) demonstrates the upper bound in corollary [ cor : zbound ] when @xmath209 , @xmath210 , @xmath211 . in section [ sec : pickcauchyloewnermatrices ] we combine our upper bound on the singular values in theorem [ thm : ratio ] with our upper bound on zolotarev numbers to derive explicit bounds on the singular values of certain pick , cauchy , and lwner matrices . in this section we bound the singular values of pick ( see section [ sec : pickmatrices ] ) , cauchy ( see section [ sec : cauchymatrices ] ) , and lwner ( see section [ sec : loewnermatrices ] ) matrices . in view of theorem [ thm : ratio ] and corollary [ cor : zbound ] , our first idea is to construct matrices @xmath27 and @xmath28 so that the rank of @xmath212 is small with the additional hope that @xmath86 and @xmath87 are contained in real and disjoint intervals . for the three classes of matrices in this section , this first idea works out under mild `` separation conditions '' . in section [ sec : hankelmatrices ] the more challenging cases of krylov , real vandermonde , and real positive definite hankel matrices are considered . an @xmath1 matrix @xmath29 is called a pick matrix if there exists a vector @xmath213 , and a collection of real numbers @xmath214 from an interval @xmath142 $ ] with @xmath83 such that @xmath215 all pick matrices satisfy the following sylvester matrix equation : @xmath216 where @xmath217 . since diagonal matrices are normal matrices and in this case the spectrum of @xmath218 is contained in @xmath142 $ ] , we have the following bounds on the singular values of @xmath29 . let @xmath29 be the @xmath1 pick matrix in . then , for @xmath92 we have @xmath219^{-2k } \sigma_j(p_n),\qquad 1\leq j+2k\leq n,\ ] ] where @xmath220 is the grtzsch ring function ( see section [ sec : zolotarev ] ) . the bound remains valid , but is slightly weaken , if @xmath221 is replaced by @xmath222 . [ cor : pickmatrices ] from , we know that @xmath223 , @xmath224 , @xmath225 , @xmath226 $ ] , and @xmath227 $ ] in theorem [ thm : ratio ] . therefore , for @xmath92 we have @xmath228 , @xmath229 . the result follows from the upper bound in corollary [ cor : zbound ] . there are two important consequences of corollary [ cor : pickmatrices ] : ( 1 ) pick matrices are usually ill - conditioned unless @xmath230 is large and/or @xmath56 is small and ( 2 ) all pick matrices can be approximated , up to an accuracy of @xmath231 with @xmath8 , by a rank @xmath232 matrix . more precisely , for any pick matrix in we have @xmath233^{2\lceil \tfrac{n}{2}-1\rceil } , \label{eq : pickconditionnumber}\ ] ] where for an even integer @xmath56 we used @xmath234 . moreover , by setting @xmath0 to be the smallest integer so that @xmath235 , we find the following bound on the @xmath43-rank of @xmath29 ( see ): @xmath236 in both and , the bound can be slightly improved by replacing the @xmath222 term by @xmath221 . previously , bounds on the minimum and maximum singular values of pick matrices were derived under the additional assumption that @xmath29 is a positive definite matrix @xcite . figure [ fig : pickcauchymatrices ] ( left ) demonstrates the bound in corollary [ cor : pickmatrices ] on three @xmath237 pick matrices . the black line bounding the singular values has a stepping behavior because the inequality in corollary [ cor : pickmatrices ] for @xmath238 only bounds odd indexed singular values of @xmath29 and to bound @xmath239 we use the trivial inequality @xmath240 . at this time we can offer little insight into why the singular values of the tested pick matrices also have a similar stepping behavior . pickmatrices ( 35,0 ) @xmath241 ( 28,30 ) ( 57,26 ) ( 70,35 ) cauchymatrices ( 33,0 ) @xmath242 ( 31,30 ) ( 42,34 ) ( 65,33 ) an @xmath85 matrix @xmath31 with @xmath11 is called a ( generalized ) cauchy matrix if there exists vectors @xmath243 and @xmath244 , points @xmath245 on an interval @xmath142 $ ] with @xmath246 , and points @xmath247 ( all distinct from @xmath248 ) in an interval @xmath249 $ ] with @xmath250 such that @xmath251 generalized cauchy matrices satisfy the following sylvester matrix equation : @xmath252 where @xmath253 and @xmath254 . if we make the further assumption that either @xmath255 or @xmath256 so that the intervals @xmath142 $ ] and @xmath249 $ ] are disjoint , then we can bound the singular values of @xmath31 . this separation condition " is an extra assumption on cauchy matrices that simplifies the analysis . if the intervals @xmath142 $ ] and @xmath249 $ ] overlapped , then one would have to consider discrete zolotarev numbers to estimate the singular values and we want to avoid this in this paper . let @xmath31 be an @xmath85 cauchy matrix in with @xmath11 and either @xmath255 or @xmath256 . then , @xmath257^{-2k}\sigma_j(c_{m , n } ) , \qquad 1\leq j+k\leq n,\ ] ] where @xmath258 is the absolute value of the cross - ratio , @xmath259 , @xmath260 , and @xmath261 the cross - ratio is given by @xmath262 . ] of @xmath263 , @xmath259 , @xmath260 , and @xmath261 . if @xmath264 and @xmath265 , then @xmath266 . the bound remains valid , but is slightly weaken , if @xmath267 is replaced by @xmath268 . [ cor : cauchymatrices ] from , we know that @xmath223 , @xmath269 , @xmath270 , @xmath226 $ ] , and @xmath271 $ ] in theorem [ thm : ratio ] . therefore , we conclude that @xmath272 for @xmath273 . the value of @xmath24 is invariant under mbius transformations . that is , if @xmath274 is a mbius transformation , then @xmath24 and @xmath275 are equal . therefore , we can transplant @xmath142\cup[c , d]$ ] onto @xmath276\cup[1,\alpha]$ ] for some @xmath277 using a mbius transformation . if @xmath255 , then the transformation satisfies @xmath278 , @xmath279 , @xmath280 , @xmath281 . since @xmath78 is a mbius transformation the cross - ratio of the four collinear points @xmath263 , @xmath259 , @xmath260 , and @xmath261 equals the cross - ratio of @xmath282 , @xmath283 , @xmath284 , and @xmath285 . hence , if @xmath255 or @xmath256 then we know that @xmath286 must satisfy @xmath287 therefore , by solving the quadratic and noting that @xmath288 we have @xmath289 from corollary [ cor : zbound ] , we conclude that @xmath290^{-2k}\sigma_j(c_{m , n } ) , \qquad 1\leq j+k\leq n.\ ] ] by gauss transformation in , we note that @xmath291 and the result follows . it is interesting to observe that the decay rate of the singular values of cauchy matrices only depends on the absolute value of the cross - ratio of @xmath263 , @xmath259 , @xmath260 , and @xmath261 . hence , the separation " of two real intervals @xmath142 $ ] and @xmath249 $ ] for the purposes of singular value estimates is measured in terms of the cross - ratio of @xmath263 , @xmath259 , @xmath260 , and @xmath261 , not the separation distance @xmath292 . corollary [ cor : cauchymatrices ] shows that the cauchy matrix in ( when @xmath255 or @xmath256 ) has an @xmath43-rank of at most @xmath293 where @xmath258 is absolute value of the cross - ratio of @xmath263 , @xmath259 , @xmath260 , and @xmath261 . bounds on the numerical rank of the cauchy matrix have also been obtained via the cauchy function , i.e. , @xmath294 on @xmath142\times [ c , d]$ ] , by exploiting the hierarchical low rank structure of @xmath31 @xcite ( for more details , see ( * ? ? ? * chapter 3 ) ) . furthermore , when @xmath295 ( and @xmath255 or @xmath256 ) we have a lower bound on the condition number of @xmath296 : @xmath297^{2(n-1)},\qquad \gamma = \frac{\|c - a\| \|d - b\|}{\|c - b\| \|d - a\|}.\ ] ] corollary [ cor : cauchymatrices ] also includes the important hilbert matrix , i.e. , @xmath57 for @xmath58 . by setting @xmath298 , @xmath299 , and @xmath300 , the matrix in is the hilbert matrix . in particular , corollary [ cor : cauchymatrices ] with @xmath142 = [ -n+1/2,-1/2]$ ] and @xmath249 = [ 1/2,n-1/2]$ ] shows that @xmath301^{-2k}\sigma_1(h_n),\qquad 1\leq k\leq n-1 . \label{eq : hilbertmatrix}\ ] ] therefore , the hilbert matrix can be well - approximated by a low rank matrix and has exponentially decaying singular values . on @xmath142\times [ a , b]$ ] with @xmath83 has exponentially decaying singular values @xcite . even though there is an error in ( 4 ) in the infinite product formula and the stated lower bound ( see appendix [ appendix : correction ] ) , we believe the proved upper bound in ( * ? ? ? * ( 15 ) ) is correct . ] in particular , it has an @xmath43-rank of at most @xmath302 . the hilbert matrix is an example of a real positive definite hankel matrix and in section [ sec : hankelmatrices ] we show that bounds similar to hold for the singular values of all such matrices . figure [ fig : pickcauchymatrices ] ( right ) demonstrates the bound in corollary [ cor : cauchymatrices ] on three @xmath1 cauchy matrices , where @xmath303 . in practice , the derived bound is relatively tight for singular values @xmath304 when @xmath60 is small with respect the @xmath56 . an @xmath1 matrix @xmath33 is called a lwner matrix if there exist vectors @xmath305 , points @xmath214 in @xmath142 $ ] with @xmath246 , and points @xmath247 ( all different from @xmath306 ) in @xmath249 $ ] with @xmath250 such that @xmath307 in the special case when @xmath308 and @xmath309 , a lwner matrix is a pick matrix ( see section [ sec : pickmatrices ] ) . lwner matrices satisfy the sylvester matrix equation given by @xmath310 where @xmath217 . from theorem [ thm : ratio ] we can bound the singular values of @xmath33 provided that @xmath142 $ ] and @xmath249 $ ] are disjoint , i.e. , either @xmath255 or @xmath256 . we emphasis that the separation condition of the intervals @xmath142 $ ] and @xmath249 $ ] if an extra assumption on a lwner matrix that allows us to proceed with the methodology we have developed . let @xmath33 be an @xmath1 lwner matrix in with @xmath255 or @xmath256 . then , for @xmath92 we have @xmath311^{-2k}\sigma_j(l_n ) , \qquad 1\leq j+2k\leq n,\ ] ] where @xmath258 is the absolute value of the cross - ratio of @xmath263 , @xmath259 , @xmath260 , and @xmath261 ( see ) . if @xmath264 and @xmath265 , then @xmath266 . the bound remains valid , but is slightly weaken , if @xmath267 is replaced by @xmath268 . [ cor : loewnermatrices ] the same argument as in corollary [ cor : cauchymatrices ] , but with @xmath20 . corollary [ cor : loewnermatrices ] shows that many lwner matrices can be well - approximated by low rank matrices with @xmath312 and are exponentially ill - conditioned . the three types of matrices considered in section [ sec : pickcauchyloewnermatrices ] allowed for direct applications of theorem [ thm : ratio ] and corollary [ cor : zbound ] . in this section , we consider the more challenging tasks of bounding the singular values of krylov matrices with hermitian arguments , real vandermonde matrices , and real positive definite hankel matrices . an @xmath313 matrix @xmath35 with @xmath11 is said to be a krylov matrix with hermitian argument if there exists a hermitian matrix @xmath12 and a vector @xmath314 such that @xmath315 vandermonde matrices of size @xmath85 with real abscissas @xmath316 , i.e. , @xmath317 , are also krylov matrices with @xmath318 and @xmath319 . krylov matrices satisfy the following sylvester matrix equation : @xmath320 where @xmath243 and @xmath321 . since @xmath27 is a normal matrix , we attempt to use theorem [ thm : ratio ] to bound the singular values of @xmath35 . for the analysis that follows , we require that @xmath56 is an even integer . this is not a loss of generality because by the interlacing theorem for singular values @xcite . to see this , let @xmath322 be the @xmath323 krylov matrix obtained from @xmath35 by removing its last column . if @xmath56 is odd , then . let @xmath324 be a best rank @xmath325 approximation to @xmath322 so that @xmath326 and consider @xmath14 obtained from @xmath324 by concatenating ( on the right ) the last column of @xmath35 . then , the rank of @xmath14 is at most @xmath327 and hence , @xmath328 . ] @xmath329 and one can bound @xmath330 instead . from now on in this section we will assume that @xmath56 is an even integer . the sylvester matrix equation in contains matrices @xmath27 and @xmath331 , which are both normal matrices . the eigenvalues of @xmath27 are contained in @xmath199 and the eigenvalues of @xmath331 are the @xmath56 ( shifted ) roots of unity , i.e. , @xmath332 since @xmath56 is even , the spectrum of @xmath331 and the real line are disjoint . using theorem [ thm : ratio ] we find that for @xmath92 and @xmath273 @xmath333 where @xmath334 and @xmath335 are complex sets defined by @xmath336\ } , \quad f_- = \ { e^{it } : t \in [ - \pi + \tfrac{\pi}{n } , - \tfrac{\pi}{n}]\}. \label{eq : fplusfminus}\ ] ] figure [ fig : zolotarevforhankel ] shows the two sets @xmath25 and @xmath26 in the complex plane . as @xmath337 the sets @xmath334 and @xmath335 approach the real line , suggesting that our bound on the singular values must depend on @xmath56 somehow . our task is to bound the quantity @xmath338 a zolotarev number that is not immediately related to one of the form @xmath82,[a , b])$ ] . ; ( 5.9754 , 3.3129 ) arc ( 9.000:171.000:2 ) ; ( 2.0246 , 2.6871 ) arc ( 189.000:351.000:2 ) ; ( 5.9754,2.6871 ) circle ( .1 cm ) ; ( 5.7820,2.0920 ) circle ( .1 cm ) ; ( 5.4142,1.5858 ) circle ( .1 cm ) ; ( 4.9080,1.2180 ) circle ( .1 cm ) ; ( 4.3129,1.0246 ) circle ( .1 cm ) ; ( 3.6871,1.0246 ) circle ( .1 cm ) ; ( 3.0920,1.2180 ) circle ( .1 cm ) ; ( 2.5858,1.5858 ) circle ( .1 cm ) ; ( 2.2180,2.0920 ) circle ( .1 cm ) ; ( 2.0246,2.6871 ) circle ( .1 cm ) ; ( 2.0246,3.3129 ) circle ( .1 cm ) ; ( 2.2180,3.9080 ) circle ( .1 cm ) ; ( 2.5858,4.4142 ) circle ( .1 cm ) ; ( 3.0920,4.7820 ) circle ( .1 cm ) ; ( 3.6871,4.9754 ) circle ( .1 cm ) ; ( 4.3129,4.9754 ) circle ( .1 cm ) ; ( 4.9080,4.7820 ) circle ( .1 cm ) ; ( 5.4142,4.4142 ) circle ( .1 cm ) ; ( 5.7820,3.9080 ) circle ( .1 cm ) ; ( 5.9754,3.3129 ) circle ( .1 cm ) ; ( 1,3 ) ( 7,3 ) ; ( 7,3 ) ( 8,3 ) node[anchor = south ] @xmath339 ; ( 0,3 ) ( 1,3 ) ; at ( 6,4.5 ) ( a ) @xmath340 ; at ( 6,1.5 ) ( a ) @xmath341 ; at ( 8.5,2 ) ( a ) @xmath342 ; at ( 4.2,2.8 ) ( a ) @xmath343 ; ( 4,3 ) node[cross , rotate=0 ] ; ( -288,113)@xmath344 ( -270,98)@xmath345 ( -278,105)@xmath346 ( -275,101 ) the following lemma relates the quantity @xmath347 to the zolotarev number @xmath348 , [ \ell,1/\ell])$ ] with @xmath349 : let @xmath149 be an integer and @xmath350 . then , @xmath351 and @xmath352 , [ \ell,1/\ell]),\ ] ] where @xmath349 , the complex sets @xmath334 and @xmath335 are defined in , and @xmath56 is an even integer . [ lem : zolotarevrelation ] let @xmath353 be the extremal function for @xmath354,[\ell,1/\ell])$ ] characterized in theorem [ thm : zextremalrational ] , where @xmath349 . since the mbius transform given by @xmath355 maps @xmath334 to @xmath356 $ ] , @xmath335 to @xmath357 $ ] , and @xmath199 to @xmath358 , we have @xmath359 \cup [ \ell,1/\ell ] ) = \inf_{r\in\mathcal{r}_{2k,2k}}\frac{\sup_{z\in\mathbb{r } } \|r(iz)\|}{\inf_{z\in[-1/\ell,-\ell ] \cup [ \ell,1/\ell ] } \|r(z)\|}.\ ] ] now , consider the rational function @xmath360 where we used the fact that @xmath361 ( see theorem [ thm : zextremalrational ] , ( b ) ) . since @xmath183 for @xmath198 ( seetheorem [ thm : zextremalrational ] , ( c ) ) , we have @xmath362 moreover , since @xmath363 for @xmath364 $ ] ( see theorem [ thm : zextremalrational ] , ( a ) ) and @xmath365 is a nondecreasing function on @xmath366 $ ] and @xmath367 , we have @xmath368 \cup [ \ell,1/\ell ] } \left\|s(z)\right\| & = \sup_{z\in[-1/\ell,-\ell ] } \left\|\frac{2}{r(z)+1/r(z)}\right\| \\ & = \sup_{z\in[-1/\ell,-\ell ] } \left\|\frac{2r(z)}{1+r(z)^2}\right\|\\ & \leq \frac{2\sqrt{z_k}}{1+z_k}. \end{aligned}\ ] ] therefore , @xmath369 as required . the bound @xmath351 trivially holds from the definition of zolotarev numbers . by corollary [ cor : zbound ] we have the slightly weaker upper bound for @xmath370 : @xmath371,[\ell,1/\ell ] ) } \leq 4 \rho^{-k},\ ] ] where since @xmath372 for @xmath373 , we have @xmath374 if @xmath56 is an even integer , then we can immediately conclude a bound on the singular values from theorem [ thm : ratio ] . if @xmath56 is an odd integer , then one must employ first . the singular values of @xmath35 can be bounded as follows : @xmath375^{-k+[n]_2 } \sigma_{j}(k_{m , n } ) , \quad 1\leq j+2k\leq n , \label{eq : krylovbound3}\ ] ] where @xmath152 is the grtzsch function and @xmath376_2=1 $ ] if @xmath56 is odd and is @xmath377 if @xmath56 is even . the bound above remains valid , but is slightly weaken , if @xmath378 is replaced by @xmath379 . [ cor : krylovbounds ] figure [ fig : krylovmatrices ] demonstrates the bound on the singular values in on @xmath1 krylov matrices , where @xmath380 . the step behavior of the bound is due to the fact that only bounds @xmath381 when @xmath56 is even and we use the trivial inequality @xmath382 otherwise . one also observes that the singular values of krylov matrices with hermitian arguments can decay at a supergeometric rate ; however , the analysis in this paper only realizes a geometric decay . therefore , is only a reasonable bound on @xmath383 when @xmath60 is a small integer with respect to @xmath56 . if @xmath384 and @xmath385 , then improved bounds on @xmath383 may be possible by bounding discrete zolotarev numbers @xcite . the bound in provides an upper bound on the @xmath43-rank of @xmath35 : @xmath386 which allows for either an odd or even integer @xmath56 . krylovmatrices ( 47,0 ) index ( 54,25 ) ( 69,38 ) ( 69,48 ) hankelmatrices ( 47,0 ) index ( 54,25 ) ( 69,38 ) ( 69,50 ) recall that vandermonde matrices with real abscissas are also krylov matrices with hermitian arguments . therefore , the bounds in this section also apply to vandermonde matrices with real abscissas and shows that they have rapidly decaying singular values and are exponentially ill - conditioned . an observation that has been extensively investigated in the literature @xcite . an @xmath1 matrix @xmath2 is a hankel matrix if the matrix is constant along each anti - diagonal , i.e. , @xmath387 for @xmath58 . clearly , not all hankel matrices have decaying singular values , for example , the exchange matrix has repeated singular values of @xmath202 . this means that any displacement structure that is satisfied by all hankel matrices , for example , @xmath388 where @xmath331 is given in , does not result in a zolotarev number that decays . motivated by the hilbert matrix in section [ sec : cauchymatrices ] , we show that every real and positive definite hankel matrix has rapidly decaying singular values . previous work has led to bounds that can be calculated by using a pivoted cholesky algorithm @xcite , bounds for very special cases @xcite , as well as incomplete attempts @xcite . in order to exploit the positive definite structure we recall that the hamburger moment problem states that a real hankel matrix is positive semidefinite if and only if it is associated to a nonnegative borel measure supported on the real line . a real @xmath389 hankel matrix , @xmath2 , is positive semidefinite if and only if there exists a nonnegative borel measure @xmath390 supported on the real line such that @xmath391 for a proof , see ( * ? ? ? * theorem 7.1 ) . let @xmath2 be a real positive definite hankel matrix associated to the nonnegative weight @xmath390 in supported on @xmath199 . let @xmath392 and @xmath393 be the gauss quadrature nodes and weights associated to @xmath390 . then , since a gauss quadrature is exact for polynomials of degree @xmath394 or less , we have @xmath395 therefore , every real positive definite hankel matrices has a so - called _ fiedler factorization _ @xcite , i.e. , @xmath396 where @xmath397 is a krylov matrix with hermitian argument and @xmath398 is the conjugate transpose of @xmath397 . this means that @xmath399 for @xmath41 . that is , a bound on the singular values of @xmath2 and the @xmath43-rank of @xmath2 directly follows from . let @xmath2 be an @xmath1 real positive definite hankel matrix . then , @xmath400^{-2k+2 } \sigma_{j}(h_n ) , \qquad 1\leq j+2k\leq n , \label{eq : hankelbound}\ ] ] and @xmath401 where both bounds allow for @xmath56 to be an even or odd integer . we conclude that all real positive definite hankel matrices have an @xmath43-rank of at most @xmath402 , explaining why low rank techniques are usually advantageous in computational mathematics on such matrices . since a real positive semidefinite hankel matrix can be arbitrarily approximated by a real positive definite hankel matrix , the results from this section immediately extend to such hankel matrices . by replacing @xmath56 by the rank of @xmath2 . ] this fact was exploited , but not proved in general , in @xcite to derive quasi - optimal complexity fast transforms between orthogonal polynomial bases . many of the inequalities in this paper were numerically verified using rktoolbox @xcite and we thank stefan gttel for providing support . we thank yuji nakatsukasa for carefully reading a draft of this manuscript and pointing us towards @xcite and @xcite . we also thank sheehan olver , gil strang , marcus webb , and heather wilber for discussions . , _ elements of the theory of elliptic functions _ , transl . of math . monographs , 79 ams , providence ri , 1990 . , _ on the decay rate of hankel singular values and related issues _ , systems and control letters , 46 ( 2002 ) , pp . 323342 . , _ spectral sets _ , chapter 37 of l. hogben , handbook of linear algebra , second edition , chapman and hall / crc,2013 . , _ fast singular value decay for lyapunov solutions with nonnormal coefficients _ , siam j. mat . appl . , 36 ( 2015 ) , pp . 656668 . , _ on the numerical condition of polynomial bases : estimates for the condition number of vandermonde , krylov and hankel matrices _ , habilitationsschrift , universitt hannover , april 1996 . , _ the condition number of real vandermonde , krylov and positive definite hankel matrices _ , numer . math . , 85 ( 2000 ) , pp . 553577 . , _ singular value estimates for matrices with small displacement rank _ , presentation at structured matrices , cortona , sept . 2024 , 2004 . , _ an error analysis for rational galerkin projection applied to the sylvester equation _ , siam j. numer . anal . , 49 ( 2011 ) , pp . 24302450 . , _ extremal rational functions on symmetric discrete sets and superlinear convergence of the adi method _ , approx . , 32 ( 2010 ) , pp . 393428 . , _ error estimation and evaluation of matrix functions via the faber transform _ , siam j. num . , 47 ( 2009 ) , pp . 38493883 . , _ a rational krylov toolbox for matlab _ , mims eprint 2014.56 , the university of manchester , uk , 2014 . , _ on the decay of the off - diagonal singular values in cyclic reduction _ , arxiv preprint arxiv:1608.01567 , ( 2016 ) . , _ alternating direction implicit methods _ , advances in computers , 3 ( 1962 ) , pp . 189273 . , _ nonlinear approximation theory _ , springer - verlag , berlin - heidelberg , new york , 1986 . , _ rational approximation of stieltjes functions by the carathodory fejr method _ , approx . , 3 ( 1987 ) , pp . 4350 . , _ approximation of @xmath403 by exponential sums in @xmath404 _ , i m a j. numer . , 25 ( 2005 ) , pp . 685697 . , _ exact matrix completion via convex optimization _ , comput . math . , 9 ( 2009 ) , pp . 717772 . , _ how bad are symmetric pick matrices _ , in : structured matrices in mathematics , computer science , and engineering i , contemporary mathematics series , vol . 280 , pp . 301312 , 2001 . , _ special matrices and their applications in numerical mathematics _ , martinus nijhoff publishers , 1986 . , _ lower bounds for the condition number of vandermonde matrices _ , numer . math . , 52 ( 1988 ) , 241250 . , _ zolotarev problems connected with rational functions _ , sbornik : mathematics , 7 ( 1969 ) , pp . 623635 . , _ existence of a low rank or @xmath405 matrix approximant to the solution of a sylvester equation _ , numer . appl . , 11 ( 2004 ) , pp . 371389 . , _ a fast algorithm for particle simulations _ , j. comp . phys . , 73 ( 1987 ) , pp . 325348 . , _ zolotarev quadrature rules and load balancing for the feast eigensolver _ , siam j. sci . comput . , 37 ( 2015 ) , a2100a2122 . , _ a sparse matrix arithmetic based on @xmath405-matrices . part i : introduction to @xmath405-matrices _ , computing , 62 ( 1999 ) , pp . 89108 . , _ algebraic methods for toeplitz - like matrices and operators _ , vol . 13 of operator theory : advances and applications , birkhuser verlag , basel , 1984 . , _ on a zolotarev problem in the method of alternating directions _ , ussr comput . math . mathematical phys . , 17 ( 1977 ) , pp . 5876 . , _ variable time steps optimization of @xmath406-stable crank nicolson method _ , russian journal of numerical analysis and mathematical modelling , 20 ( 2005 ) , pp . 283303 . , _ fast algorithms for multivariable systems _ , phd thesis , stanford university , 1974 . , _ computing fundamental matrix decompositions accurately via the matrix sign function in two iterations : the power of zolotarev s functions _ , siam review , 58 ( 2016 ) , pp . 461493 . , _ nist handbook of mathematical functions _ , cambridge university press , 2010 . , _ lower bounds for separable approximations of the hilbert kernel _ , sbornik : mathematics , 198 ( 2007 ) , pp . 425432 . , _ how bad are vandermonde matrices ? _ , siam j. mat . anal . appl . , 37 ( 2016 ) , pp . 676694 . , _ hankel operators and their applications _ , springer , 2012 . , _ eigenvalue decay bounds for solutions of lyapunov equations : the symmetric case _ , systems control lett . , 40 ( 2000 ) , , _ solution of large - scale lyapunov equations via the block modified smith method _ , ph.d . thesis , rice university , houston , tx , 2006 . , _ computational methods for linear matrix equations _ , siam review , 58 ( 2016 ) , pp . 377441 . , _ principal submatrices ix : interlacing inequalities for singular values of submatrices _ , lin appl . , 5 ( 1972 ) , pp . 112 . , _ computing with functions in two dimensions _ , dphil thesis , university of oxford , 2014 . , _ fast polynomial transforms based on toeplitz and hankel matrices _ , arxiv preprint arxiv:1604.07486 , ( 2016 ) . , _ more on hankel matrices _ , manuscript , 1998 . , _ how bad are hankel matrices ? _ , numer . math . , 67 ( 1994 ) , pp . 261269 . , _ matrix computations _ , fourth edition , the johns hopkins university press , baltimore , 2013 . , _ finite sections of some classical inequalities _ , springer , heidelberg , 1970 . , _ application of elliptic functions to questions of functions deviating least and most from zero _ , zap . nauk . st . petersburg , 30 ( 1877 ) , pp . 159 . in section [ sec : zolotarev ] we noted that there were typos in an infinite product formula given by lebedev ( * ? ? ? * ( 1.11 ) ) . the mistake has unfortunately been copied several times in the literature . here , we attempt to correct these typos . lebedev @xcite and his successors @xcite were not concerned with the zolotarev problem in , but instead the equivalent problem of minimal blaschke products in the half plane , i.e. , @xmath407 ) = \min_{z_1, ... ,z_k\in \mathbb{c } } \max_{z\in [ a , b ] } \left\| \prod_{s=1}^k \frac{z - z_s}{z+\overline{z_s } } \right\| , \qquad 0<a < b<\infty . \label{eq : ek}\ ] ] in ( * ? ? ? * ( 1.11 ) ) , lebedev presented an infinite product formula for @xmath408 that unfortunately contained typos and resulted in an erroneous lower bound for @xmath408 in ( * ? ? ? * ( 1.12 ) ) . more recently , other erroneous lower bounds have been claimed in ( * ? ? ? * ( 4.1 ) ) for a related problem based on ( * ? ? ? * ( 3.17 ) ) . to correct the situation we first show that with @xmath150,[a , b])$ ] we have @xmath409)=e_k([a , b])=e_k([a / b,1]),\ ] ] where the last two equalities are immediate from symmetry considerations and scaling . since any @xmath410 describes a rational function for @xmath411)$ ] in , the solution to describes a rational function that is a candidate for the zolotarev problem in and we have @xmath412)$ ] . conversely , taking @xmath413 as in theorem [ thm : zextremalrational ] we get from property ( c ) that @xmath413 has a set of poles being closed under complex conjugation . property ( b ) tells us that , if @xmath414 is a pole of @xmath208 , then @xmath415 is a zero of @xmath208 . thus , from theorem [ thm : zextremalrational ] we have @xmath416 which implies that @xmath411)\leq \max_{z\in [ -b ,- a ] } \| r(z)\| \leq \sqrt{z_k}$ ] . here , in the last inequality we have applied property ( a ) . we conclude that @xmath411 ) = \sqrt{z_k}$ ] . therefore , an infinite product formula for @xmath417)$ ] that corrects ( * ? ? ? * ( 1.11 ) ) is obtained by taking square roots ( and setting @xmath418 ) in theorem [ thm : zproduct ] . that is , for @xmath419 we have @xmath420 ) = 2\rho^{-k}\prod_{\tau=1}^\infty \frac{(1+\rho^{-8\tau k})^2}{(1+\rho^{4k}\rho^{-8\tau k})^2 } , \qquad \rho=\exp\left(\frac{\pi^2}{2\mu(\eta)}\right),\ ] ] where @xmath152 is the grtzsch ring function . from , one obtains upper and lower bounds for @xmath417)$ ] that correct ( * ? ? ? * ( 1.12 ) ) , ( * ? ? ? * ( 3.17 ) ) , and ( * ? ? ? * ( 15 ) ) , namely @xmath421 ) \leq \frac{2\rho^{-k}}{1+\rho^{-4k } } \leq 2\rho^{-k } , \qquad \rho=\exp\left(\frac{\pi^2}{2\mu(\eta)}\right ) . \label{eq : ebounds}\ ] ] more refined estimates than in can be obtained by taking more terms from the infinite product in . more recently , the best rational approximation of the sign function on @xmath141\cup [ a , b]$ ] has become important in numerical linear algebra because of a recursive construction of spectral projectors of matrices @xcite . in this setting , if @xmath422,[a , b])$ ] then @xmath423\cup[a , b ] } \left\| r(z ) - { \rm sgn}(z ) \right\|,\quad { \rm sgn}(z ) = \begin{cases}1,&z\in [ a , b],\\ -1,&z\in [ -b ,- a].\end{cases}\ ] ] unfortunately , lower and upper bounds for @xmath424 are claimed in ( * ? ? ? * ( 4.1 ) ) based on the erroneous infinite product formula in ( * ? ? ? * ( 3.17 ) ) and for @xmath425 in ( * ? ? ? * ( 3.8 ) ) by incorrectly citing the fundamental work of gonar ( * ? ? ? * ( 32 ) ) . we believe it is therefore useful to state infinite product formulas for @xmath426 and the resulting estimates . we recall from the proof of theorem [ thm : zproduct ] and theorem [ thm : zextremalrational ] that we have @xmath427 thus , in the proof of theorem [ thm : zproduct ] we select @xmath428 and obtain @xmath429 again , this infinite product in results in asymptotically tight corrected lower and upper bounds on @xmath426 : @xmath430 similarly , more refined estimates than in can be obtained by taking more terms from the infinite product in .	matrices with displacement structure such as pick , vandermonde , and hankel matrices appear in a diverse range of applications . in this paper , we use an extremal problem involving rational functions to derive explicit bounds on the singular values of such matrices . for example , we show that the @xmath0th singular value of a real @xmath1 positive definite hankel matrix , @xmath2 , is bounded by @xmath3 with explicitly given constants @xmath4 and @xmath5 , where @xmath6 is the spectral norm . this means that a real @xmath1 positive definite hankel matrix can be approximated , up to an accuracy of @xmath7 with @xmath8 , by a rank @xmath9 matrix . analogous results are obtained for pick , cauchy , real vandermonde , lwner , and certain krylov matrices . singular values , displacement structure , zolotarev , rational 15a18 , 26c15
a set @xmath19 is _ symmetric _ if there exists a number @xmath20 ( the center of @xmath21 ) such that @xmath22 if and only if @xmath23 . given a set @xmath24 of positive measure , is there necessarily a symmetric subset @xmath25 of positive measure ? the main topic of this paper is to determine how large , in terms of the lebesgue measure of @xmath26 , one may take the symmetric set @xmath21 . in other words , for each @xmath27 we are interested in @xmath28 it is not immediately obvious , although it turns out to be true , that @xmath29 . we have dubbed this sort of question `` continuous ramsey theory '' , and we direct the reader to later in this section for problems with a similar flavor ; some of these have appeared in the literature and some are given here for the first time . we determine a lower bound for @xmath2 using tools and methods from harmonic analysis , some of which were spurred by ideas from nonstandard analysis and the theory of wavelets . we also construct sets without large symmetric subsets using results from probabilistic number theory . these two lines of attack complement each other , and our bounds on @xmath2 yield new results in additive number theory as well . the following theorem , proved in sections [ easy.section ] and [ deriving.upper.bounds.sec ] , states some fundamental properties of the function @xmath2 . [ delta.properties.thm ] the function @xmath2 is continuous and , in fact , satisfies the lipschitz condition @xmath30 for all @xmath31 $ ] . furthermore , the function @xmath32 is increasing on @xmath33 $ ] , and hence @xmath34 exists . we turn now to stating our quantitative bounds for @xmath2 . the lower bound @xmath35 , which we call the trivial lower bound on @xmath2 ( see lemma [ delta.trivial.lower.bound.lem ] below ) , is not so far from the best we can derive . in fact , the bulk of this paper is devoted to improving the constant in this lower bound from @xmath36 to @xmath37 . moreover , we are able to establish a complementary upper bound for @xmath2 using results on an analogous problem in combinatorial number theory . figure [ overallpic ] shows the precise upper and lower bounds we obtain for @xmath38 as functions of @xmath4 , which we present as theorem [ delta.summary.thm ] . [ delta.summary.thm ] we have : 1 . @xmath39 for @xmath6 , and @xmath40 for @xmath41 ; 2 . @xmath42 for all @xmath43 ; 3 . @xmath44 for all @xmath43 ; 4 . @xmath45 for @xmath8 ; 5 . @xmath46 for all @xmath43 . ( 384,240 ) ( 12,230)@xmath38 ( 370,12)@xmath4 ( 0,0)[overallpic],title="fig : " ] note that @xmath47 . the upper bound in part ( v ) of the theorem is superior to the one in part ( iv ) in the range @xmath48 . the five parts of the theorem are proved separately in proposition [ line.2e-1.prop ] , proposition [ the.full.de.bound.prop ] , proposition [ delta.one.half.thm ] , corollary [ part.iii.cor ] , and proposition [ part.iv.prop ] , respectively . perhaps surprisingly , the upper bounds given in theorem [ delta.summary.thm](iv)(v ) are derived from number - theoretic considerations . a set @xmath9 of integers is called a @xmath49 $ ] set if for any given @xmath11 there are at most @xmath12 ordered pairs @xmath13 with @xmath14 . we shall use constructions of large @xmath49 $ ] sets to derive upper bounds on @xmath2 in section [ deriving.upper.bounds.sec ] . conversely , our bounds on @xmath2 improve the best known upper bounds on the size of @xmath49 $ ] sets for large @xmath12 , as we show in @xcite . see the article @xcite of the second author for a survey of @xmath49 $ ] sets . we note that the difficulty of determining @xmath2 is in stark contrast to the analogous problem where we consider subsets of the circle @xmath50 instead of subsets of the interval @xmath3 $ ] . in this analogous setting , we completely determine the corresponding function @xmath51 ; in fact , we show ( corollary [ circle.answer.cor ] ) that @xmath52 for all @xmath43 . as it turns out , the methods that allow the proof of the upper bound @xmath53 , namely constructions of large @xmath49 $ ] sets in @xmath54 , are also helpful to us in constructing the large @xmath49 $ ] sets themselves . schinzel and schmidt @xcite consider the problem of bounding @xmath55 where the supremum is taken over all nonnegative functions supported on @xmath3 $ ] ; they showed that @xmath56 . the proof of theorem [ delta.summary.thm](ii ) improves the value @xmath57 to @xmath58 . we remark briefly on the phrase `` continuous ramsey theory '' . a `` coloring theorem '' has the form : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ given some fundamental set @xmath59 colored with a finite number of colors , there exists a highly structured monochromatic subset , provided that @xmath59 is sufficiently large . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ the prototypical example is ramsey s theorem itself : however one colors the edges of the complete graph @xmath60 with @xmath61 colors , there is a monochromatic complete subgraph on @xmath62 vertices , provided that @xmath63 is sufficiently large in terms of @xmath61 and @xmath62 . another example is van der waerden s theorem : however one colors the integers @xmath16 with @xmath61 colors , there is a monochromatic arithmetic progression with @xmath62 terms , provided that @xmath63 is sufficiently large in terms of @xmath61 and @xmath62 . in many cases , the coloring aspect of a ramsey - type theorem is a ruse and one may prove a stronger statement with the form : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ given some fundamental set @xmath59 , any large subset of @xmath59 contains a highly structured subset , provided that @xmath59 itself is sufficiently large . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ such a result is called a `` density theorem . '' for example , van der waerden s theorem is a special case of the density theorem of szemerdi : every subset of @xmath16 with cardinality at least @xmath64 contains a @xmath62-term arithmetic progression , provided that @xmath63 is sufficiently large in terms of @xmath65 and @xmath62 . ramsey theory is the study of such theorems on different types of structures . by `` continuous ramsey theory '' we refer to ramsey - type problems on continuous measure spaces . in particular , this paper is concerned with a density - ramsey problem on the structure @xmath66 with lebesgue measure . the type of substructure we focus on is a symmetric subset . other appearances of continuous ramsey theory in the literature are in the work of @xcite ( see also ( * ? ? ? * problem c17 ) ) , @xcite , @xcite , and @xcite . in all cases , there is an analogous discrete ramsey theory problem . however , see @xcite for an interesting example where the quantities involved in the discrete setting do not tend in the limit to the analogous quantity in the continuous setting . we give easy lower bounds for @xmath2 and prove that @xmath2 is continuous in section [ easy.section ] . section [ introduction.section ] below makes explicit the connection between @xmath2 and harmonic analysis . section [ basic.argument.section ] gives a simple , but quite good , lower bound on @xmath2 . in section [ main.bound.section ] , we give a more general form of the argument of section [ basic.argument.section ] . using an analytic inequality established in section [ useful.inequalities.section ] , we investigate in section [ full.bound.section ] the connection between @xmath67 and the fourier coefficients of @xmath68 , which , when combined with the results of section [ basic.argument.section ] , allows us to show that @xmath69 . the bound given in section [ basic.argument.section ] and improved in section [ full.bound.section ] depends on a kernel function with certain properties ; in section [ kernel.problem.section ] we discuss how we chose our kernel . in section [ delta.half.section ] , we use a different approach to derive a lower bound on @xmath2 which is superior for @xmath70 . we now turn our attention to the investigation of the function @xmath2 defined in eq . ( [ deltadef ] ) . in this section we establish several simple lemmas describing basic properties of @xmath71 . let @xmath72 denote lebesgue measure on @xmath73 . we find the following equivalent definition of @xmath2 easier to work with than the definition given in eq . : if we define @xmath74 then @xmath75 [ delta.trivial.bounds.lem ] @xmath76 for all @xmath77 . for @xmath24 , the centrally symmetric set @xmath78 has measure equal to @xmath79 therefore @xmath80 from the definition ( [ ddef ] ) of the function @xmath81 taking the infimum over all subsets @xmath26 of @xmath82 with measure @xmath4 , this becomes @xmath40 as claimed . while this bound may seem obvious , it is in many situations the state of the art . as we show in proposition [ line.2e-1.prop ] below , @xmath2 actually equals @xmath83 for @xmath84 ; and @xmath85 is the best lower bound of which we are aware in the range @xmath86 . one is tempted to try to sharpen the bound @xmath85 by considering the symmetric subsets with center @xmath87 , @xmath88 , or @xmath89 , for example , instead of merely @xmath88 . unfortunately , it can be shown that given any @xmath90 and any finite set @xmath91 , one can construct a sequence @xmath92 of sets , each with measure @xmath4 , that satisfies @xmath93 thus , no improvement is possible with this sort of argument . @xmath94 for all @xmath95 . [ delta.trivial.lower.bound.lem ] given a subset @xmath26 of @xmath82 of measure @xmath4 , let @xmath96 denote the indicator function of @xmath26 , so that the integral of @xmath96 over the interval @xmath82 equals @xmath4 . if we define @xmath97 , then @xmath98 is the measure of the largest symmetric subset of @xmath26 with center @xmath20 , and we seek to maximize @xmath98 . but @xmath68 is clearly supported on @xmath82 , and so @xmath99 since @xmath100 is supported on @xmath3 \subseteq[-x,2-x]$ ] . since @xmath26 was an arbitrary subset of @xmath82 of measure @xmath4 , we have shown that . it is obvious from the definition of @xmath71 that @xmath2 is an increasing function ; the next lemma shows that @xmath101 is also an increasing function . later in this paper ( see proposition [ de.over.e2.increasing.prop ] ) , we will show that in fact even @xmath32 is an increasing function . [ delta.line.to.origin.lem ] @xmath102 for all @xmath103 . in particular , @xmath104 . if @xmath105 is a scaled copy of a set @xmath26 , then clearly @xmath106 . applying this with any set @xmath24 of measure @xmath107 and with @xmath108 , we see that @xmath109 is a subset of @xmath82 with measure @xmath4 , and so by the definition of @xmath71 we have @xmath110 . taking the infimum over all sets @xmath24 of measure @xmath107 , we conclude that @xmath111 . the second assertion of the lemma follows from the first assertion with the trivial value @xmath112 . let @xmath113 denote the symmetric difference of @xmath9 and @xmath114 . ( while this operation is more commonly denoted with a triangle rather than with a @xmath115 , we would rather avoid any potential confusion with the function @xmath71 featured prominently in this paper . ) if @xmath9 and @xmath114 are two sets of real numbers , then @xmath116 . [ diamond.lem ] let @xmath117 be any symmetric subset of @xmath9 , and let @xmath20 be the center of @xmath117 , so that @xmath118 . define @xmath119 , which is a symmetric subset of @xmath114 with center @xmath20 . we can write @xmath120 using the inclusion - exclusion formula @xmath121 rearranging terms , and noting that @xmath122 , we see that @xmath123 because reflecting a set in the point @xmath20 does not change its measure , this is the same as @xmath124 therefore , since @xmath125 is a symmetric subset of @xmath114 , @xmath126 taking the supremum over all symmetric subsets @xmath117 of @xmath9 , we conclude that @xmath127 . if we now exchange the roles of @xmath9 and @xmath114 , we see that the proof is complete . the function @xmath71 satisfies the lipschitz condition @xmath128 for all @xmath107 and @xmath129 in @xmath3 $ ] . in particular , @xmath71 is continuous . [ lipschitz.condition.lem ] without loss of generality assume @xmath130 . in light of the monotonicity @xmath131 , it suffices to show that @xmath132 . let @xmath133 have measure @xmath129 . choose any set @xmath134 with measure @xmath135 , and set @xmath136 . then @xmath137 , and so by lemma [ diamond.lem ] , @xmath138 . therefore @xmath139 by the definition of @xmath71 . taking the infimum over all sets @xmath133 of measure @xmath129 yields @xmath132 as desired . there are many ways to define the basic objects of fourier analysis ; we follow @xcite . unless specifically noted otherwise , all integrals are over the circle group @xmath50 ; for example , @xmath140 denotes the class of functions @xmath68 for which @xmath141 is finite . for each integer @xmath142 , we define @xmath143 , so that for any function @xmath144 , we have @xmath145 almost everywhere . we define the convolution @xmath146 , and we note that @xmath147 for every integer @xmath142 ; in particular , @xmath148 . we define the usual @xmath149 norms @xmath150 and @xmath151 with these definitions , hlder s inequality is valid : if @xmath152 and @xmath153 are conjugate exponents that is , @xmath154then @xmath155 . we also note that @xmath156 when @xmath68 and @xmath12 are nonnegative functions ; in particular , @xmath157 . we shall also employ the @xmath158 norms for bi - infinite sequences : if @xmath159 , then latexmath:[$\\|a\\|_p : = \big ( \sum_{j\in{{\mathbb z } } } latexmath:[$\\|a\\|_\infty : = \lim_{p\to\infty } \\|a\\|_p = \sup_{j\in{{\mathbb z } } } @xmath158 norms , no confusion should arise , as the object inside the norm symbol will either be a function on @xmath161 or its sequence of fourier coefficients . with this notation , we recall parseval s identity @xmath162 ( assuming the integral and sum both converge ) ; in particular , if @xmath163 is real - valued ( so that @xmath164 is the conjugate of @xmath165 for all @xmath142 ) , this becomes @xmath166 . the hausdorff - young inequality , @xmath167 whenever @xmath152 and @xmath153 are conjugate exponents with @xmath168 , can be thought of as a generalization of this latter version of parseval s identity . we also require the definition @xmath169 for any sequence @xmath159 , so that @xmath170 , for example . we note that for any fixed sequence @xmath159 , the @xmath158-norm @xmath171 is a decreasing function of @xmath152 . to see this , suppose that @xmath172 and @xmath173 . then @xmath174 for all @xmath175 , whence @xmath176 ( since @xmath177 ) and so @xmath178 . summing both sides over all @xmath175 yields @xmath179 , and taking @xmath153th roots gives the desired inequality @xmath180 . finally , we define a `` pdf '' , short for `` probability density function '' , to be a nonnegative function in @xmath181 whose @xmath140-norm ( which is necessarily finite , since @xmath161 is a finite measure space ) equals 1 . also , we single out a special type of pdf called an `` nif '' , short for `` normalized indicator function '' , which is a pdf that only takes one nonzero value , that value necessarily being the reciprocal of the measure of the support of the function . ( we exclude the possibility that an nif takes the value 0 almost everywhere . ) specifically , we define for each @xmath182 the nif @xmath183 note that if @xmath68 is a pdf , then @xmath184 . we are now ready to reformulate the function @xmath2 in terms of this notation . [ fourier.connection.lem ] we have @xmath185 the first infimum being taken over all pdfs @xmath12 that are supported on @xmath186 $ ] , and the second infimum being taken over all nifs @xmath68 whose support is a subset of @xmath187 $ ] of measure @xmath188 . the inequality is trivial , since every nif is a pdf ; it remains to prove the equality . for each measurable @xmath24 , define @xmath189 $ ] . the sets @xmath26 and @xmath190 differ only by translation and scaling , so that @xmath191 and @xmath192 . thus @xmath193,\ , \lambda(e)=\tfrac{{\varepsilon}}{2 } \right\}. \end{aligned}\ ] ] for each @xmath194 $ ] with @xmath195 , the function @xmath196 is an nif supported on a subset of @xmath186 $ ] with measure @xmath197 , and it is clear that every such nif arises from some set @xmath117 . thus , it remains only to show that @xmath198 , i.e. , that @xmath199 . fix @xmath194 $ ] , and let @xmath200 be the indicator function of @xmath117 . note that @xmath201 . the maximal symmetric subset of @xmath117 with center @xmath20 is @xmath202 , and this has measure @xmath203 . thus @xmath204 as desired . the convolution in lemma [ fourier.connection.lem ] may be taken over @xmath73 or over @xmath161 , the two settings being equivalent since @xmath205 is supported on an interval of length 1 . in fact , the reason we scale @xmath68 to be supported on an interval of length @xmath88 is so that we may replace convolution over @xmath73 , which is the natural place to study @xmath2 , with convolution over @xmath161 , which is the natural place to do harmonic analysis . we begin the process of improving upon the trivial lower bound for @xmath2 by stating a simple version of our method that illustrates the ideas and techniques involved . [ cf.first.result.prop ] let @xmath206 be any continuous function on @xmath161 satisfying @xmath207 when @xmath208 $ ] , and let @xmath68 be a pdf supported on @xmath186 $ ] . then @xmath209 we have @xmath210 by parseval s identity . hlder s inequality now gives @xmath211 , which we restate as the inequality @xmath212 . now @xmath213 by another application of parseval s identity . since @xmath214 , integration yields @xmath215 . combining the last three sentences , we see that @xmath216 as claimed . this reasonably simple theorem already allows us to give a nontrivial lower bound for @xmath2 . the step function @xmath217 has @xmath218 ( the elaborate constant used in the definition of @xmath219 was chosen to minimize @xmath220 ) . a careful reader may complain that @xmath219 is not continuous . the continuity condition is not essential , however , as we may approximate @xmath219 by a continuous function @xmath221 with @xmath222 arbitrarily close to @xmath223 . green @xcite used a discretization of the kernel function @xmath224 with a suitably chosen @xmath225 to get @xmath226 . we get a slightly larger value of @xmath227 in the following corollary with a much more complicated kernel . see section [ kernel.problem.section ] for a discussion of how we came to find our kernel . [ first.cor ] if @xmath68 is a pdf supported on @xmath187 $ ] , then @xmath228 consequently , @xmath229 for all @xmath95 . set @xmath230 @xmath231 is pictured in figure [ goodk ] . ( 384,128 ) ( 176,114)@xmath231 ( 356,0)@xmath107 ( 0,-10)[goodk],title="fig : " ] we do not know how to rigorously bound @xmath232 , but we can rigorously bound @xmath233 where @xmath234 is a piecewise linear function ` close ' to @xmath235 . specifically , let @xmath236 be the even piecewise linear function with corners at @xmath237 we calculate ( using proposition [ piecewise.linear.prop ] below ) that @xmath238 . therefore , by proposition [ cf.first.result.prop ] we have @xmath239 using lemma [ fourier.connection.lem ] , we now have @xmath240 . the constants in the definition of @xmath231 were numerically optimized to minimize @xmath233 and otherwise have no special significance . the definition of @xmath231 is certainly not obvious , and there are much simpler kernels that do give nontrivial bounds . in section [ kernel.problem.section ] below , we indicate the experiments that led to our choice . we note that the function @xmath241 has @xmath242 and @xmath243 . although @xmath244 is not a pdf ( it is not in @xmath181 ) , it provides strong evidence that the bound on @xmath245 given in corollary [ first.cor ] is not far from best possible . this bound on @xmath245 may be nearly correct , but the resulting bound on @xmath67 is not : we prove below that @xmath246 , and believe that @xmath247 . we have tried to improve the argument given in proposition [ cf.first.result.prop ] in the following four ways : 1 . instead of considering the sum @xmath248 as a whole , we separate the central terms from the tails and establish inequalities that depend upon the two in distinct ways . this generalized form of the above argument is expounded in the next section . the success of this generalization relies on certain inequalities restricting the possible values of these central coefficients ; establishing these restrictions is the goal of sections [ useful.inequalities.section ] and [ full.bound.section ] . the final lower bound derived from these methods is given in section [ full.bound.section ] . we have searched for more advantageous kernel functions @xmath249 for which we can compute @xmath250 in an accurate way . a detailed discussion of our search for the best kernel functions is in section [ kernel.problem.section ] . the application of parseval s identity can be replaced with the hausdorff - young inequality , which leads to the conclusion @xmath251 , where @xmath252 and @xmath253 are conjugate exponents . numerically , the values @xmath254 appear to be optimal . however , beckner s sharpening @xcite of the hausdorff - young inequality leads to the stronger conclusion @xmath255 where @xmath256 . we have not experimented to see whether a larger lower bound can be obtained from this stronger inequality by taking @xmath257 . notice that we used the inequality @xmath258 with the function @xmath259 . this inequality is sharp exactly when the function @xmath12 takes only one nonzero value ( i.e. , when @xmath12 is a nif ) , but the convolution @xmath205 never behaves that way . perhaps for these autoconvolutions , an analogous inequality with a stronger constant than 1 could be established . unfortunately , we have not been able to realize any success with this idea , although we believe conjecture [ infinity.two.cnj ] below . if true , the conjecture implies the bound @xmath260 . [ infinity.two.cnj ] if @xmath68 is a pdf supported on @xmath187 $ ] , then @xmath261 with equality only if either @xmath262 or @xmath263 equals @xmath264 on the interval @xmath265 . we remark that proposition [ cf.first.result.prop ] can be extended from a twofold convolution in one dimension to an @xmath266-fold convolution in @xmath267 dimensions . let @xmath206 be any continuous function on @xmath268 satisfying @xmath269 when @xmath270^d$ ] , and let @xmath68 be a pdf supported on @xmath271^d$ ] . then @xmath272 every subset of @xmath3^d$ ] with measure @xmath4 contains a symmetric subset with measure @xmath273 . the proof proceeds as above , with the conjugate exponents @xmath274 in place of @xmath275 , and the kernel function @xmath276 in place of the kernel function @xmath249 defined in the proof of corollary [ first.cor ] . the second assertion of the proposition follows on taking @xmath277 . we now present a more subtle version of proposition [ cf.first.result.prop ] . recall that the notation @xmath278 was defined in eq . ( [ lnorm.def ] ) . we also use @xmath279 to denote the real part of the complex number @xmath280 . [ main.bound.prop ] let @xmath281 . suppose that @xmath68 is a pdf supported on @xmath186 $ ] and that @xmath206 is even , continuous , satisfies @xmath282 for @xmath283 , and @xmath284 . set @xmath285 . then @xmath286 the equality follows from parseval s formula @xmath287 as in the proof of proposition [ cf.first.result.prop ] , we have @xmath288 since @xmath206 is even , @xmath289 is real , and since @xmath68 is real valued , @xmath290 . we have @xmath291 which we can also write as @xmath292 . taking absolute values and applying hlder s inequality , we have @xmath293 which we recast in the form @xmath294 we add @xmath295 to both sides and observe that @xmath296 and @xmath297 to finish the proof of the inequality . with @xmath298 , the bound of proposition [ main.bound.prop ] becomes @xmath299 this is a quartic polynomial in the @xmath300 , and consequently it is not difficult to minimize , giving an absolute lower bound on @xmath301 . this minimum occurs at @xmath302 where @xmath303 is the real cube root of @xmath304 . a substitution and simplification of the resulting expression then yields @xmath305 which is nothing more than the bound that proposition [ main.bound.prop ] gives with @xmath306 . moreover , @xmath307 ( the details of this calculation are given in section [ kernel.problem.section ] ) so that proposition [ main.bound.prop ] , by itself , does not give a different bound on @xmath301 than proposition [ cf.first.result.prop ] . however , we shall obtain additional information on @xmath308 in terms of @xmath67 in section [ full.bound.section ] below , and this information can be combined with proposition [ main.bound.prop ] to provide a stronger lower bound on @xmath67 than that given by proposition [ cf.first.result.prop ] . [ cf.main.bound.cor ] let @xmath68 be a pdf supported on @xmath187 $ ] , and set @xmath309 . then @xmath310 set @xmath311 denote by @xmath312 the even piecewise linear function with corners at @xmath313 we find ( using proposition [ piecewise.linear.prop ] ) that @xmath314 , @xmath315 , and @xmath316 . apply proposition [ main.bound.prop ] with @xmath317 to finish the proof . hardy , littlewood , and plya @xcite call a function @xmath318 _ symmetric decreasing _ if @xmath319 and @xmath320 for all @xmath321 , and they call @xmath322 the _ symmetric decreasing rearrangement _ of @xmath68 . for example , if @xmath68 is the indicator function of a set with measure @xmath323 , then @xmath324 is simply the indicator function of the interval @xmath325 . another example is any function @xmath68 defined on an interval @xmath326 $ ] and is periodic with period @xmath327 , where @xmath63 is a positive integer , and that is symmetric decreasing on the subinterval @xmath328 $ ] ; then @xmath329 for all @xmath330 $ ] . in particular , on the interval @xmath187 $ ] , we have @xmath331 for any nonzero integer @xmath142 . we shall need the following result ( * ? ? ? * theorem 378 ) : @xmath332 we say that _ @xmath333 is more focused than @xmath68 _ ( and @xmath68 is less focused than @xmath333 ) if for all @xmath334 $ ] and all @xmath335 we have @xmath336 for example , @xmath324 is more focused than @xmath68 . in fact , we introduce this terminology because it refines the notion of symmetric decreasing rearrangement in a way that is useful for us . to give another example , if @xmath68 is a nonnegative function , set @xmath337 to be @xmath338 times the indicator function of the interval @xmath339 $ ] ; then @xmath337 is more focused than @xmath68 . [ focus.lem ] let @xmath318 be a symmetric decreasing function , and let @xmath340 be pdfs with @xmath341 more focused than @xmath266 . then for all @xmath335 , @xmath342 without loss of generality we may assume that @xmath343 , since if @xmath344 is more focused than @xmath345 , then it is also more focused than @xmath346 . also , without loss of generality we may assume that @xmath347 are continuous and strictly positive on @xmath161 , since any nonnegative function in @xmath140 can be @xmath140-approximated by such . define @xmath348 and @xmath349 , so that @xmath350 , and note that the more - focused hypothesis implies that @xmath351 for all @xmath334 $ ] . now @xmath266 is continuous and strictly positive , which implies that @xmath352 is differentiable and strictly increasing on @xmath353 $ ] since @xmath354 . therefore @xmath355 exists as a function from @xmath3 $ ] to @xmath353 $ ] . similar comments hold for @xmath356 . since @xmath357 , we see that @xmath358 for all @xmath359 $ ] . then , since @xmath360 and @xmath360 are positive and @xmath361 is decreasing for positive arguments , we conclude that @xmath362 , and so @xmath363 on the other hand , making the change of variables @xmath364 , we see that @xmath365 since @xmath361 is symmetric . similarly @xmath366 , and so inequality ( [ h.inverse.inequality ] ) becomes @xmath367 as desired . to use proposition [ main.bound.prop ] to bound @xmath2 , we need to develop a better understanding of the central fourier coefficients @xmath308 for small @xmath142 . in particular , we wish to apply proposition [ main.bound.prop ] with @xmath317 , i.e. , we need to develop the connections between @xmath67 and the real part of the fourier coefficient @xmath368 . we turn now to bounding @xmath369 in terms of @xmath67 . the guiding principle is that if @xmath205 is very concentrated then @xmath67 will be large , and if @xmath205 is not very concentrated then @xmath369 will be small . green ( * ? ? ? * lemma 26 ) proves the following lemma in a discrete setting , but since we need a continuous version we include a complete proof . [ fhat.green.bound.lem ] let @xmath68 be a pdf supported on @xmath187 $ ] . for @xmath370 , @xmath371 let @xmath372 be defined by @xmath373 , with @xmath374 chosen so that @xmath375 is real and positive ( clearly @xmath376 and @xmath377 ) . set @xmath345 to be the symmetric decreasing rearrangement of @xmath378 , and @xmath379 , where @xmath380 is the indicator function of @xmath381 $ ] . we have @xmath382 by the inequality . we now apply lemma [ focus.lem ] to find @xmath383 with this technical result in hand , we can finally establish the lower bound on @xmath2 given in theorem [ delta.summary.thm](ii ) . [ the.full.de.bound.prop ] @xmath42 for all @xmath95 . this gist of the proof of proposition [ the.full.de.bound.prop ] is that if @xmath67 is small , then @xmath384 is small by lemma [ fhat.green.bound.lem ] , and so @xmath245 is not very small by corollary [ cf.main.bound.cor ] , whence @xmath67 is not small . if @xmath385 , then we get a contradiction . let @xmath68 be a pdf supported on @xmath187 $ ] , and assume that @xmath386 set @xmath309 . since @xmath68 is supported on @xmath187 $ ] , we see that @xmath387 . by lemma [ fhat.green.bound.lem ] , @xmath388 however , we already know from corollary [ cf.main.bound.cor ] that @xmath389 routine calculus shows that there are no simultaneous solutions to the inequalities ( [ main.bound.0.eq ] ) , ( [ main.bound.1.eq ] ) , and ( [ main.bound.3.eq ] ) . therefore @xmath390 , whence lemma [ fourier.connection.lem ] implies that @xmath42 . let @xmath391 be the class of functions @xmath392 satisfying @xmath393 on @xmath187 $ ] . proposition [ cf.first.result.prop ] suggests the problem of computing @xmath394 in proposition [ cf.first.result.prop ] the case @xmath395 arose , but using the hausdorff - young inequality in place of parseval s identity we are led to consider @xmath396 . also , we assumed in proposition [ cf.first.result.prop ] that @xmath206 was continuous , but this assumption can be removed by taking the pointwise limit of continuous functions . as similar problems occur in @xcite and in @xcite , we feel it is worthwhile to detail the thoughts and experiments that led to the kernel functions chosen in corollaries [ first.cor ] and [ cf.main.bound.cor ] . our first observation is that if @xmath397 , then so is @xmath398 , and since latexmath:[$\|\hat{k}(j)\|=\|\re\hat{g}(j)\|\leq thus , we may restrict our attention to the _ even _ functions in @xmath391 . we also observe that @xmath400 decays more rapidly if many derivatives of @xmath206 are continuous . this suggests that we should restrict our attention to continuous @xmath206 , perhaps even to infinitely differentiable @xmath206 . however , computations suggest that the best functions @xmath206 are continuous but _ not _ differentiable at @xmath401 ( see in particular figure [ optimal.k.p43.fig ] below ) . in the argument of proposition [ cf.first.result.prop ] we used the inequality @xmath402 , which is an equality if we take @xmath206 to be equal to 1 on @xmath187 $ ] , instead of merely at least 1 . in light of this , we should not be surprised if the optimal functions in @xmath391 are exactly 1 on @xmath187 $ ] . this is supported by our computations . finally , we note that if @xmath403 , and @xmath404 with @xmath405 , then @xmath406 also . this is particularly useful with @xmath407 . specifically , given any @xmath408 with known @xmath409 ( we stipulate @xmath410 to avoid technicalities ) , we may easily compute the @xmath411 $ ] for which @xmath412 is minimized , where @xmath413 . we have @xmath414 where we have set @xmath415 and @xmath416 . taking the derivative with respect to @xmath225 , we obtain @xmath417 the only root of which is @xmath418 ( where @xmath154 ) . it is straightforward ( albeit tedious ) to check by substituting @xmath225 into the second derivative of the expression ( [ hatkpp.expression ] ) that this value of @xmath225 yields a local maximum for @xmath419 . the maximum value attained is then calculated to equal @xmath420 , which is easily computed from the known function @xmath421 . notice that when @xmath395 ( so @xmath422 ) , applying proposition [ cf.first.result.prop ] with our optimal function @xmath206 yields @xmath423 whereupon we recover the conclusion of proposition [ main.bound.prop ] with @xmath306 . haar wavelets provide a convenient basis for @xmath424)$ ] . we have numerically optimized the coefficients in various spaces of potential kernel functions @xmath206 spanned by short sums of haar wavelets to minimize @xmath425 within those spaces . the resulting functions are shown in figure [ optimal.k.p43.fig ] . this picture justifies restricting our attention to continuous functions that are constant on @xmath187 $ ] , and also implies that the optimal kernels are non - differentiable at @xmath426 , indeed that their derivatives become unbounded near these points . ( 384,360 ) ( 24,108)@xmath249 ( 24,228)@xmath249 ( 24,348)@xmath249 ( 215,108)@xmath249 ( 214,228)@xmath249 ( 213,348)@xmath249 ( 178,64)@xmath107 ( 178,184)@xmath107 ( 178,304)@xmath107 ( 372,64)@xmath107 ( 372,184)@xmath107 ( 372,304)@xmath107 ( 85,88)@xmath427 ( 86,208)@xmath428 ( 86,328)@xmath429 ( 275,88)@xmath430 ( 275,208)@xmath431 ( 276,328)@xmath432 , title="fig : " ] ( 384,128 ) ( 25,122)@xmath249 ( 186,102)@xmath433 ( 378,68)@xmath107 ( 0,0),title="fig : " ] for computational reasons , we further restrict attention to the class of continuous piecewise - linear even functions whose vertices all have abscissae with a given denominator . let @xmath434 denote the hurwitz zeta function . if @xmath435 is a vector , define @xmath436 to be the vector whose coordinates are the @xmath152th powers of the absolute values of the corresponding coordinates of @xmath435 . let @xmath114 be a positive integer , @xmath63 a nonnegative integer , and @xmath437 a real number . for each integer @xmath438 , define @xmath439 , and let @xmath440 be an arbitrary real number , except that @xmath441 . let @xmath249 be the even function on @xmath161 that is linear on @xmath442 $ ] and on each of the intervals @xmath443 $ ] @xmath444 , satisfying @xmath445 and @xmath446 @xmath447 . then @xmath448 where @xmath449 is the @xmath114-dimensional vector @xmath450 , @xmath451 is the @xmath452 matrix whose @xmath453-th component is @xmath454 and @xmath455 is the @xmath456-dimensional vector @xmath457 [ piecewise.linear.prop ] note that @xmath458 where @xmath459 and @xmath460 are the slope and @xmath129-intercept of the line going through @xmath461 and @xmath462 . if we define @xmath463 , then integrating by parts we have @xmath464 the first term of this expression is @xmath465 since @xmath466 and @xmath467 , @xmath468 , this entire expression is a telescoping sum whose value is zero . ( [ first.term.o1 ] ) thus becomes @xmath469 using @xmath470 . each @xmath471 is rational and can be written with denominator @xmath456 , so we see that the sequence of normalized fourier coefficients @xmath472 is periodic with period @xmath456 . we proceed to compute @xmath473 with @xmath63 positive and @xmath474 . @xmath475 because of the periodicity of @xmath472 , we may write this as @xmath476 which concludes the proof . proposition [ piecewise.linear.prop ] is useful in two ways . the first is that only @xmath449 depends on the chosen values @xmath440 . that is , the vector @xmath455 and the matrix @xmath451 may be precomputed ( assuming @xmath114 is reasonably small ) , enabling us to compute @xmath473 quickly enough as a function of @xmath449 to numerically optimize the @xmath440 . the second use is through eq . ( [ piecewise.linear.eq ] ) . for a given @xmath206 , we set @xmath477 , whereupon @xmath472 is computed for each @xmath142 using the formula in eq . . thus we can use eq . ( [ piecewise.linear.eq ] ) to compute @xmath478 with arbitrary accuracy , where @xmath219 is almost equal to @xmath206 . we have found that with @xmath479 one can generally compute @xmath478 quickly . in performing these numerical optimizations , we have found that `` good '' kernels @xmath480 have a very negative slope at @xmath481 . see figure [ optimal.k.p43.fig ] , for example , where the `` @xmath482 '' picture denotes the step function @xmath206 for which @xmath483 is minimal among all step functions whose discontinuities all lie within the set @xmath484 . viewing graphs of these numerically optimized kernels suggests that functions of the form @xmath485 which have slope @xmath486 at @xmath487 , may be very good . ( note that the graph of @xmath488 between @xmath489 and @xmath490 is the lower half of an ellipse . ) more good candidates are functions of the form @xmath491 where @xmath492 and @xmath493 are positive . we have used a function of the form @xmath494 in the proof of corollary [ cf.main.bound.cor ] and a function of the form @xmath495 in the proof of corollary [ first.cor ] . we begin with a fundamental relationship between @xmath497 and @xmath498 . [ fhat.f1.f2.domain.lem ] let @xmath68 be a pdf supported on @xmath186 $ ] . then @xmath499 to prove the first inequality , set @xmath500 ( with @xmath501 ) and observe that for @xmath502 , we have @xmath503 . thus @xmath504 rearranging , we arrive at @xmath505 . setting @xmath506 , we find that @xmath507 . as for the second inequality , since @xmath508 is at least 1 for @xmath502 , we have @xmath509 rearranging , we arrive at @xmath510 . from the inequality @xmath511 ( lemma [ fhat.f1.f2.domain.lem ] ) one easily computes that @xmath512 this yields @xmath513 , a non - trivial bound . if one assumes that @xmath68 is an nif supported on a subset of @xmath187 $ ] with large measure , then one can do much better than lemma [ fhat.f1.f2.domain.lem ] . the following proposition establishes the lower bound on @xmath2 given in theorem [ delta.summary.thm](iii ) . [ delta.one.half.thm ] let @xmath68 be an nif supported on a subset of @xmath187 $ ] with measure @xmath514 . then @xmath515 and consequently @xmath516 in particular , @xmath517 . for @xmath518 , this proposition is weaker than lemma [ delta.trivial.bounds.lem ] , and for @xmath519 it is weaker than proposition [ the.full.de.bound.prop ] , so we restrict our attention to @xmath70 . let @xmath520 be a parameter and set @xmath521 . if we define @xmath522 , then @xmath523 on the one hand , and @xmath524 on the other , where @xmath525 is the symmetric decreasing rearrangement of @xmath526 on the interval @xmath187 $ ] . thus @xmath527 the right - hand side may be computed explicitly as a function of @xmath4 and @xmath244 and then the value of @xmath244 chosen in terms of @xmath4 to maximize the resulting expression . one finds that for @xmath528 , the optimal choice of @xmath244 lies in the interval @xmath529 , and the resulting lower bound for @xmath125 is @xmath530 from lemma [ fhat.green.bound.lem ] we know that @xmath531 . we compare these bounds on @xmath125 to conclude the proof . specifically , @xmath532 where the expression on the right - hand side of this equation is from the taylor expansion of @xmath533 at @xmath534 , and @xmath535 where the expression on the right - hand side is from the taylor expansion of the middle expression at @xmath536 . comparing eqs . ( [ f2.lower.bound ] ) and ( [ f2.upper.bound ] ) gives a lower bound on @xmath67 , say @xmath537 with certain constants @xmath538 . it is easily checked that @xmath539 and @xmath540 , concluding the proof of the first asserted inequality . the second inequality then follows from lemma [ fourier.connection.lem ] . a symmetric set consists of pairs @xmath542 all with a fixed midpoint @xmath543 . if there are few pairs in @xmath544 with a given sum @xmath545 , then there will be no large symmetric subset of @xmath117 with center @xmath20 . we take advantage of the constructions of large integer sets whose pairwise sums repeat at most @xmath12 times to construct large real subsets of @xmath82 with no large symmetric subsets . more precisely , a set @xmath9 of integers is called a @xmath10 $ } } $ ] set if for any given @xmath11 there are at most @xmath12 ordered pairs @xmath13 with @xmath14 . ( in the case @xmath15 , these are better known as sidon sets . ) define @xmath546 $ } } $ set } \big\}. \label{rdef}\ ] ] for any integers @xmath547 , we have @xmath548 [ delta.from.r.connection.prop ] let @xmath549 be a @xmath49 $ ] set with @xmath550 . define @xmath551 and note that @xmath552 $ ] and that the measure of @xmath553 is exactly @xmath554 . thus it suffices to show that the largest symmetric subset of @xmath553 has measure at most @xmath555 . notice that the set @xmath553 is a finite union of intervals , and so the function @xmath556 , which gives the measure of the largest symmetric subset of @xmath553 with center @xmath20 , is piecewise linear . ( figure [ as.and.convolution ] contains a typical example of the set @xmath553 portrayed in dark gray below the @xmath20-axis , together with the function @xmath556 shown as the upper boundary of the light gray region above the @xmath20-axis , for @xmath557 . ) without loss of generality , therefore , we may restrict our attention to those symmetric subsets of @xmath553 whose center @xmath20 is the midpoint of endpoints of any two intervals @xmath558 . in other words , we may assume that @xmath559 . ( 384,178 ) ( 0,163)@xmath560 ( 370,20)@xmath20 ( 30,60)@xmath561 ( 30,99.5)@xmath562 ( 30,139.5)@xmath563 ( 10,0 ) , and the function @xmath560 , with @xmath557[as.and.convolution],title="fig : " ] suppose @xmath361 and @xmath564 are elements of @xmath553 such that @xmath565 . write @xmath566 and @xmath567 for integers @xmath568 and real numbers @xmath569 satisfying @xmath570 . ( we may ignore the possibility that @xmath571 or @xmath572 is an integer , since this is a measure - zero event for any fixed @xmath20 . ) then @xmath573 , and since @xmath574 , @xmath575 , and @xmath576 are all integers , we see that @xmath577 is also an integer . but @xmath578 , so @xmath579 and @xmath580 . since @xmath9 is a @xmath10 $ } } $ ] set , there are at most @xmath12 solutions @xmath581 to the equation @xmath580 . if it happens that @xmath582 , the interval @xmath583 ( a set of measure @xmath584 ) is contributed to the symmetric subset with center @xmath20 . otherwise , the set @xmath585 ( a set of measure @xmath586 ) is contributed to the symmetric subset with center @xmath20 , but this counts for the two solutions @xmath581 and @xmath587 . in total , then , the largest symmetric subset having center @xmath20 has measure at most @xmath555 . this establishes the proposition . using proposition [ delta.from.r.connection.prop ] , we can translate lower bounds on @xmath2 into upper bounds on @xmath541 , as in corollary [ delta.r.basic.cor ] . [ delta.r.basic.cor ] if @xmath588 , then @xmath589 for all @xmath590 . we remark that we may take @xmath591 by theorem [ delta.summary.thm](ii ) , and so this corollary implies that @xmath592 . this improves the previously - best bound on @xmath541 ( given in @xcite ) for @xmath18 and @xmath63 large . combining the hypothesized lower bound @xmath593 with proposition [ delta.from.r.connection.prop ] , we find that @xmath594 which is equivalent to @xmath595 . we have been unable to prove or disprove that @xmath596 i.e. , that corollary [ delta.r.basic.cor ] is best possible as @xmath597 . at any rate , for small @xmath12 it is possible to do better by taking advantage of the shape of the set @xmath553 used in the proof of proposition [ delta.from.r.connection.prop ] this is the subject of the companion paper @xcite of the authors . proposition [ delta.from.r.connection.prop ] provides a one - sided inequality linking @xmath598 and @xmath541 . it will also be useful for us to prove a theoretical result showing that the problems of determining the asymptotics of the two functions are , in a weak sense , equivalent . in particular , the following proposition implies that the trivial lower bound @xmath599 and the trivial upper bound @xmath600 are actually equivalent . further , any nontrivial lower bound on @xmath2 gives a nontrivial upper bound on @xmath541 , and vice versa . @xmath601 for all @xmath95 . [ delta.as.infimum.prop ] that @xmath598 is bounded above by the right - hand side follows immediately from proposition [ delta.from.r.connection.prop ] and the fact that @xmath71 is an increasing function . for the complementary inequality , let @xmath133 with @xmath602 . basic lebesgue measure theory tells us that given any @xmath603 , there exists a finite union @xmath114 of open intervals such that @xmath604 , and it is easily seen that @xmath114 can be chosen to meet the following criteria : @xmath605 , the endpoints of the finitely many intervals comprising @xmath114 are rational , and @xmath606 . choosing a common denominator @xmath63 for the endpoints of the intervals comprising @xmath114 , we may write @xmath607 ( up to a finite set of points ) for some set of integers @xmath608 ; most likely we have greatly increased the number of intervals comprising @xmath114 by writing it in this manner , and @xmath609 contains many consecutive integers . let @xmath12 be the maximal number of solutions @xmath610 to @xmath611 as @xmath612 varies over all integers , so that @xmath609 is a @xmath49 $ ] set and thus @xmath613 by the definition of @xmath59 . it follows that @xmath614 . now @xmath114 is exactly the set @xmath615 as defined in eq . ; hence @xmath616 as we saw in the proof of proposition [ delta.from.r.connection.prop ] . therefore by lemma [ diamond.lem ] , @xmath617 taking the infimum over appropriate sets @xmath9 and noting that @xmath603 was arbitrary , we derive the desired inequality @xmath618 . we begin by considering a modular version of sets . a set @xmath9 is a set if for any given @xmath11 there are at most @xmath12 ordered pairs @xmath13 with @xmath619 ( equivalently , if the coefficients of the least - degree representative of @xmath620 are bounded by @xmath12 ) . for example , the set @xmath621 is a @xmath622\pmod{7}$ ] set , and @xmath623 is a @xmath624\pmod{12}$ ] set . note that @xmath625 , so that @xmath623 is not a `` modular sidon set '' as defined by some authors , e.g. , @xcite or ( * ? ? ? * problem c10 ) . just as we defined @xmath541 to be the largest possible cardinality of a set contained in @xmath626 , we define @xmath627 to be the largest possible cardinality of a set . the mnemonic is `` r '' for the @xmath73eal problem and `` c '' for the circular problem . we demonstrate the existence of large sets via a probabilistic construction in this section , and we give a similar probabilistic construction of large sets in section [ probabilistic.bg.section ] . we rely upon the following two lemmas , which are quantitative statements of the central limit theorem . let @xmath628 be real numbers in the range @xmath3 $ ] , and set @xmath629 . define mutually independent random variables @xmath630 such that @xmath631 takes the value @xmath632 with probability @xmath633 and the value @xmath634 with probability @xmath632 ( so that the expectation of each @xmath631 is zero ) , and define @xmath635 . then for any positive number @xmath636 , @xmath637 < \exp \big ( \frac{-a^2}{2pn } + \frac{a^3}{2p^2n^2 } \big ) \quad\text{and}\quad { \mathop{\rm pr } } [ x<-a ] < \exp \big ( \frac{-a^2}{2pn } \big).\ ] ] [ prob.method.lem ] these assertions are theorems a.11 and a.13 of @xcite . let @xmath628 be real numbers in the range @xmath3 $ ] , and set @xmath638 . define mutually independent random variables @xmath639 such that @xmath640 takes the value 1 with probability @xmath633 and the value 0 with probability @xmath632 , and define @xmath641 ( so that the expectation of @xmath642 equals @xmath117 ) . then @xmath643 < \exp \big ( \frac{-a^2}{3e } \big)$ ] for any real number @xmath644 , and @xmath645 < \exp \big ( \frac{-a^2}{2e } \big)$ ] for any positive real number @xmath636 . [ our.application.lem ] this follows immediately from lemma [ prob.method.lem ] upon defining @xmath646 for each @xmath647 and noting that @xmath648 and that @xmath649 under the assumption @xmath644 . we now give the probabilistic construction of large sets . we write that @xmath650 ( and @xmath651 ) if @xmath652 is at least 1 . for every @xmath43 , there is a sequence of ordered pairs @xmath653 of positive integers such that @xmath654 and @xmath655 . [ circle.probabilistic.prop ] let @xmath63 be an odd integer . we define a random subset @xmath9 of @xmath656 as follows : for every @xmath657 , let @xmath640 be 1 with probability @xmath4 and 0 with probability @xmath658 with the @xmath640 mutually independent , and let @xmath659 . we see that @xmath660 has expectation @xmath661 . setting @xmath662 , lemma [ our.application.lem ] gives @xmath663 < \frac12.\ ] ] now for any integer @xmath612 , define the random variable @xmath664 so that @xmath665 is the number of representations of @xmath666 as the sum of two elements of @xmath9 . observe that @xmath665 is the sum of @xmath667 random variables taking the value 1 with probability @xmath668 and the value 0 otherwise , plus one random variable ( corresponding to @xmath669 ) taking the value 1 with probability @xmath4 and the value 0 otherwise . therefore the expectation of @xmath665 is @xmath670 . setting @xmath671 , and noting that @xmath672 when @xmath63 is sufficiently large in terms of @xmath4 , lemma [ our.application.lem ] gives @xmath673 < \frac1{2n}\ ] ] for each @xmath674 . the random set @xmath9 is a @xmath10\pmod{n}$ } } $ ] set with @xmath675 and @xmath676 unless @xmath677 or @xmath678 or @xmath679 or @xmath680 or @xmath681 . for any events @xmath682 , @xmath683 < \sum_{i } { \mathop{\rm pr}}[a_i],\ ] ] and consequently , @xmath684 \\ < \frac 12 + 2n\,\cdot \frac{1}{2n } < 1 . \end{gathered}\ ] ] therefore , there exists a @xmath10\pmod{n}$ } } $ ] set @xmath685 , with @xmath686 , with @xmath687 . this establishes the proposition . define @xmath688 to be the supremum of those real numbers @xmath65 such that every subset of @xmath161 with measure @xmath4 has a subset with measure @xmath65 that is fixed by a reflection @xmath689 . the function @xmath688 stands in relation to @xmath627 as @xmath2 stands to @xmath541 . however , it turns out that @xmath690 is much easier to understand . every subset of @xmath161 with measure @xmath4 contains a symmetric subset with measure @xmath668 , and this is best possible for every @xmath4 : @xmath691 for all @xmath95 . [ circle.answer.cor ] in the proof of the trivial lower bound for @xmath2 ( lemma [ delta.trivial.lower.bound.lem ] ) , we saw that every subset of @xmath3 $ ] with measure @xmath4 contains a symmetric subset with measure at least @xmath692 . the proof is easily modified to show that every subset of @xmath161 with measure @xmath4 contains a symmetric subset with measure @xmath668 . this shows that @xmath693 for all @xmath4 . on the other hand , the proof of proposition [ delta.from.r.connection.prop ] is also easily modified to show that @xmath694 , as is the proof of lemma [ lipschitz.condition.lem ] to show that @xmath690 is continuous . then , by virtue of proposition [ circle.probabilistic.prop ] and the monotonicity of @xmath690 , we have @xmath695 . we can use the probabilistic methods employed in section [ circular.probabilistic.section ] to construct large sets in @xmath696 . the proof is more complicated because it is to our advantage to endow different integers with different probabilities of belonging to our random set . although all of the constants in the proof could be made explicit , we are content with inequalities having error terms involving big - o notation . let @xmath697 be a real number and @xmath698 be an integer . there exists a set @xmath699 , where @xmath700 , with @xmath701 . [ probabilistic.bg.prop ] define mutually independent random variables @xmath702 , taking only the values 0 and 1 , by @xmath703 ( notice that @xmath704 for all @xmath705 . ) these random variables define a random subset @xmath706 of the integers from 1 to @xmath63 . we shall show that , with positive probability , @xmath9 is a large @xmath707 $ ] set with @xmath12 not much bigger than @xmath708 . the expected size of @xmath9 is @xmath709 if we set @xmath710 , then lemma [ our.application.lem ] tells us that @xmath711 < \exp\big ( \frac{-a_0 ^ 2}{2e_0 } \big ) = \frac13.\ ] ] now for any integer @xmath712 $ ] , let @xmath713 the number of representations of @xmath612 as @xmath714 with @xmath568 . ( here we adopt the convention that @xmath715 if @xmath612 is odd ) . notice that in this latter sum , @xmath716 and the @xmath717 are mutually independent random variables taking only values 0 and 1 , with @xmath718 = p_jp_{k - j}$ ] . thus the expectation of @xmath665 is @xmath719 using the inequalities @xmath704 and @xmath720 . if we set @xmath721 , then lemma [ our.application.lem ] tells us that @xmath722 < { \mathop{\rm pr}}[r_k > e_k + a ] < \exp\big ( \frac{-a^2}{3e_k } \big ) < \exp\big ( \frac{-a^2}{3(\gamma+1 ) } \big ) = \frac1{3n}\ ] ] for every @xmath612 in the range @xmath723 . note that @xmath724 trivially for @xmath612 in the range @xmath725 . therefore , with probability at least @xmath726 , the set @xmath9 has at least @xmath727 elements and satisfies @xmath728 for all @xmath729 . setting @xmath730 , we conclude that any such set @xmath9 is a set . this establishes the proposition . schinzel and schmidt @xcite conjectured that among all pdfs supported on @xmath353 $ ] , the function @xmath731 , \\ 0 & \text{otherwise } \end{cases}\ ] ] has the property that @xmath67 is minimal . we have @xmath732 , \\ \tfrac\pi2 - 2\arctan \sqrt{2x-1 } & x\in[\frac12,1 ] , \\ 0 & \text{otherwise } \end{cases}\ ] ] and so @xmath733 . we have adapted the function @xmath68 for our definition ( [ schinzel.pk.def ] ) of the probabilities @xmath734 ; the constant @xmath735 appears as the value of the last integral in eq . ( [ pi.over.2.appearance ] ) . if schinzel s conjecture were false , then we could immediately incorporate any better function @xmath68 into the proof of proposition [ probabilistic.bg.prop ] and improve the lower bound on @xmath736 . for any @xmath737 , we have @xmath738 if both @xmath739 and @xmath740 are sufficiently large in terms of @xmath65 . [ g.growing.lower.bound.thm ] in the proof of proposition [ probabilistic.bg.prop ] , we saw that @xmath741 and @xmath742 ; this implies that @xmath743 . therefore the size of the constructed set @xmath9 was at least @xmath744 this establishes the theorem . in this section use the lower bounds on @xmath541 established in section [ probabilistic.bg.section ] into upper bounds for @xmath2 . our first proposition verifies the statement of theorem [ delta.summary.thm](i ) . @xmath5 for @xmath745 . [ line.2e-1.prop ] we already proved in lemma [ delta.trivial.bounds.lem ] that @xmath40 for all @xmath43 . recall from lemma [ lipschitz.condition.lem ] that the function @xmath71 satisfies the lipschitz condition @xmath128 . therefore to prove that @xmath746 for @xmath745 , it suffices to prove simply that @xmath747 . for any positive integer @xmath12 , it was shown by the authors ( * ? ? ? * theorem 2(vi ) ) that @xmath748 we combine this with proposition [ delta.from.r.connection.prop ] and the monotonicity of @xmath71 to see that @xmath749 since @xmath71 is continuous by lemma [ lipschitz.condition.lem ] , we may take the limit of both sides as @xmath597 to obtain @xmath750 as desired . _ in light of the lipschitz condition @xmath128 , the lower bound @xmath40 for all @xmath43 also follows easily from the trivial value @xmath112 . the function @xmath32 is increasing on @xmath33 $ ] . [ de.over.e2.increasing.prop ] the starting point of our proof is the inequality ( * ? ? ? * theorem 2(v ) ) @xmath751 with the monotonicity of @xmath2 and proposition [ delta.from.r.connection.prop ] , this gives @xmath752 choose @xmath753 . let @xmath754 be such that @xmath755 and @xmath756 , which is possible by proposition [ delta.as.infimum.prop ] . by proposition [ circle.probabilistic.prop ] , we may choose sequences of integers @xmath757 and @xmath758 such that @xmath759 and @xmath760 as @xmath761 . this implies @xmath762 so that , again using the monotonicity and continuity of @xmath71 , @xmath763 as @xmath761 . this shows that @xmath764 as desired . we can immediately deduce two nice consequences of this proposition . @xmath34 exists . this follows from the fact that the function @xmath32 is increasing and bounded below by @xmath36 on @xmath33 $ ] by the trivial lower bound ( lemma [ delta.trivial.lower.bound.lem ] ) . @xmath765 for @xmath766 . [ part.iii.cor ] this follows from the value @xmath767 calculated in proposition [ line.2e-1.prop ] and the fact that the function @xmath32 is increasing . the corollary above proves part ( iv ) of theorem [ delta.summary.thm ] , leaving only part ( v ) yet to be established . the following proposition finishes the proof of theorem [ delta.summary.thm ] . @xmath768 for all @xmath43 . [ part.iv.prop ] define @xmath769 , so that @xmath770 . if we set @xmath771 in the proof of proposition [ probabilistic.bg.prop ] , then the sets constructed are sets with @xmath772 and have size at least @xmath773 from eq . ( [ second.main.term.eq ] ) . therefore , for these values of @xmath12 and @xmath63 , @xmath774 as @xmath63 goes to infinity , by the continuity of @xmath71 . on the other hand , we see by proposition [ delta.from.r.connection.prop ] that @xmath775 as @xmath63 goes to infinity . combining these two inequalities yields @xmath776 as desired . we group the problems in this section into three categories , although some problems do not fit clearly into any of the categories and others fit into more than one . notice that the upper bounds given in theorem [ delta.summary.thm ] are not too far from this conjecture , the difference between the constants @xmath778 and @xmath779 in the middle range for @xmath4 being the only discrepancy . in fact , we believe it might be possible to prove that the expression in conjecture [ bold.de.cnj ] is indeed an upper bound for @xmath2 by a more refined application of the probabilistic method employed in section [ probabilistic.bg.section ] . the key would be to show that the various events @xmath780 are more or less independent of one another ( as it stands we have to assume the worst that they are all mutually exclusive in obtaining our bound for the probability of obtaining a `` bad '' set ) . there are some intermediate qualitative results about the function @xmath2 that might be easier to resolve . it seems likely that @xmath2 is convex , for example , but we have not been able to prove this . a first step towards clarifying the nature of @xmath2 might be to prove that @xmath781 also , we would not be surprised to see accomplished an exact computation of @xmath782 , but we have been unable to make this computation ourselves . we do at least obtain @xmath783 in proposition [ delta.one.half.thm ] . note that conjecture [ bold.de.cnj ] would imply that @xmath784 . we do not believe that there is always a set with measure @xmath4 whose largest symmetric subset has measure precisely @xmath2 . in fact , we do not believe that there is a set with measure @xmath785 whose largest symmetric subset has measure @xmath786 , but we do not even know the value of @xmath787 . in proposition [ line.2e-1.prop ] , we showed that @xmath788 , but this was found by rather limited computations and is unlikely to be sharp . the quantity @xmath789 in theorem [ delta.summary.thm](iv ) is of the form @xmath790 , and thus any improvement in the bound @xmath791 would immediately result in an improvement to theorem [ delta.summary.thm](iv ) . we remark that conjecture [ bold.de.cnj ] implies that @xmath792 , which in turn would allow us to replace the constant @xmath778 in theorem [ delta.summary.thm](iv ) by @xmath779 . let @xmath391 be the class of functions @xmath793 satisfying @xmath393 on @xmath187 $ ] . how small can we make @xmath412 for @xmath794 ? we are especially interested in @xmath395 , but a solution for any @xmath152 may be enlightening . to give some perspective to this problem , note that a trivial upper bound for @xmath795 can be found by taking @xmath206 to be identically equal to 1 , which yields @xmath796 . one can find functions that improve upon this trivial choice ; for example , the function @xmath206 defined in eq . ( [ easy.k.def ] ) is an example where @xmath797 . on the other hand , since the @xmath158-norm of a sequence is a decreasing function of @xmath152 , parseval s identity immediately gives us the lower bound @xmath798 , and of course @xmath799 is the exact minimum for @xmath800 . we remark that proposition [ cf.first.result.prop ] and the function @xmath801 defined after the proof of corollary [ first.cor ] provide a stronger lower bound for @xmath802 . by direct computation we have @xmath803 , and by proposition [ cf.first.result.prop ] we have @xmath804 for any @xmath805 . together these imply that @xmath806 in particular , for @xmath395 we know the value of @xmath807 to within one part in ten thousand . the problem of determining the actual infimum for @xmath808 seems quite mysterious . we remark that green @xcite considered the discrete version of a similar optimization problem , namely the minimization of @xmath809 over all pdfs @xmath206 supported on @xmath810 $ ] . as mentioned at the end of section [ basic.argument.section ] , we used the inequality @xmath811 which is exact when @xmath12 takes on one non - zero value , i.e. , when @xmath12 is an nif . we apply this inequality when @xmath259 with @xmath68 supported on an interval of length @xmath36 , which usually looks very different from an nif . in this circumstance , the inequality does not seem to be best possible , although the corresponding inequality in the exponential sums approach of @xcite and in the discrete fourier approach of @xcite clearly is best possible . specifically , we ask for a lower bound on @xmath812 that is strictly greater than 1 . we know that this infimum is at most @xmath813 , and in fact conjecture [ infinity.two.cnj ] would imply that the infimum is exactly @xmath814 . more generally , for any subset @xmath117 of an abelian group endowed with a measure , we can define @xmath815 , where @xmath816 is defined in the same way as in eq . ( [ ddef ] ) . for example , @xmath817}({\varepsilon})$ ] is the function @xmath2 we have been considering throughout this paper , and @xmath688 was considered in section [ circular.probabilistic.section ] . most of the work in this paper generalizes easily from @xmath818 $ ] to @xmath818^d$ ] . we have had difficulties , however , in finding good kernel functions in higher dimensions . that is , we need functions @xmath819 such that @xmath820 is as small as possible , while @xmath821 if all components of @xmath822 are less than @xmath489 in absolute value . this restricts @xmath206 on one - half of the space in 1 dimension , one - quarter of the space in 2 dimensions , and only @xmath823 of the space in @xmath267 dimensions . for this reason one might expect that better _ kernels exist in higher dimensions , but the computational difficulties have prevented us from finding them .	a symmetric subset of the reals is one that remains invariant under some reflection @xmath0 . we consider , for any @xmath1 , the largest real number @xmath2 such that every subset of @xmath3 $ ] with measure greater than @xmath4 contains a symmetric subset with measure @xmath2 . in this paper we establish upper and lower bounds for @xmath2 of the same order of magnitude : for example , we prove that @xmath5 for @xmath6 and that @xmath7 for @xmath8 . this continuous problem is intimately connected with a corresponding discrete problem . a set @xmath9 of integers is called a @xmath10 $ } } $ ] set if for any given @xmath11 there are at most @xmath12 ordered pairs @xmath13 with @xmath14 ; in the case @xmath15 , these are better known as sidon sets . our lower bound on @xmath2 implies that every @xmath10 $ } } $ ] set contained in @xmath16 has cardinality less than @xmath17 . this improves a result of green for @xmath18 . conversely , we use a probabilistic construction of @xmath10 $ } } $ ] sets to establish an upper bound on @xmath2 for small @xmath4 . * ams msc ( 2000 ) : * * 05d99 extremal combinatorics , * 42a16 fourier series of functions with special properties , * 11b83 special sequences . * keywords : * ramsey theory , continuous combinatorics , sidon sets
nuclear clustering has been one of the main subjects in nuclear physics ; for a recent review , see ref . @xcite . as a new topic , @xmath4-clustering on the surface of heavy nuclei , sn isotopes , is theoretically predicted in ref . this result itself is interesting and important because it has been believed that @xmath4-clustering is developed mainly in light nuclei , although some indication for @xmath4-clustering in @xmath5ca and @xmath6ti was discussed @xcite . furthermore , the result gives a significant impact on the nuclear equation of state @xcite . as emphasized in ref . @xcite , however , one should keep it in mind that a large spectroscopic factor of @xmath4 does not necessarily indicate the @xmath4-clustering , because of the _ duality _ of the mean - field - type structure and the cluster structure @xcite . on the other hand , the localization of @xmath4 in the nuclear surface region is direct evidence of the @xmath4-clustering in a nucleus . in this view , the @xmath4 transfer reaction , ( @xmath7li,@xmath8 ) in particular , has been utilized for investigating the @xmath4-clustering . very recently , a three - body reaction model with a microscopic cluster wave function was applied to the @xmath9o(@xmath7li,@xmath8)@xmath10ne reaction and the transfer cross section was shown to be sensitive to the @xmath4 distribution on the nuclear surface of @xmath10ne @xcite . in the present study , as an alternative way to the @xmath4 transfer reaction , we consider the proton - induced @xmath4 knockout reaction on @xmath0sn , and investigate how clearly it can probe the @xmath4 distribution in the surface region of @xmath0sn , i.e. , the @xmath4-clustering of @xmath0sn . we adopt the distorted wave impulse approximation ( dwia ) framework to describe the ( @xmath1,@xmath2 ) reaction ; dwia has successfully been used in the analysis of various nucleon knockout @xcite and @xmath4 knockout @xcite experiments . in many preceding studies , however , the so - called factorization approximation , which factors out the nucleon - nucleon ( @xmath11 ) transition amplitude in the evaluation of the total transition matrix element of the knockout process , has been adopted . in this paper we explicitly examine the validity of the factorization approximation by means of the local semi classical approximation ( lsca ) @xcite to the distorted waves . it was argued in ref . @xcite that the factorization approximation becomes questionable when the distortion effect is large . it is thus important to examine its validity for the @xmath4 knockout process for a heavy nucleus , in which the distortion on @xmath4 by the reaction residue is expected to be very strong . in sec . [ secform ] we describe the dwia formalism for the ( @xmath1,@xmath2 ) reaction , with introducing the lsca that is a key prescription for discussing the accuracy of the factorization approximation . in sec . [ secresult ] first we show the comparison between the present calculation and the experimental data . next we discuss the validity of the factorization approximation in the @xmath0sn(@xmath1,@xmath2)@xmath3cd reaction at 392 mev . we then show that the @xmath0sn(@xmath1,@xmath2)@xmath3cd reaction probes the @xmath4 distribution in the surface region with high selectivity . the dependence of these findings on the @xmath4 wave function is also discussed . finally , a summary is given in sec . [ secsum ] . we consider the a(@xmath1,@xmath2)b reaction in normal kinematics in the dwia framework . the incoming proton in the initial channel is labelled as particle 0 , and the outgoing proton and @xmath4 are particles 1 and 2 , respectively . a ( b ) denotes the target ( residual ) nucleus . @xmath12 and @xmath13 ( @xmath14 ) represent the momentum and its solid angle , respectively , and @xmath15 ( @xmath16 ) is the total ( kinetic ) energy of particle @xmath17 . all quantities with and without superscript l indicate that they are evaluated in the laboratory ( l ) and center - of - mass ( c.m . ) frame , respectively . the transition amplitude in the dwia formalism is given by @xmath18 where @xmath19 , @xmath20 , and @xmath21 are the scattering wave functions of the @xmath1-a , @xmath1-b , and @xmath4-b systems , respectively , @xmath22 is the transition interaction between @xmath1 and @xmath4 , and @xmath23 is the @xmath4-cluster wave function . @xmath24 , @xmath25 , @xmath26 , and @xmath27 are , respectively , the principal quantum number , the orbital angular momentum , the total angular momentum , and its third component of @xmath4 in the nucleus a. the superscripts @xmath28 and @xmath29 specify the outgoing and incoming boundary conditions on @xmath30 , respectively . the definition of the coordinates is given in fig . [ figcoord ] . , @xmath2)b reaction . , scaledwidth=35.0% ] by using @xmath31 @xmath32 are written by @xmath33 where @xmath34 and @xmath35 with @xmath36 the mass number of a. we make the lsca @xcite that describes the propagation of the scattering wave for a short distance @xmath37 by a plane wave , i.e. , @xmath38 the norm of the local momentum @xmath39 is given by @xmath40 is determined so as to satisfy the local energy conservation : @xmath41 with @xmath42 and @xmath43 the reduced mass of the scattering particles and the distorting potential for particle @xmath17 , respectively . the direction of @xmath39 is taken to be parallel to the flux of @xmath44 . the validity of the lsca is discussed in sec . [ seclsca ] . equation ( [ kotmtx ] ) is then reduced to @xmath45 where @xmath46 and @xmath47 are defined by @xmath48 here , @xmath49 ( @xmath50 ) is the @xmath1-@xmath4 relative momentum in the initial ( final ) channel : @xmath51 @xmath52 is determined by the momentum conservation of the @xmath1-@xmath4 system : @xmath53 in taking the squared modulus of eq . ( [ redtmat ] ) , we make the on - the - energy - shell ( on - shell ) approximation to @xmath54 : @xmath55 where @xmath56 is the angle between @xmath49 and @xmath50 , i.e. , the local @xmath1-@xmath4 scattering angle , and @xmath57 is the local scattering energy defined by @xmath58 in eqs . ( [ onshell ] ) and ( [ locale ] ) @xmath59 is the reduced mass of the @xmath1-@xmath4 system . with the lsca and the on - shell approximation , the triple differential cross section ( tdx ) of the ( @xmath60 ) reaction is given by @xmath61 where @xmath62 is the spectroscopic factor of the alpha - cluster and the kinematical factor @xmath63 is defined by @xmath64^{-1}\end{aligned}\ ] ] with @xmath65 the jacobian from the c.m frame to the l frame , and @xmath66 the reduced transition amplitude is given by @xmath67 in the preceding studies on knockout reactions @xcite , further simplification of @xmath68 was made by replacing @xmath39 with the asymptotic momentum @xmath12 . we then obtain @xmath69 where @xmath70 and @xmath71 are given in the same way as for @xmath72 and @xmath57 , respectively , but with using the asymptotic @xmath1-@xmath4 relative momenta : @xmath73 this prescription is called the factorization approximation . one sees that this approximation is equivalent to use the asymptotic momentum @xmath12 instead of the local momentum @xmath39 in eq . ( [ eqlsca ] ) , i.e. , @xmath74 which we call the asymptotic momentum approximation ( ama ) . therefore the accuracy of the factorization approximation can be judged , in principle , by that of the ama . for the bound state wave function @xmath23 , we assume that the @xmath4 particle is bound in the 4@xmath75 orbit in a woods - saxon potential @xmath76)$ ] with @xmath77 fm and @xmath78 fm . the depth of the potential @xmath79 is adjusted so as to reproduce the @xmath4 separation energy of @xmath0sn , 4.81 mev . in the calculation shown below , the @xmath4 spectroscopic factor @xmath80 for @xmath0sn is taken to be 0.022 @xcite . it should be noted that the purpose of the present study is not to determine @xmath80 but to understand the property of the @xmath81 knockout reaction and to examine the reliability of dwia with the factorization approximation . one of the most important ingredients of the present dwia is the @xmath1-@xmath4 differential cross section @xmath82 that determines the transition strength of the @xmath81 process . because @xmath83 for various scattering energies and angles are needed , we adopt the microscopic single folding model @xcite with implementing the phenomenological nuclear density of @xmath4 and the melbourne @xmath11 @xmath84-matrix interaction @xcite . as shown in fig [ figelm ] , with no free parameter , the calculated @xmath85 agrees very well with the experimental data @xcite at 297 mev and 500 mev . calculated by the single folding model calculation and the experimental data @xcite at 297 mev and 500 mev . , scaledwidth=45.0% ] as for the distorting potential for @xmath4 in the final channel , for consistency , we employ the double folding model @xcite using the same ingredients as used in the @xmath1-@xmath4 calculation ; we use the nuclear density of @xmath3cd calculated by the hartree - fock method in the same way as ref . @xcite . it is known that to phenomenologically determine a low - energy scattering potential of @xmath4 is quite difficult because of the discrete ambiguities @xcite . in fact , there have been many attempts @xcite to microscopically determine an @xmath4 potential with the double - folding model approach . it should be noted , however , that in the present study we evaluate both the real and imaginary parts of the @xmath4 potential with no free adjustable parameter , in contrast to those preceding studies . for the distorting potential of proton in the initial and final channels , we use the edad1 parameter set of the dirac phenomenology @xcite . the coulomb terms of the distorting potentials are constructed by assuming that the target ( residual ) nucleus is a uniformly charged sphere with the radius of @xmath86 ( @xmath87 ) . the effect of the nonlocality of the proton and alpha distorting potentials is taken into account by multiplying the scattering waves by the perey factor @xcite @xmath88^{-1/2}$ ] , where @xmath89 is the reduced mass between the two scattering particles . the range of nonlocality @xmath90 for @xmath1 ( @xmath4 ) is taken to be 0.85 fm ( 0.2 fm ) @xcite . we take the following kinematical condition on the @xmath0sn(@xmath1,@xmath2)@xmath3cd reaction at 392 mev ; the madison convention is adopted . the kinetic energy of particle 1 is fixed at 328 mev and its emission angle is set to @xmath91 . as for particle 2 , @xmath92 is fixed at @xmath93 and @xmath94 is varied around @xmath95 ; the kinetic energy @xmath96 changes around 59 mev and @xmath97 , @xmath98 mev , accordingly @xcite . we always adopt the relativistic kinematics for all the scattering particles in this study . we test the present model calculation by comparing the calculated result of the energy sharing cross section , which is a tdx with fixed @xmath99 and @xmath100 , as a function of @xmath101 for @xmath102zn(@xmath1,@xmath2)@xmath103ni reaction with measured experimental data @xcite ; the incident energy is 101.5 mev . the present result and the experimental data are shown in fig . [ fig3 ] . zn(@xmath1,@xmath2)@xmath103ni reaction at 101.5 mev . the experimental data are from ref . , scaledwidth=35.0% ] the edad parameter set are used for the distorting potential of @xmath1-@xmath102zn and @xmath1-@xmath103ni , and the double folding model is adopted for @xmath4-@xmath103ni , in the same way as in [ subsecinput ] . according to ref . @xcite , we assume that the @xmath4 particle is bound in the 6@xmath75 state in a woods - saxon potential with @xmath104=1.30 fm and @xmath78 fm , and the depth of the potential @xmath79 is adjusted so as to reproduce the @xmath4 separation energy 4.58 mev . one can see that the present calculation well reproduces the observed energy sharing cross section ; the deduced @xmath4 spectroscopic factor is 0.84 , which is sizably larger than the value 0.42 obtained in the previous study @xcite . it should be noted , however , that the double folding model for the distorting potential of @xmath4-@xmath103ni will have some ambiguities due to the relatively low scattering energy of @xmath105 mev . furthermore , the calculated result in ref . @xcite showed quite large ambiguities ( @xmath106% ) of the deduced @xmath4 spectroscopic factors due to the @xmath4-@xmath103ni potential . considering these facts , it can be concluded that the present result is consistent with the experimental data and its analysis . the validity of the lsca for the scattering of nucleon has been examined in refs . @xcite and it was concluded that at energies higher than about 50 mev , the lsca works for the propagation within 1.5 fm . furthermore , at those energies the ama is found to work at almost the same level as of the lsca @xcite . considering the aforementioned kinematical condition on particles 0 and 1 , one may conclude that for proton both the lsca and the ama are valid in the description of the @xmath0sn(@xmath1,@xmath2)@xmath3cd reaction . on the other hand , such a validation for particle 2 , the knocked out @xmath4 particle , has not been done before . in fig . [ figlsca ] we show the validity of the lsca and the ama for @xmath107 with @xmath108 , which corresponds to the quasi - free condition , i.e. , the residual nucleus @xmath3cd is at rest in the l frame . figures [ figlsca](a ) and [ figlsca](b ) correspond to the propagation from @xmath109 , @xmath95 , @xmath110 and @xmath111 , @xmath112 , @xmath113 , respectively , in the spherical coordinate representation . with no approximation ( solid line ) , with the lsca ( dashed line ) , and with the ama ( dotted line ) are plotted . in figs . [ figlsca](a ) and [ figlsca](b ) , the propagation from @xmath114 , @xmath95 , @xmath110 and @xmath114 , @xmath112 , @xmath113 are investigated , respectively ; @xmath108 is chosen for the kinematics of the @xmath4 particle . + , title="fig:",scaledwidth=45.0% ] with no approximation ( solid line ) , with the lsca ( dashed line ) , and with the ama ( dotted line ) are plotted . in figs . [ figlsca](a ) and [ figlsca](b ) , the propagation from @xmath114 , @xmath95 , @xmath110 and @xmath114 , @xmath112 , @xmath113 are investigated , respectively ; @xmath108 is chosen for the kinematics of the @xmath4 particle . + , title="fig:",scaledwidth=45.0% ] in each panel the solid , dashed , and dotted lines show , respectively , the real part of the exact wave function , that with the lsca , and that with the ama . since @xmath115 ( @xmath116 ) corresponds to the foreside ( left side ) of @xmath3cd with respect to the outgoing @xmath4 , the distortion effect on @xmath4 at @xmath115 ( @xmath116 ) is weak ( strong ) . with weak distortion , as shown in fig . [ figlsca](a ) , both approximations work well within about 0.5 fm of the propagation . it should be noted that , with considering the range of the @xmath1-@xmath4 interaction of about 2 fm and the constant @xmath117 in front of @xmath118 in eq . ( [ r2 ] ) , the lsca and the ama are required to be valid for the propagation of about 0.4 fm . the two approximations are thus validated for the propagation from @xmath115 . in case of the strong distortion , as shown in fig . [ figlsca](b ) and suggested in ref . @xcite , the ama can not describe the behavior of the exact scattering wave function ; since the radial direction from @xmath116 is almost orthogonal to the direction of the asymptotic momentum @xmath119 , the dotted line is almost constant , whereas the solid line shows clear variation . on the other hand , the lsca reproduces well the exact solution at almost the same level as in the case of weak distortion . thus , one sees that the kinematics of @xmath4 at @xmath116 is significantly changed from that in the asymptotic region by the distorting potential of @xmath3cd ; this kinematical change is well traced by using the lsca , i.e. , the local momentum of the @xmath4 particle . therefore one can conclude that the lsca works for the @xmath4 scattering wave function that is strongly distorted , whereas the ama not . this may cast doubt on the use of the factorization approximation for the @xmath81 reaction investigated in the present study . in the following subsections we discuss this in view of the tdx . the calculated tdx is shown in fig . [ figmd ] as a function of the recoil momentum @xmath120 defined by @xmath121 the solid and dashed lines represent the results without and with the factorization approximation , respectively . one sees from the good agreement between the solid and dashed lines that the factorization approximation , or equivalently , the ama , affects very little the tdx , although the ama for @xmath4 is shown to be invalid around @xmath122 . this is due to the strong absorption of @xmath4 in that region as shown in sec . [ sec_ana_integ ] . the dotted line in fig . [ figmd ] represents the result of the plane wave impulse approximation ( pwia ) calculation divided by 200 . the renormalization factor @xmath123 shows the strong absorption mainly caused by the @xmath4-@xmath3cd distorting potential . in the pwia , the tdx is essentially proportional to the absolute square of the fourier transform of the @xmath4 distribution @xmath23 inside @xmath0sn . since we take a 4@xmath75 state , the dashed line in fig . [ figrinteg ] shown below , the tdx calculated with the pwia shows an oscillation pattern accordingly . the shape of the tdx calculated with the dwia is quite different from that with the pwia . the widening of the width of the tdx caused by distortion suggests that , because of the uncertainty principle , only a limited region of @xmath23 is probed by the @xmath81 reaction , as shown in sec . [ sec_ana_integ ] . it should be noted that the slight shift of the peak of the tdx with the dwia from @xmath124 is understood by the shift of the momentum of particles 2 due to the real part of the distorting potential @xcite . in fig [ figrinteg ] , we show by the solid line the absolute value of the integrand on the r.h.s . ( [ localtmt ] ) after integrated over the solid angle @xmath125 of @xmath126 : @xmath127 the plotted result corresponds to @xmath124 , i.e. , the quasi - free condition . at @xmath124 ( solid line ) , the same but calculated with only the @xmath4-@xmath3cd distorting potential @xmath128 ( dashed line ) , and the result with pwia ( dotted line ) . the results are normalized to unity at the peak position . , scaledwidth=45.0% ] the dashed line shows @xmath129 calculated with including only @xmath128 , the distorting potential of the @xmath4-@xmath3cd system in the final state , and the dotted line shows that with pwia . each line is normalized to unity at the peak position . one sees that the magnitude of @xmath130 is strongly suppressed in the interior region , @xmath131 fm , mainly because of the absorption due to the @xmath4-@xmath3cd distorting potential . the slight shift of the peak position is due to the suppression in the interior region . it should be noted , however , that the product of the oscillating three distorted waves and a bound - state wave function can make nontrivial cancellation . this property also may contribute to the aforementioned suppression . furthermore , in fig . [ figrmin ] the tdxs calculated with changing the minimum value @xmath132 of the integration over @xmath133 are shown ; we take @xmath134 0 , 6 , 6.5 , 7 , and 8 fm . but with changing @xmath132 . the solid , dashed , dotted , dot dashed and two - dot dashed lines correspond to @xmath134 0 , 6 , 6.5 , 7 , and 8 fm , respectively . , scaledwidth=45.0% ] it is found that the calculated tdx does not change for @xmath135 05.5 fm , and decrease drastically for @xmath135 68 fm . the slight increase of tdx with @xmath136 fm is due to the interference of the integrand . this result shows that the ( @xmath1,@xmath2 ) reaction on heavy nuclei probes the @xmath4-cluster wave function on the nuclear surface with high selectivity , as required for the reaction to be a good probe for @xmath4-clustering . with this peripherality of the ( @xmath1,@xmath2 ) reaction , one can understand naturally the mechanism that makes the width of the tdx wider when the distortion is taken into account . for more detailed analysis , the absolute value of the integrand on the r.h.s . of eq ( [ localtmt ] ) @xmath137 on the @xmath138-@xmath139 plane for @xmath1400 , 1 , 3 , 5 , 6 , and 7 fm are shown in fig . [ figrinteg2d](a)fig . [ figrinteg2d](f ) . for @xmath140 0 , 1 , and 3 fm , it is clearly seen that the amplitude is located in the foreside region with @xmath141@xmath142 fm , where @xmath143 is not absorbed and @xmath144 has a finite amplitude . for @xmath145 fm , the localization of the amplitude becomes rather vague , because the absorption property of @xmath143 does not strongly depend on @xmath138 and @xmath139 for such values of @xmath146 . nevertheless , one may see that the main part of @xmath147 exist in the foreside region . figures [ figrinteg2d](a)[figrinteg2d](f ) therefore show that the @xmath81 reaction has selectivity not only in the radius but also the direction of the target nucleus . it is found that the peak at the rear side on @xmath148 plane , around @xmath149@xmath150 , @xmath151 , @xmath113 in fig . [ figrinteg2d](a ) comes from the focus of @xmath152 due to the attraction of the distorting potential and the increase in @xmath85 caused by that . it should be noted that this rear - side peak exists only at around @xmath148 as shown in fig . [ figrinteg2d ] , and makes no major contribution to the tdx . in fact , it is found that about 90% of the tdx comes from the @xmath153 region . this means that the possible interference between the amplitudes in the foreside and rear - side regions is very small , which realizes an intuitive picture that the ( @xmath60 ) reaction of our interest takes place in a limited region of space . these features support that the ama is valid for the calculation of the tdx . since a very naive model for @xmath154 is adopted in the present study , it is important to see the @xmath154 dependence of the findings discussed above . it is obvious that the validity of the lsca itself has nothing to do with @xmath154 . thus , we discuss the @xmath154 dependence of the tdx as well as effect of the ama on that . . the solid ( dotted ) line is the tdx with @xmath155 ( @xmath156 ) fm . the dashed line is the same as fig . [ figmd ] for comparison . , scaledwidth=45.0% ] in fig . [ figrange ] the solid ( dotted ) line shows the tdx calculated with @xmath154 with increasing ( decreasing ) the range parameter @xmath104 by 10% , @xmath155 ( @xmath156 ) fm ; these results are obtained by using the lsca . the dashed line is the same as the solid line in fig . [ figmd ] . one can see that the 10% difference of @xmath104 changes the magnitude of the tdx significantly , i.e. , about a factor of three difference . this is also understood by the absorption in the interior region . since only the surface region contributes to the tdx , small extension of @xmath154 to the exterior region changes the magnitude of tdx drastically . it is found that the tdx at @xmath124 calculated with the ama differs from that with the lsca by only 6% at most . furthermore , the qualitative features shown in figs . [ figrmin ] and [ figrinteg2d ] turned out to be independent of @xmath104 . we have examined the @xmath0sn(@xmath1,@xmath2)@xmath3cd reaction at 392 mev in the dwia framework . to show the validity of the dwia model , we have demonstrated that it reproduces the observed energy sharing cross section data of @xmath102zn(@xmath60)@xmath103ni at 101.5 mev . it was clarified that the so - called factorization approximation adopted in many preceding studies is equivalent to the ama to the distorted waves , which is a further simplification of the lsca . although the ama does not work for the propagation of @xmath4 in the region where the nuclear deflection is significant , it does not affect the tdx because of the strong absorption in that region . in other words , the integrand of the transition matrix has a contribution only in the region where the ama works well . as a result , the factorization approximation was verified for the calculation of the tdx of the @xmath81 reaction . it should be kept in mind , however , that the inaccuracy of the ama may affect the tdx if a scattering particle feels a potential having a strong real part and a weak imaginary part ; this can be realized , for instance , for nucleon scattering at lower energies . the strong absorption due to the @xmath4-@xmath3cd distorting potential makes the @xmath81 reaction very peripheral , which allows one to clearly probe the @xmath4-clustering of nuclei . furthermore , the @xmath81 reaction has high selectivity also in the direction of the target nucleus ; only the foreside region with respect to the emitting @xmath4 with the radius of 69 fm is probed . it is also shown that the factorization approximation and the peripherality of the reaction are valid for different choices of @xmath154 , but the magnitude of tdxs are strongly dependent on them . this result suggests that it is essential to employ a reliable alpha - cluster wave function for the qualitative discussion . validation of the on - shell approximation to the @xmath1-@xmath4 transition amplitude will be important for more reliable description of the knockout processes . the authors thank s. kawase and t. uesaka for fruitful discussions . the computation was carried out with the computer facilities at the research center for nuclear physics , osaka university . this work was supported in part by grants - in - aid of the japan society for the promotion of science ( grants no . jp15j01392 and no . jp25400255 ) and by the impact program of the council for science , technology and innovation ( cabinet office , government of japan ) . c. w. wang , p. g. roos , n. s. chant , g. ciangaru , f. khazaie , d. j. mack , a. nadasen , s. j. mills , r. e. warner , e. norbeck , f. d. becchetti , j. w. janecke , and p. m. lister , phys . c * 31 * , 1662 ( 1985 ) . proceedings of the 2nd louvain cracow seminar on the alpha nucleus interaction , louvain - la - neuve,1978 , edited by g.gregoire and k. grotowski ( universite de lourain - la - neuve , 1978 ) , and references therein .	the @xmath0sn(@xmath1,@xmath2)@xmath3cd reaction at 392 mev is investigated with the distorted wave impulse approximation ( dwia ) framework . we show that this reaction is very peripheral mainly because of the strong absorption of @xmath4 by the reaction residue @xmath3cd , and the @xmath4-clustering on the nuclear surface can be probed clearly . we investigate also the validity of the so - called factorization approximation that has frequently been used so far . it is shown that the kinematics of @xmath4 in the nuclear interior region is significantly affected by the distortion of @xmath3cd , but it has no effect on the reaction observables because of the strong absorption in that region .
recently there has been considerable progress in the determination of the vector and axial - vector hadronic spectral functions from the semileptonic decays of the @xmath0 lepton @xcite@xcite . using the qcd predictions for the moments of these spectral functions one may obtain constraints on @xmath3 and the parameters characterizing the nonperturbative qcd dynamics ( @xcite@xcite , @xcite@xcite ) . the accuracy of the obtained value of @xmath3 appears to be quite high . in order to have a proper understanding of the phenomenological relevance of this determination of @xmath3 it is important to make a careful estimate the theoretical uncertainties in the qcd predictions @xcite@xcite . in this note we investigate the uncertainties in the evaluation of the perturbative part of the qcd prediction for the spectral moments , for which the next - to - next - to - leading order ( nnlo ) approximation is available . we study in some detail the sensitivity of the perturbative qcd predictions to the choice of the renormalization scheme ( rs ) and the effect of the rs dependence on the fitted values of the qcd parameters . let us begin by summarizing the theoretical framework adopted in the analysis of the @xmath0 decay data @xcite@xcite . in @xcite it was suggested to use the @xmath4 moments , defined by the relation @xmath5 where @xmath6 denotes the invariant mass distribution for the cabbibo allowed semileptonic @xmath0 decays in the vector ( v ) or axial - vector ( a ) channel and @xmath7 denotes the electronic width of the @xmath0 lepton . the qcd predictions for @xmath4 have the form : @xmath8 where @xmath9 is the ckm matrix element ( @xmath10 ) and @xmath11 represents the correction from electroweak interactions @xcite . the @xmath12 factor denotes the parton model prediction for later use we shall need @xmath13 and @xmath14 . the @xmath15 term denotes the perturbative contribution , evaluated for 3 massless quarks . ( the @xmath16 and @xmath17 quark mass effects are negligible , the @xmath18 quark mass effects are very small for non - strange decays and the @xmath19 quark is considered to be decoupled @xcite . ) the @xmath15 term is universal for the v and a channels . the @xmath20 term in ( [ eq : rklpred ] ) denotes the contribution from the nonperturbative qcd effects , which are estimated using the svz approach @xcite : @xmath21 where @xmath22 are the vacuum expectation values of the gauge invariant operators of dimension @xmath23 and @xmath24 are coefficients specific for the considered spectral moment and the type of the operator . the @xmath24 coefficients are in principle power series in the strong coupling constant . the object of greatest interest is of course the total decay rate for the cabbibo allowed semileptonic @xmath0 decays in the vector / axial - vector channels , @xmath25 . the perturbative correction to this moment is sizeable and it is highly sensitive to the value of the strong coupling , due to the low characteristic energy scale of @xmath26 . the higher @xmath27 moments are introduced to take advantage of the the full information contained in the hadronic spectral functions . by using the predictions for several @xmath27 moments one may obtain a simultaneous fit of @xmath1 and of some of the condensates @xmath22 . in this way the whole analysis becomes self - consistent and in addition one obtains a check on both the perturbative and the nonperturbative qcd contributions . the perturbative qcd corrections @xmath28 are evaluated using a contour integral expression @xcite , which relates them to the qcd correction @xmath29 to the so called adler function @xcite , i.e. the logarithmic derivative of the transverse part of the vector / axial - vector current correlator @xmath30 : @xmath31 . \label{eq : defadler}\ ] ] ( in the approximation of massless quarks the perturbative contributions to the adler functions for the vector and axial - vector current correlators are identical . ) we have : @xmath32 where @xmath33 is a weight function specific to the considered moment and @xmath34 is a contour running clockwise from @xmath35 to @xmath36 away from the region of small @xmath37 . in the following we shall need the weight functions @xmath38 and @xmath39 : @xmath40 @xmath41 the nnlo renormalization group improved perturbative expansion for @xmath29 may be written in the form : @xmath42 , \label{eq : dpert}\ ] ] where @xmath43 denotes the running coupling constant that satisfies the nnlo renormalization group ( rg ) equation : @xmath44 in the modified minimal subtraction @xmath2 scheme ( i.e. using @xmath2 subtraction procedure and choosing @xmath45 ) we have @xcite@xcite for @xmath46 @xmath47 and @xmath48 . the renormalization group coefficients for @xmath46 are @xmath49 , @xmath50 and @xmath51 . the qcd predictions for the @xmath28 are usually calculated in the modified minimal subtraction ( @xmath2 ) renormalization scheme @xcite . however , in the nnlo approximation with massless quarks there is a two - parameter freedom in choosing the rs . this is a consequence of the fact that in each order of perturbation expansion the finite parts of the renormalization constant for the coupling constant may be chosen arbitrarily . different choices of the finite parts of the renormalization constant result in different definitions of the coupling constant , which are related by a finite renormalization . this results in a change of values of the coefficients @xmath52 , @xmath53 and @xmath54 . ( we restrict our discussion to the class of mass and gauge independent schemes , for which the coefficients @xmath55 and @xmath56 are universal . ) the formulas describing how the redefinition of the coupling affects the coefficients @xmath57 and @xmath54 are collected for example in @xcite . the dimensional qcd parameter @xmath58 also depends on the choice of the rs @xcite . in the nnlo there exists however a rs invariant combination of the expansion coefficients @xcite@xcite : @xmath59 for the @xmath60 we have @xmath61 . the change in the expansion coefficients and the change in the coupling constant compensate each other , but of course in the finite order of perturbation expansion such compensation may only be approximate , which results in the numerical rs dependence of the perturbative predictions . this rs dependence is formally of higher order in the coupling constant , but in the case of the @xmath0 decay it may be significant numerically , since the coupling is not very small at the energy scale of @xmath62 . it is therefore very important to verify to what extent the rs dependence affects the predictions and the fits to the experimental data . the authors of @xcite used the @xmath27 moments with @xmath63 and @xmath64 , and fitted the @xmath65 condensates . since the aim of this note is primarily to study the theoretical uncertainties in the whole procedure we shall adopt a simplified approach , which still has considerable phenomenological relevance . this approach is based on the fact that the dominant nonperturbative contribution to @xmath25 in the svz expansion comes from the @xmath66 term @xcite , because the @xmath67 term is suppressed by additional power of @xmath3 . we shall therefore neglect the @xmath67 contribution to the total decay rate and for consistency the higher order correction to the @xmath66 coefficient . ( such an approximation was in fact made already in @xcite . ) in the following we shall also neglect contributions from the @xmath68 condensates . in order to be able to fit the @xmath66 condensate together with @xmath3 we shall use the qcd prediction for the @xmath69 moment , which similarly to @xmath25 has a suppressed contribution from the @xmath67 condensate . neglecting in @xmath69 the contributions from @xmath68 condensates we obtain a simple set of self - consistent formulas . in our approximation : @xmath70 where @xmath71 is the leading @xmath66 contribution to the transverse part of the hadronic vacuum polarization function : @xmath72 this contribution is dominated by the four - quark condensates @xcite . in the phenomenological analysis it is usually expressed in a simplified form , motivated by the chiral symmetry and the vacuum saturation approximation @xcite : @xmath73 where @xmath74 is the quark condensate , @xmath75 , @xmath76 and @xmath77 is an effective parameter , characterizing the deviation from the strict vacuum saturation . in our study of the rs dependence effects we parametrize the freedom of choice of the rs by the parameters @xmath52 and @xmath54 , following the conventions of our previous work @xcite . to obtain the perturbative qcd corrections @xmath28 we evaluate the contour integral numerically , using under the integral a numerical solution of the rg equation ( [ eq : rge ] ) in the complex energy plane . in this way we take full advantage of the rg - invariance properties of the perturbative prediction and we resum to all orders some of the large terms which would otherwise appear in the perturbation expansion @xcite . we assume that the integration contour @xmath34 is a circle @xmath78 , @xmath79 $ ] . to determine the numerical value of the running coupling constant on the contour @xmath34 we solve the transcendental equation , which results from integration of the rg equation ( [ eq : rge ] ) with a suitable boundary condition and analytic continuation to the complex energy plane : @xmath80 where in nlo @xmath81 and in nnlo for @xmath82 @xmath83 the presence of @xmath84 in the expression valid for arbitrary scheme follows from our taking into account explicitly the relation between @xmath58 parameters in different schemes @xcite and using @xmath84 as a reference parameter . after evaluating the predictions for @xmath85 and @xmath86 we perform the fits to the experimental data for @xmath25 and @xmath87 . we use the experimental values reported recently by aleph @xcite : @xmath88 , @xmath89 , @xmath90 , @xmath91 . in the fits we assume for simplicity that the experimental errors for @xmath92 and @xmath93 are not correlated . we express the results of our fits in terms of @xmath94 and @xmath95 . ( we actually fit the value of @xmath84 , which we then convert to @xmath94 using the rg equation in the @xmath96 scheme . ) for comparison with other determinations of the strong coupling constant we extrapolate @xmath3 from @xmath97 to @xmath98 . our procedure for extrapolation relies on the matching formula relating @xmath99 to @xmath100 : @xmath101 where @xmath102 and @xmath103 is the running quark mass @xmath104 of the heavy quark evaluated at the scale @xmath105 . ( the nnlo matching formula of that form was originally proposed in @xcite , see also discussion in @xcite . however , in @xcite it was found that a numerical coefficient in the nnlo term is actually different , which was subsequently confirmed in @xcite . we use the coefficient of @xcite . ) in order to evolve @xmath3 from scale @xmath106 to the scale @xmath107 we solve the equation : @xmath108 to obtain @xmath109 from the given value of @xmath1 we first evolve @xmath3 from the scale @xmath62 to the scale of @xmath110 using the @xmath46 rg equation , then we use the matching formula ( [ eq : match ] ) to obtain @xmath111 , evolve this to the scale of @xmath112 using the @xmath113 rg equation , use the matching formula to obtain @xmath114 , and finally evolve this to the scale of @xmath115 using the @xmath116 rg equation . we use @xmath117 and @xmath118 , which are the central values recommended by the review of particle properties @xcite . in table [ tb : match ] we give for reference some values of @xmath1 and @xmath109 as a function of @xmath119 in the nnlo approximation . .table of values of @xmath1 and @xmath109 in nnlo related by the matching procedure described in the text . [ cols="^,^,^,^",options="header " , ] to have a broader picture of the renormalization scheme dependence we perform the fit in a more general class of schemes . it is clear that regardless of our choice of the concrete optimal scheme there is a continuum of schemes , which seem to be equally reasonable . predictions in such schemes should also be somehow taken into account in the phenomenological analysis . a natural way to do this is to supplement the prediction in a preferred scheme with an estimate of the variation of the predictions over a whole set of _ a priori _ acceptable schemes . a concrete realization of this idea was presented in @xcite , based on the existence of the rs invariant @xmath120 , which provides a natural rs independent characterization of the magnitude of the nnlo corrections for the considered physical quantity . in @xcite it was proposed to calculate variation of the predictions over the set of schemes for which the expansion coefficients satisfy the condition : @xmath121 where @xmath122 a motivation for the condition ( [ eq : condition ] ) is that it eliminates schemes in which the expansions ( [ eq : dpert ] ) and ( [ eq : rge ] ) involve unnaturally large expansion coefficients that introduce large cancellations in the expression for the rs invariant @xmath120 . the constant @xmath123 in the condition ( [ eq : condition ] ) controls the degree of cancellation that we want to allow in the expression for @xmath120 . in our case we have for the pms parameters @xmath124 , so in order to take into account the schemes , which have the same or smaller degree of cancellation as the pms scheme we take @xmath125 . one may expect , that the estimate of the rs dependence obtained according to this prescription would be useful for a quantitative comparison of reliability of perturbative predictions for different physical quantities , evaluated at different energies . it is also clear that any large variation of the predictions over a set of schemes satisfying the constraint ( [ eq : condition ] ) with @xmath125 would be an unambiguous sign of a limited applicability of the nnlo expression . in fig . [ fig : alfitvecrsdep ] we show how the value of @xmath94 resulting from the fit depends on the parameters @xmath52 and @xmath54 specifying the renormalization scheme in nnlo . in the region of scheme parameters satisfying condition ( [ eq : condition ] ) with @xmath125 the minimum value of @xmath126 ( @xmath127 , @xmath128 ) is attained for @xmath129 and @xmath130 , and the maximum value of @xmath131 ( @xmath132 , @xmath133 ) is attained for @xmath134 and @xmath130 . it should be noted that the @xmath2 scheme parameters lie outside the @xmath125 region , but the fitted value of @xmath1 in this scheme coincides with the maximal value quoted above . it is also important to note that one of the other potentially interesting schemes , for which the nnlo expressions recently became available @xcite the so called v scheme @xcite corresponds to the rs parameters @xmath135 and @xmath136 , which lie very far outside the @xmath125 region . it seems that the large value of @xmath54 in the v scheme restricts its usefulness for the low energy perturbative predictions , because of the influence of the landau pole in the rg equation . it is of some interest to perform the same fits using instead the nlo predictions . using the nlo predictions in the @xmath2 scheme , we obtain @xmath137 and @xmath138 , which translates via nlo rg equation into @xmath139 , which in turn corresponds via nlo extrapolation to @xmath140 . we see that the value of @xmath141 is practically identical to that obtained in the nnlo fit . however , the value of the strong coupling constant is surprisingly high and it is significantly different from the nnlo value . using the nlo predictions in the pms scheme we obtain @xmath142 and @xmath143 , which corresponds to @xmath144 and @xmath145 . we see that although the difference in the nlo and nnlo values of @xmath84 obtained in the pms fits is still considerable , it is nevertheless much smaller than in the case of the @xmath2 scheme . ( if we look at @xmath146 instead of @xmath84 this difference is even smaller , although this may be partially a result of a fortuitous compensation between different factors influencing the evaluation of @xmath3 in various orders . ) for completeness we give the nlo results in the ec scheme : @xmath147 and @xmath148 ( @xmath149 ) . these numbers provide a nice illustration of the fact , that the difference between the nlo and nnlo predictions is strongly scheme dependent . therefore , if such a difference is to be used in any way to estimate the precision of the qcd prediction , some way of making a preferred choice of the renormalization scheme must be employed . so far our discussion was concentrated primarily on the value of @xmath94 coming from the fit . in fig . [ fig : d6fitvecrsdep ] we show the rs dependence of the fitted value of @xmath141 . we find that in the set of schemes satisfying the constraint ( [ eq : condition ] ) with @xmath125 the fitted value of @xmath141 changes in the range 0.01450.0182 . using the eq . ( [ eq : vacsat ] ) we may translate this into the range @xmath150 for the commonly used parameter @xmath151 . this seems to be in reasonable agreement with the values obtained previously by other authors , for example @xmath152 obtained in the original work of svz @xcite or @xmath153 obtained in a more recent analysis @xcite . similarly as in the case of the vector channel we start with the fit in the @xmath2 scheme . we obtain @xmath154 and @xmath155 , which corresponds to @xmath156 and @xmath157 . the result of this fit should be compared with @xmath158 , obtained by aleph in a more complete fit @xcite . we find that the difference between our results and the aleph result is slightly bigger than in the case of vector channel . it should be noted however that the aleph fit in the axial - vector channel has surprisingly large @xmath67 contribution , which would be difficult to justify theoretically . performing the nnlo fit in the pms scheme we obtain @xmath159 and @xmath160 , which corresponds to @xmath161 and @xmath162 . similarly as in the case of the vector channel we find appreciable reduction in the extracted value of @xmath1 as compared to the values obtained in the @xmath2 scheme . to make our results useful in the case of future improvements of the experimental analysis we show in fig . [ fig : ald6ax ] the results of the nnlo fit of @xmath1 and @xmath163 , obtained with the pms predictions , as a function of the experimental values of @xmath164 and @xmath165 . for comparison we indicate the results of the nnlo fit in the @xmath2 scheme ( dashed lines ) . in fig . [ fig : alfitaxrsdep ] we show how the fitted value of @xmath1 depends on the rs parameters @xmath52 and @xmath54 . in the region of the scheme parameters satisfying the condition ( [ eq : condition ] ) with @xmath125 we have @xmath166 ( @xmath167 , @xmath168 ) . similar figure may be obtained for the rs dependence of @xmath163 : it is found that @xmath169 . having obtained the results for @xmath95 in the vector and axial - vector channels it is of some interest to verify the simple relation between them , implied by the generalized vacuum saturation approximation : @xmath170 . using the numbers from the nnlo pms fits in both channels we obtain @xmath171 . in order to have a full picture of the perturbative uncertainties in the axial - vector channel we also perform the nlo fits . using the @xmath2 scheme , we obtain @xmath172 and @xmath173 , which translates via nlo rg equation into @xmath174 , which in turn corresponds via nlo extrapolation to @xmath175 . in the pms scheme we obtain @xmath176 and @xmath177 , which implies @xmath178 and @xmath179 . similarly as in the vector channel case we note a rather large difference between nlo and nnlo in the case of the @xmath2 scheme , which is significantly reduced if the preferred scheme is pms . in nlo in the ec scheme we obtain @xmath176 and @xmath180 ( @xmath181 ) . our discussion of the rs dependence of the qcd predictions and the fits may be summarized as follows : * changing the scheme from the @xmath2 scheme to the pms scheme we obtain a reduction in the extracted value of @xmath94 by approximately 0.01 ( @xmath182 is reduced by 0.001 ) . also , the difference between the nlo and nnlo results is much smaller in the pms scheme than in the @xmath2 scheme . * varying the scheme parameters @xmath52 and @xmath54 in the region satisfying the condition ( [ eq : condition ] ) with @xmath125 we obtain an uncertainty in the extracted value of @xmath94 of approximately 0.02 ( uncertainty in @xmath182 is 0.002 ) . as was argued in the text , this set of scheme parameters seems to be a minimal set that one should take into account in the discussion of the rs dependence . our general conclusion is that the perturbative predictions for the qcd effects in the inclusive decay rates for the semileptonic @xmath0 decays appear to be relatively precise , despite the rather low energy scale . it should be emphasized however , that in the discussion of the final precision of @xmath3 extracted from the @xmath0 decays one should also take into account the approximate character of the svz expansion itself @xcite . k. g. chetyrkin , a. l. kataev and f. v. tkachov , phys . b 85 ( 1979 ) 277 , m. dine and j. sapirstein , phys . ( 1979 ) 668 , w. celmaster and r. j. gonsalves , phys . 44 ( 1979 ) 560 , phys . d 21 ( 1980 ) 3112 .	the qcd corrections to the moments of the invariant mass distribution in the semileptonic @xmath0 decays are considered . the effect of the renormalization scheme dependence on the fitted values of @xmath1 and the condensates is discussed , using a simplified approach where the nonperturbative contributions are approximated by the dimension six condensates . the fits in the vector and the axial - vector channel are investigated in the next - to - leading and the next - to - next - to - leading order . the next - to - next - to - leading order results are found to be relatively stable with respect to change of the renormalization scheme . a change from the @xmath2 scheme to the minimal sensitivity scheme results in the reduction of the extracted value of @xmath1 by 0.01 . pacs 13.35.dx , 12.38.cy , 11.25.db .4 cm
attempting to uncover the mix of stellar ages and metallicities within a composite system such as an elliptical galaxy can be compared to trying to puzzle out the recipe for an elaborate gourmet dish on the basis of appearance and aroma alone . the basic ingredients may be obvious , but without the benefit of actually biting in , deciding how to prepare and combine them to reproduce the final product is guesswork . historically , most attempts at modeling the stellar populations within elliptical galaxies have concentrated on reproducing broad - band photometric colours using various ` empirical ' or ` evolutionary ' schemes ( e.g. , spinrad & taylor 1971 ; faber 1972 ; tinsley 1972 , 1978 ; bruzual 1983 ) to combine stars of different masses , metallicities , and/or ages . more recently , stellar population models have also included predictions for the absorption line strengths in the integrated spectra ( e.g. , worthey 1994 ; buzzoni 1995 ; weiss , peletier , & matteucci 1995 ; borges et al . 1995 ; vazdekis et al . the combination of broadband colours and absorption line indices allows for better constraints on the stellar content of galaxies . however , the spectral indices reveal at least two clear shortcomings of stellar population synthesis : model populations of reasonable ages ( @xmath7gyr ) can not reproduce the weak balmer absorption indices observed in the more metal rich galactic globular clusters ( cohen , blakeslee , & ryzhov 1998 ; gibson et al . 1999 ; vazdekis & arimoto 1999 ) and the model metal abundances do not match the ratios observed in giant ellipticals or metal - rich globulars , both of which exhibit enhancement of mg and the other alpha elements ( e.g. , peletier 1989 ; worthey , faber , & gonzalez 1992 ; vazdekis et al . 1997 ; worthey 1998 ; cohen et al . moreover , because of the degenerate behaviour of the integrated colours and indices with respect to age and metallicity variations ( see worthey 1994 ) , even some of the successes of the models prove ambiguous . clearly , new insights beyond those afforded by the usual measures of the integrated stellar spectrum are needed for unraveling the stellar content of early - type galaxies . observations of surface brightness fluctuations ( sbf ) provide a powerful , and hitherto fairly neglected , class of constraints for stellar population models . the sbf method is most well known for studies of the extragalactic distance scale ( e.g. , tonry & schneider 1988 ; tonry , ajhar , & luppino 1990 ; tonry et al . 1997 , hereafter sbf - i ; lauer et al . 1998 ; pahre et al . 1999 ) and peculiar velocity field ( tonry et al . 2000a , hereafter sbf - ii ; blakeslee et al . however , discussions of optical sbf in the context of stellar populations studies have been given by tonry et al . ( 1990 ) , worthey ( 1993a , b ) , buzzoni ( 1993 ) , ajhar & tonry ( 1994 ) , and sodemann & thomsen ( 1996 ) . the implications of near - infrared sbf observations for distances and stellar populations have been discussed by luppino & tonry ( 1993 ) , pahre & mould ( 1994 ) , jensen , luppino , & tonry ( 1996 ) , and jensen , tonry , & luppino ( 1998 ) . blakeslee , ajhar , & tonry ( 1999a ) give a recent comprehensive review of the sbf method and its applications . more recently , liu , charlot , & graham ( 2000 ) have presented sbf predictions using a completely updated version of the bruzual & charlot ( 1993 ) stellar population models . their work allows for an up - to - date and independent ( appearing after this paper was submitted ) comparison to our own . sbf distances are based on a measurement of the quantity , the apparent magnitude of the luminosity - weighted mean luminosity of the individual stars within a galaxy . this mean luminosity , called , is a well - defined property of any stellar system in any given bandpass . sbf thus begins to provide some taste of the constituent stars of a galaxy . multiband sbf observations yield distance - independent ` fluctuation colours , ' e.g. , . the behaviour of the absolute sbf magnitudes and fluctuation colours , when plotted for instance against integrated colours , absorption indices , or each other , must be reproduced before any population synthesis model can be deemed fully successful . in the following section , we present new @xmath0-band sbf measurements for five galaxies in the fornax cluster . most sbf observations to date have been done in the @xmath2-band and near - ir bands , as the fluctuations in old stellar populations are much brighter in these bands and expected to show less scatter . the only sizable @xmath0-band sbf data set so far published has been by tonry et al . ( 1990 ) for galaxies in the virgo cluster . however , quantities such as may be useful in stellar population studies , and because of its compactness , the fornax cluster is a favourite testbed for such ideas . in sec . [ sec : models ] , we describe an updated set of the models of vazdekis et al . ( 1996 ) and present sbf magnitude predictions from these . we compare these predictions to observations of elliptical galaxies and globular clusters in sec . [ sec : disc ] . the comparisons provide insight into the observations and help in gauging the accuracy of theoretical calibrations of the sbf distance method . in sec . [ sec : theorydists ] we construct composite populations from our simple population models and use these to derive theoretical calibrations as a function of . we also compare the mg@xmath5 and indices as sbf calibrators . [ sec : nbar ] discusses the new tonry fluctuation number ( tonry et al . 2000b , hereafter sbf - iv ) and provides an explanation of its observed properties . the final section summarizes the major results and conclusions . = ? [ cols="^,^,^,^,^,^,^,^,^,^,^,^ " , ] the @xmath2-band sbf distance indicator has also been calibrated against the mg@xmath5 absorption index ( thomsen et al . this calibration , using data from sodemann & thomsen ( 1995 ) and the mg@xmath5 measurements of davies , sadler , & peletier ( 1993 ) , is based solely on the internal gradients in ngc3379 . it is only strictly applicable over the range @xmath8 . the top panels of figure [ fig : mg2cal ] show the corresponding relation for our composite models . the relation in the top right panel is @xmath9\,,\;\ ; \pm\,0.07\,\hbox{mag } \ , . \label{eq : ibarmg2}\ ] ] the slope agrees well with the thomsen et al . calibration , but the zero point is again 0.15mag too faint . although the scatter is small , calibrating the observations is much more difficult . it requires a well - orchestrated combination of imaging and spectroscopy , and absorption measurements are difficult well outside the bright galaxy centre . also , there is the unresolved ( and unmodeled ) issue over whether the colours , and sbf magnitudes for that matter , should follow the fe abundance or the mg overabundance . the consistency between the mg@xmath5 and calibrations is illustrated in the lower panels of figure [ fig : mg2cal ] . we can tie these two quantities together via their respective calibrations , which for is based on @xmath1150 galaxies and for mg@xmath5 is based on one galaxy . the resulting ` empirical ' line , extended well beyond the true limits of applicability , almost perfectly overlays the model predictions . however , the agreement may be somewhat coincidental . a shift of only 0.02mag in the mg@xmath5 data for ngc3379 ( as , for instance , in the measurements by vazdekis et al . 1997 ) would eliminate the offset in the top panel of fig . [ fig : mg2cal ] and cause an offset in the lower panel . for a further comparison we show in the bottom panel galaxies in common between the sbf and smac ( smith et al . 2000 ; hudson et al . 2000 ) surveys . the smac mg@xmath5 values have been shifted by @xmath10 mag as an approximate allowance for aperture effects . the galaxy points follow a shallower slope than the models , perhaps because of the increasing mg enhancement for redder , more luminous ellipticals . figure [ fig : ircals ] shows the composite model relations between colour and the near - ir sbf magnitudes . the calibrations derived from the simple least - squares fits are @xmath11\ , , & \pm\,0.09\,\hbox{mag } \ , , \label{eq : jbar } \\ \overline{m}_h & = \;\ ; -4.92 \,+\ , 3.0\,[\vi - 1.15]\ , , & \pm\,0.10\,\hbox{mag } \ , , \label{eq : hbar } \\ \overline{m}_k & = \;\ ; -5.38 \,+\ , 2.9\,[\vi - 1.15]\ , , & \pm\,0.10\,\hbox{mag } \ , . \label{eq : kbar } \end{aligned}\ ] ] again , the slopes show variations at about the 12% level for different ( reasonable ) composite modeling assumptions , while the zero points remain fixed at the @xmath120.02mag level . the selection by fluctuation colours decreases the rms scatters about eqs.([eq : jbar])-([eq : kbar ] ) by @xmath170% in each case . thus , a measurement of the actual scatter in a large and diverse sample of early - type galaxies could provide insight into the variety of the stellar mixtures in these galaxies . our model @xmath6-band zero point at @xmath13 is 0.240.36 mag fainter than the empirical one ( depending on whether the group or direct cepheid tie is used ) , which is based on measurements by jensen et al . ( we should note that the @xmath6-band data are more weighted towards the red than the @xmath2-band survey data , so a better fiducial colour might be @xmath14 , in which case the zero point discrepancy is 0.30.42mag . interestingly , the former padua isochrones ( bertelli et al . 1994 ) , because of their more infrared luminous agb , make the model @xmath6 zero point brighter by 0.5mag or more , i.e. , about 0.2mag brighter than the empirical zero point . however , the agreement becomes much worse in the @xmath2 band , with the theoretical zero point being 0.40.5 mag too faint because of the cooler rgb . again , all the same caveats apply as in [ ssec : optcals ] with regard to the sensitivity of these calibrations to stellar evolution . clearly we will have learned much by the time the models reproduce the sbf data in all bands . it is noteworthy that the slopes of eqs.([eq : jbar])([eq : kbar ] ) are positive , i.e. , in the same sense as the slopes for optical sbf ( although the @xmath6-band slope is negative before the selection by fluctuation colours ) . this is in the sense opposite that given by either the worthey models or the liu et al . ( 2000 ) models , which both predict a brightening of the near - ir sbf magnitudes for redder populations . the dimming of our model near - ir sbf magnitudes in redder populations agrees better with observations ( jensen et al . 2000 , in preparation ) . in particular , we disagree with the conclusion by liu et al . ( 2000 ) that the @xmath6-band data imply that ellipticals all have roughly solar metallicity but a large spread in age . in the context of our models , the observed trend could well be due to metallicity variations . redder than @xmath14 , figure [ fig : ircals ] shows a flattening of the predicted relations , so that a constant calibration in the near - ir would work best for the reddest ellipticals . interestingly , just the opposite happens in the @xmath2-band , as we discuss in the following section . ajhar et al . ( 1997 ) measured sbf magnitudes in the wfpc2 f814w bandpass for 16 early - type galaxies observed with _ hst_. although f814w is similar to cousins @xmath2 , the data gave a steeper slope of @xmath15 for the calibration against . in an effort to understand this , we have computed sbf magnitudes for our models in the wfpc2 filter system using the empirical and synthetic photometric transformations from holtzman et al . unfortunately , the empirical transformations are only determined for stars with @xmath16 . for redder stars , including higher metallicity rgb stars ( the most important ones for us ) , they derived synthetic transformations using template stellar spectra and model bandpasses . it is this sort of theoretical transformation that we have sought to avoid in using the new vazdekis models , but here we are forced to it . as holtzman et al . discuss , their empirical and synthetic transformations do not always agree . however , we noticed that the two types of @xmath2f814w transformation intersect at @xmath17 . for consistency , we then shifted the @xmath0f555w and @xmath18f675w synthetic transformations by @xmath10.023mag into agreement with the empirical ones at @xmath17 . we used the empirical transformation for model stars bluer than this and the synthetic one for redder model stars . after transforming all the stars , we calculate the sbf magnitudes . figure [ fig : hstdiffs ] shows the differences between the standard and wfpc2 @xmath0 and @xmath2 sbf magnitudes as a function of , and compares these differences to the stellar transformations . even though these are precisely the transformations used for the model stars in this colour range , there are systematic ( though small ) differences with the final sbf transformations . this should be no surprise : the spectrum of the fluctuations is quite different from a stellar spectrum , but it is a point often casually overlooked . figure [ fig : f814cal ] shows the resulting relation between @xmath19 and for the culled sample of composite models . it has the same overall 4.5 slope as in the standard @xmath2 band , and is shifted fainter by 0.06mag ( but this result depends on the uncertain synthetic transforms ) . however , there is a steepening of the relation starting around @xmath14 . to illustrate , we compute the running average of @xmath20 in steps of 0.02 and plot the results in the figure . the ajhar et al . ( 1997 ) calibrating galaxies are distributed from @xmath13 to @xmath17 , with 75% being redder than 1.20 . this is because the wfpc2 images show only the red galaxy centres . the slope of the running averages at @xmath21 is 6.2 [ and changes by @xmath22 as the starting is changed by @xmath23 , in good agreement with the ajhar et al . result . thus , these new models for the first time resolve the puzzle of the steeper f814w slope : it boils down to a slightly nonlinear relation coupled with differences in sample selection . recently , sbf - iv introduced the parameter as the ratio , expressed in magnitudes , of the total apparent flux from a galaxy to the flux in the fluctuation signal . this ratio is independent of distance , photometric calibration , and extinction ; it corresponds to the total luminosity of a galaxy in units of a typical giant star within that galaxy : @xmath24 \ , , \label{eq : nbardef}\ ] ] where @xmath25 is the total apparent magnitude and @xmath26 the total luminosity . put another way , if all stars in a galaxy were of magnitude , then would be 2.5 times the logarithm of the total number of stars . sbf - iv therefore called the `` fluctuation star count , '' but it is known to the authors for convenience as the `` tonry number . '' sbf - iv showed that can be used as an alternative calibration in measuring sbf distances . the calibration they derived , adjusted to our group - based zero point ( [ ssec : vemp ] ) , is @xmath27 although this introduces covariance between and , the final distance modulus is actually 14% less sensitive to errors in . moreover , the necessity of measuring to an accuracy of 0.022 mag in order to limit the distance error from photometry to @xmath28mag is replaced by the much less stringent requirement of measuring @xmath25 to an accuracy of 0.7mag . similarly , the sensitivity to errors in the galactic extinction is cut nearly in half for the calibration . however , sbf - iv cautioned against dropping altogether in favour of . the calibration is based purely on stellar population properties , whereas the calibration essentially relies on the fundamental plane ( fp ) of elliptical galaxies and may be subject to systematic effects , such as environmental influences or the presence of a disk component . sbf - iv showed that has a tight correlation with stellar velocity dispersion for sbf survey ellipticals : @xmath29 where @xmath30 is in km s@xmath31 , and the scatter is 0.070 dex . the connection with the fp becomes transparent if we invert this relation to get @xmath32 eq.([eq : fabernbar ] ) superficially resembles the classical faber - jackson relation ( faber & jackson 1976 ) , but it contains an implicit correction for the mass - to - light ratio via the term , with its strong stellar - population dependence . this explains why the scatter is similar to that of the inverse fp ( e.g. , hudson et al . 1997 ) , and less than half that of faber - jackson . in order to model the behaviour of , we must assume some relation between the galaxy luminosity , or mass , and the stellar population . perhaps the most common form this takes is the mg@xmath5-@xmath30 relation ( dressler et al . 1987 ) , conventionally interpreted as a pure mass metallicity relation ( e.g. , bender , burstein , & faber 1993 ; guzmn , lucey , & bower 1993 ; but see also trager et al . however , we wish to avoid the use of mg@xmath5 here because it explicitly brings in the alpha - enhancement issue . instead , we simply derive the following relation between the colours of the sbf survey galaxies ( sbf - iv ) and the velocity dispersions of these galaxies from the homogenized smac sample ( hudson et al . 2000 ) : @xmath33 details of the comparison of the sbf and fp parameters will be given by blakeslee et al . ( 2000 , in preparation ) . next we assume that the dynamical mass of a galaxy is given by @xmath34 ( e.g. , bender et al . 1992 ; jrgensen 1999 ) and make the zeroth - order assumption that all ellipticals have effective radii @xmath35 kpc . finally , we assume that the initial stellar mass is half the dynamical mass and then use the model mass - to - light ratios to derive @xmath26 and . figure [ fig : nbar ] illustrates that our crude set of assumptions produces an - relation very similar to the observed one . the slope of the composite model relation is 0.18 , while the empirical slope is @xmath36 , and the scatter is 0.095 mag ( the slope and scatter would change to 0.20 and 0.074 mag , respectively , if we culled as in the previous section ) . the zero points of the model and empirical relations are also similar in figure [ fig : nbar ] , but this of course depends on the values assumed for @xmath37 and the fraction of the mass in stellar material . we have provided only a fairly schematic explanation , based on scaling laws , of why the distance calibration works . a true theoretical calibration of this method must come from further developments in the semi - analytic modeling of early - type galaxy formation ( e.g. , cole et al . 1994 ; baugh , cole , & frenk 1996 ; kauffmann & charlot 1998 ) . we have presented new @xmath0-band sbf measurements for five early - type galaxies in the fornax cluster . along with virgo and leo @xmath38 data from tonry et al . ( 1990 ) , these have been used to derive an empirical , cepheid - calibrated @xmath39- distance indicator . the slope of this relation is @xmath40 , similar to the @xmath41- empirical slope of 4.5 ( sbf - i ) . we find a mean fluctuation colour @xmath42 for our fairly homogeneous set of ellipticals , with a dispersion of only 0.15mag , comparable to the measurement error . new stellar population models , updated from those of vazdekis et al.(1996 ) with the isochrones of girardi et al . ( 2000 ) and a new set of empirical transformations , have been employed to calculate sbf magnitudes and integrated colours for six different metallicities and a wide range of ages . the age - metallicity degeneracy in the optical -colour relations is less for these models than for the models of worthey ( 1994 ) , although there is still a high degree of degeneracy in the relation against mg@xmath5 at intermediate and high metallicities . in the near - ir bands , the model sbf magnitudes have much shallower dependences on colour / metallicity . below solar metallicity , they depend solely on age , while the integrated colour depends almost solely on metallicity . thus , the vs. @xmath43 plot completely breaks the age - metallicity degeneracy for the lower metallicity models . this could be an important new avenue in stellar population studies . we have examined the sensitivity of our sbf predictions to changes in the isochrones , theoretical - to - observational transformations , and the imf . we find that using the former set of padua isochrones makes the sbf magnitudes 0.10.2mag fainter in the optical and about 0.5mag brighter in the near - ir . using theoretical transformations instead of our empirical ones can change the predictions by nearly 0.5mag and by up to 0.7mag ; the changes for the other bands are at the 0.10.2 mag level . thus , the isochrones and the transformations have about equal importance for the sbf results . on the other hand , different imfs yield fairly similar -colour relations in the optical because the imf effects are largely degenerate with age and metallicity , as noted by worthey ( 1993a ) . this is less true in the near - ir , where the steeper imf still makes the model colour redder ( like metallicity ) and fainter ( like age ) , but age and metallicity are no longer degenerate . the and magnitudes of our low metallicity models have been tested against observations for 19 globular clusters . the models agree fairly well , especially if younger ages are taken at higher metallicities , with the data in both zero point and metallicity dependence , which is very weak in the @xmath2-band . model calculations at lower metallicities and sbf measurements of higher metallicity globulars ( if possible ) would allow for much better tests . at any single metallicity , the models predict an anticorrelation between fluctuation colour and integrated colour : increasing the age makes bluer and redder , although they both become redder with increasing metallicity . the models further imply that the observed spread in at a fixed should be interpreted as a spread in age among the globulars . however , some of the globulars are too blue in ( and thus too old ) to be match models . this could in part be due to errors in the extinction estimates . at high metallicities , the models predict @xmath44 , in good agreement with the observations of elliptical galaxies . because the single - metallicity , single - age models do not yield completely monotonic relations between and colour , we adopt a monte carlo approach to making composite population models from three metallicity bins with different ( fairly loose ) restrictions on age . we then select out the models that reproduce the observed ( distance - independent ) fluctuation colours of early - type galaxies . this procedure reproduces well the colour range of elliptical galaxies and the empirically well - determined slope of against . if we reduce the mean number of metal poor stars in the models from 10% to 5% , the @xmath2-band sbf slope becomes marginally too steep , and steeper than the @xmath0-band slope , contrary to observations . this is due to the flattening of the - relation in bluer populations , seen in both the data and our models , and is a palpable indication of the presence of metal - poor stars in early - type galaxies . similarly , we find that the @xmath2 slope becomes steeper for redder composite populations , those with few metal - poor stars . this explains the steep slope found by ajhar et al . ( 1997 ) for _ hst _ f814w observations of elliptical galaxy centres , which are usually quite red . although the slopes of the model - relations are fairly sensitive to the details of the composite model prescription , the zero points at @xmath13 are not the zero point of the model calibration is only 0.03 mag brighter than the empirically determined zero point from [ sec : obs ] . if the ` direct ' ( but observationally more precarious ) tie of sbf to the cepheid distance scale via spiral bulges were adopted instead of the group tie , then the theoretical calibration would be 0.09 mag fainter than the empirical one . the theoretical zero point is 0.15 mag fainter than the empirical one ( or 0.27 mag for the direct tie ) . on the other hand , if the cepheid distance scale were revised to match the dynamically measured distance to the water maser in ngc4258 , then the theoretical and empirical zero points would agree closely ( at the expense of the empirically less secure calibration ) . our theoretical calibration is @xmath10.35mag fainter than the calibration given by the worthey models ; this difference appears to stem from the different isochrones used in the two sets of models . the variation of with mg@xmath5 for the models is in good accord with that obtained by thomsen et al . ( 1997 ) from the internal gradients of ngc3379 . again , the theoretical zero point is 0.15mag fainter than observed . this relation is predicted to have less scatter than the calibration against , but the distances would be much more challenging from an observational point of view , unless the sbf measurements are restricted to the bright galaxy centres , as for example with _ hst _ imaging . also , the models do not include the effects of mg overabundance on either the linestrength predictions or the isochrone temperatures , and it is unclear how this would affect the theoretical calibration . our composite modeling predicts that the slope of the near - ir -relation will have the same sense as in the optical , in contrast to the predictions of the worthey ( 1994 ) models and the liu et al . ( 2000 ) models . our predicted relation then becomes nearly flat for @xmath45 , ( just the opposite of the behaviour seen in the @xmath2 band ) . the slope of our derived relations depends on our procedure for selecting out composite models based on their fluctuation colours . the selection criteria also greatly reduce the scatter in the near - ir - relations . observational studies of the true slope and scatter , and further investigation into the range of fluctuation colours among elliptical galaxies , would be important from the standpoint of both stellar population and distance studies . the @xmath46 zero point given by our models is 0.24 mag fainter than the empirically calibrated one ( or 0.36 mag fainter with the direct tie ) . this discrepancy depends strongly on the temperature structure of the isochrones , including the agb . finally , we have used mass - colour scaling relations to investigate the recently - introduced tonry number , which describes the luminosity of a galaxy in units of a typical giant star . the simple scaling relations combined with the stellar population models can reproduce the observed dependence of on from sbf - iv . however , a true theoretical calibration of this form of the sbf method requires more physically - motivated evolutionary models . overall , our model sbf predictions are consistent with observations at the @xmath10.2mag level in the optical and at the @xmath10.3mag in the near - ir . the present uncertainties in the empirical calibrations are thought to be about a factor of two smaller than this . these results therefore present challenges for stellar evolution and population synthesis . we have noted that sbf magnitudes are acutely sensitive to the distribution and properties of evolved stars , particularly on the rgb , which in turn are sensitive to fairly ill - constrained things such as the colour - temperature scale for the coolest stars , the helium fraction , the prescription for convection , variations in elemental abundance ratios , diffusion , etc . however , it is precisely this sensitivity that gives sbf its own unique vista , as compared to more traditional observables , and should make multi - band sbf observations an important constraint for future stellar population . we are grateful to scott trager , guy worthey , michael liu , john lucey , and john tonry for helpful comments on earlier versions of the manuscript , and to harald kuntschner and tom shanks for fruitful discussions . we thank guy worthey for making his stellar population models accessible via the world wide web and the padua group for making their isochrones electronically available . we thank our sbf survey collaborators john tonry and alan dressler , as well john lucey and the smac team , for allowing us to make use of survey data in the process of publication . jpb and eaa thank john tonry for inspiring their interests in sbf . this work made use of starlink computer facilities and was supported at the university of durham by a pparc rolling grant in extragalactic astronomy and cosmology .	we investigate the use of surface brightness fluctuations ( sbf ) measurements in optical and near - ir bandpasses for both stellar population and distance studies . new @xmath0-band sbf data are reported for five galaxies in the fornax cluster and combined with literature data to define a @xmath0-band sbf distance indicator , calibrated against cepheid distances to the leo group and the virgo and fornax clusters . the colour dependence of the @xmath0-band sbf indicator is only @xmath115% steeper than that found for the @xmath2-band , and the mean ` fluctuation colour ' of the galaxies is @xmath3 . we use new stellar population models , based on the latest padua isochrones ( girardi et al . 2000 ) transformed empirically to the observational plane , to predict optical and near - ir sbf magnitudes and integrated colours for a wide range of population ages and metallicities . we examine the sensitivity of the predicted sbf colour relations to changes in the isochrones , stellar transformations , and initial mass function . the new models reproduce fairly well the weak dependence of @xmath0 and @xmath2 sbf in globular clusters on metallicity , especially if the more metal - rich globulars are younger . below solar metallicity , the near - ir sbf magnitudes depend mainly on age while the integrated colours depend mainly on metallicity . this could prove a powerful new approach to the age - metallicity degeneracy problem ; near - ir sbf observations of globular clusters would be an important test of the models . the models also help in understanding the and fluctuation colours of elliptical galaxies , with much less need for composite stellar populations than in previous models . however , in order to obtain theoretical calibrations of the sbf distance indicators , we combine the homogeneous population models into composite models and select out those ones with fluctuation colours consistent with observations . we are able to reproduce the observed range of elliptical galaxy colours , the slopes of the @xmath0 and @xmath2 sbf distance indicators against ( fainter sbf in redder populations ) , and the flattening of the @xmath2-band relation for @xmath4 . the models also match the observed slope of @xmath2-band sbf against the mg@xmath5 absorption index and explain the steep colour dependence found by ajhar et al . ( 1997 ) for the _ hst_/wfpc2 f814w - band sbf measurements . in contrast to previous models , ours predict that the near - ir sbf magnitudes will also continue to grow fainter for redder populations . the theoretical @xmath0-band sbf zero point predicted by these models agrees well with the cepheid - calibrated @xmath0-band empirical zero point . however , the model zero point is 0.150.27mag too faint in the @xmath2-band and 0.240.36mag too faint in @xmath6 . the zero points for the @xmath2 band ( empirically the best determined ) would come into close agreement if the cepheid distance scale were revised to agree with the recent dynamical distance measured to ngc4258 . we note that the theoretical sbf calibrations are sensitive to the uncertain details of stellar evolution and conclude that the empirical calibrations remain more secure . however , the sensitivity of sbf to these finer details potentially make it a powerful , relatively unexploited , constraint for stellar evolution and population synthesis . [ firstpage ] galaxies : distances and redshifts galaxies : elliptical and lenticular , cd galaxies : fundamental parameters galaxies : stellar content
let @xmath0 be an algebraic variety , defined by an ideal @xmath3}$ ] . the * amoeba * of @xmath1 , is defined to be the image of @xmath1 under the map @xmath4 defined at the point @xmath5 by @xmath6 we denote the amoeba of @xmath1 by either of @xmath7 or @xmath8 . if @xmath9 is a hypersurface , the zero locus of a single function @xmath10 , we will also use the notation @xmath11 . we refer the reader to mikhalkin s survey article @xcite for a broad discussion of amoebas and their applications . consider @xmath12}$ ] and a point @xmath13 . write @xmath10 as a sum of monomials @xmath14 . define @xmath15 to be the list of positive real numbers @xmath16 note that since the @xmath17 are monomials , this is well defined even though @xmath18 is not injective . [ def : lopsided ] we say that a list of positive numbers is * lopsided * , if one of the numbers is greater than the sum of all the others . equivalently , a list of numbers @xmath19 is not lopsided , if it is possible to choose complex phases @xmath20 ( @xmath21 ) , so that @xmath22 . this follows from the triangle inequality . we also define @xmath23 one can easily see that if @xmath24 then @xmath15 can not be lopsided ; in other words @xmath25 . indeed if @xmath26 then @xmath27 so it is giving a way to assign complex phases to the list @xmath28 such that the sum is @xmath29 . thus one should think of @xmath30 as a crude approximation to the amoeba @xmath11 . suppose @xmath31 , and let @xmath32 . for any complex phases @xmath33 , there exist @xmath34 such that @xmath35 and @xmath36 . thus @xmath24 if and only if @xmath37 is non - lopsided ; i.e. @xmath38 . in the above example we have enough freedom to choose the phases of the monomials @xmath39 for @xmath40 that @xmath41 . however , this only works because @xmath10 has very few non - zero terms . in general @xmath30 can be quite different from @xmath11 ( see figure [ fig : la ] ) . nevertheless , we will show that for a suitable multiple of @xmath10 , we can use this lopsidedness test to get very good approximations for @xmath11 . let @xmath42 be a positive integer . we consider the polynomials @xmath43 these @xmath44 are cyclic resultants @xmath45 and as such can be practically computed . our main result for amoebas of hypersurfaces is roughly the following . the precise version is stated and proved in section [ subsec : hypersurface ] . theorem [ thm : hypersurface ] ( rough version ) as @xmath46 , the family @xmath47 converges uniformly to @xmath11 . there exists an integer @xmath48 such that to compute @xmath11 to within @xmath49 , it suffices to compute @xmath47 for any @xmath50 . moreover , @xmath48 depends only on @xmath49 and the newton polytope ( or degree ) of @xmath10 , and can be computed explicitly from these data . this will lead us to the following characterisation of the amoeba of a general subvariety of @xmath51 . theorem [ thm : generalamoeba ] let @xmath3}$ ] be an ideal . a point @xmath13 is in the amoeba @xmath8 if and only if @xmath52 is not lopsided for every @xmath53 . phrased another way , if a point @xmath54 is outside the amoeba @xmath8 , a polynomial @xmath55 may witness this fact by being lopsided at @xmath54 . theorem [ thm : generalamoeba ] then states that there is always a witness . we will actually show something slightly stronger , in both theorems [ thm : hypersurface ] and [ thm : generalamoeba ] . we will show that there is a witness @xmath10 such that @xmath15 is `` superlopsided '' , according to the following definition : let @xmath56 . we say that a list of positive numbers @xmath57 is if there exists some @xmath58 such that @xmath59 for all @xmath60 . if @xmath61 , we will simply say the list is superlopsided . as before , we also define @xmath62 if a list of positive numbers is superlopsided , it is certainly lopsided ; hence @xmath63 ( see figure [ fig : la ] ) . david speyer observed that each component of the complement of @xmath64 is given by a system of _ linear _ inequalities , making it easier than @xmath30 to compute explicitly . hence theorem [ thm : hypersurface ] actually prescribes a method for approximating @xmath11 to within @xmath49 by systems of linear inequalities . similar ideas lead to a method for approximating the _ spine _ of a hypersurface amoeba . we discuss these constructions in section [ sec : approx ] . the motivation for these results comes from tropical algebraic geometry , and from this viewpoint , lopsidedness ( rather than superlopsidedness ) is the more natural condition to consider . in tropical algebraic geometry , we work with the semiring @xmath65 . this a semiring whose underlying set is @xmath66 , but whose operations are given by * @xmath67 * @xmath68 . the operations @xmath69 and @xmath70 are known as tropical addition and tropical multiplication . one can easily check that they satisfy the usual commutative , associative , and distributive laws ; however there are no additive inverses . a polynomial @xmath71 $ ] is therefore a piecewise linear function on @xmath72 : if @xmath73 then , translated into the usual operations on @xmath66 , @xmath74 the tropical variety associated to @xmath75 is then defined to be the singular locus of this piecewise linear function . a tropical variety associated to a single polynomial @xmath75 in this way is called a * tropical hypersurface. thus there is a simple nullstellensatz for tropical hypersurfaces . a point @xmath76 is outside the tropical variety of @xmath75 , if there is a single monomial term of @xmath75 which is strictly larger than each of the others when evaluated at the point @xmath76 . in terms of the tropical operations this term is strictly greater than the tropical sum of the other terms ( c.f . definition [ def : lopsided ] ) . more generally the principal results in this paper can be seen of as an analytic analogue of a theorem of speyer and sturmfels @xcite for tropical varieties of arbitrary codimension , also known as non - archimedian amoebas . we discuss this connection in section [ sec : tropical ] . i am deeply grateful to david speyer , for providing a number of useful insights , and in particular for drawing my attention to the existence of lemma [ lem : averaging ] . i would also like to thank allen knutson providing feedback on early drafts , matthias beck for answering my questions on ehrhart polynomials , and bernd sturmfels for being particularly supportive of this project and for suggesting the title . this research was partially supported by nserc . the idea of the one variable case is simple enough . suppose that @xmath78 , and for sake of argument , assume that the absolute values of the @xmath79 are all distinct , say @xmath80 . then @xmath81 for @xmath42 large , the terms @xmath82 are small in comparison with the other terms and so this is approximately @xmath83 suppose @xmath84 . consider @xmath85 : as @xmath46 , every term tends to @xmath29 except for the constant term which is @xmath86 . thus for @xmath42 large , there is a single term in @xmath87 and likewise in @xmath44 which is much bigger in absolute value than all the others . we now formalise this heuristic argument in a way that will be useful in proving theorem [ thm : hypersurface ] . at the crux of the heuristic argument are the following three key facts about @xmath44 : 1 . it has no roots inside a certain annulus ( in the heuristic argument the annulus is @xmath88 } ) . the only non - zero terms which appear are of the form @xmath89 . the number of terms is not too large . ( this approach fails if instead of @xmath90 , we try to use @xmath91 with @xmath92 . ) to get a result which we can apply to the multivariable case , we need to be able to make a uniform statement about polynomials with these properties . this is precisely captured by the next two lemmas . by applying lemma [ lem : onevariable ] directly to the family of functions @xmath90 one immediately obtains a complete proof of theorem [ thm : hypersurface ] in the @xmath77 case . [ lem : goodestimate ] let @xmath93 be an open annulus in @xmath94 . we allow the possibilities of @xmath95 , and @xmath96 . let @xmath97 be a real number with @xmath98 , and let @xmath99 denote the closed interval @xmath100 $ ] . suppose @xmath101 is a polynomial such that 1 . @xmath101 has no roots in @xmath102 , and 2 . the only terms which appear in @xmath101 are of the form @xmath103 , i.e. @xmath104 , where @xmath105 , for some positive integer @xmath42 . then there is some @xmath106 such that @xmath107 for all @xmath108 . a similar statement can be made for laurent polynomials : if @xmath101 is a laurent polynomial satisfying properties 1 and 2 , and @xmath109 , the conclusion remains valid . here @xmath110 and @xmath111 refer respectively to the largest and smallest exponents which appear in @xmath101 . the notation in our proof assumes that @xmath101 is actually a polynomial . we can write @xmath112 where @xmath113 . we adopt the convention that @xmath114 , and @xmath115 . since @xmath116 has no roots in @xmath102 we have @xmath117 for some @xmath106 ( @xmath118 ) . we show that this @xmath106 has the desired property . consider @xmath119 . write @xmath120 where @xmath121 is a monomial of degree @xmath122 . we now estimate the size of @xmath123 at some point @xmath124 . if @xmath125 , clearly holds , so we only need to consider @xmath126 . first , we will show that if @xmath126 , we have @xmath127 certainly for all @xmath126 we have @xmath128 we use the following facts : if @xmath129 , we have @xmath130 ( since @xmath131 ) . * if @xmath132 and @xmath133 , we have @xmath134 . * if @xmath135 , then @xmath136 . * if @xmath137 , then @xmath138 . if @xmath139 we write as @xmath140 this last equation comes from the fact that @xmath141 is the number of @xmath142 with @xmath143 . similarly , if @xmath144 we write @xmath145 which establishes . thus we obtain @xmath146 in particular for each @xmath126 we have latexmath:[\[\label{eqn : badestimate } however , we can do slightly better than this , by noting that the smallest power of @xmath97 that appears on the right side of the inequality is @xmath148 . thus , whereas tells us that @xmath149 , in fact we have : @xmath150 ( although is a better estimate than , the latter will prove to be more useful to us . ) for @xmath151 we have the following estimate : @xmath152 combining the estimates and we obtain @xmath153 for @xmath126 , as required . [ rmk : domterm ] it should be noted that we have actually determined which term is the special term @xmath154 . this is done in . if @xmath155 is a polynomial , then @xmath156 is the number of roots ( counted with multiplicity ) of @xmath155 inside the disc @xmath157 . if @xmath155 is a laurent polynomial , @xmath158 will be the number of roots inside @xmath159 . [ lem : onevariable ] as before , let @xmath93 . let @xmath97 be a real number with @xmath98 , and @xmath160 $ ] . fix positive integers @xmath161 . suppose @xmath116 is a family of polynomials such that 1 . @xmath116 has no roots in @xmath102 , 2 . the only terms which appear in @xmath116 are of the form @xmath89 . 3 . @xmath162 . more generally , if @xmath155 are laurent polynomials , we can ask that @xmath163 . assume @xmath42 is large enough that @xmath164 . then @xmath165 is @xmath166-superlopsided for all @xmath167 . from lemma [ lem : goodestimate ] , if we write @xmath168 , there is some @xmath106 such that for @xmath126 , @xmath169 here we are using the weaker estimate in to bound the numerator , and for the denominator . thus , @xmath170 the last step uses calculation [ cal:1 ] , which can be found in appendix [ app : calcdetails ] . this establishes that @xmath171 is @xmath166-superlopsided . this proof of lemma [ lem : onevariable ] uses a number of inequalities to give an answer to the question ` how large does @xmath42 have to be ? ' many of these inequalities , at first glance , appear not to be very tight . the answer we obtain is certainly not the best possible , and in most situations we should not expect to need quite such a large @xmath42 . however , without assuming more about @xmath155 it is surprisingly close to the best answer , particularly as @xmath172 . the best general answer for the question will be of the same form , i.e. @xmath173 one can see this by performing the requisite analysis on the polynomials @xmath174 . as a general heuristic , the more closely the roots of @xmath155 are packed , the larger @xmath42 will have to be ; thus the family of polynomials @xmath175 , where every root has as high a multiplicity as possible , is where we should expect our worst case behaviour to occur . suppose we want @xmath42 large enough to guarantee that @xmath176 is @xmath166-superlopsided for @xmath177 . write @xmath178 . we know @xmath86 will be the dominant term as @xmath42 gets large ( since @xmath179 ) , thus we need @xmath180 , or equivalently @xmath181 if we only want to guarantee that @xmath165 is lopsided for @xmath167 , we need @xmath42 large enough so that @xmath182 or equivalently @xmath183 this will hold if we have @xmath184 so @xmath42 only needs to be about half as big to guarantee that @xmath165 is lopsided as it does to guarantee that @xmath165 is superlopsided . again , we can show see that this is fairly close to the best answer by considering @xmath185 . in this section we prove our main theorem characterising the amoeba of a hypersurface . if @xmath186}$ ] , we consider the laurent polynomials @xmath187 theorem [ thm : hypersurface ] states that for @xmath188 , and a point @xmath13 in the complement of the amoeba @xmath11 whose distance from @xmath11 is at least @xmath49 , we can choose @xmath42 large enough so that @xmath189 is superlopsided . moreover , the theorem gives a upper bound on how large @xmath42 needs to be , based only on @xmath49 and the _ newton polytope _ of @xmath10 . the idea behind the proof of theorem [ thm : hypersurface ] is to look at the family of @xmath190 and interpret this as a function of a single variable @xmath191 . at the point @xmath192 , we define @xmath193 we will apply lemma [ lem : onevariable ] to these , and a find a single dominant term in this polynomial of one variable . then by an averaging argument , we show that this implies that @xmath44 has a single dominant term . first , however , we need a few simple observations . [ prop : sameamoeba ] @xmath194 @xmath190 is a product of terms @xmath195 , where @xmath196 . since @xmath197 , @xmath198 , and so @xmath199 . we will also need to know some information about the number and degree of the terms which appear in @xmath44 . first note the following important fact . [ prop : powersofn ] the only monomials which appear in @xmath44 are of the form @xmath200 . in particular the only terms appearing in @xmath201 are of the form @xmath89 . let @xmath202 denote the cyclic group of roots of @xmath203 . @xmath44 is manifestly invariant under the group action of @xmath204 acting on @xmath205}$ ] by @xmath206 . thus each monomial of @xmath44 must be invariant under this action . the only monomials with this property are of the form @xmath200 . the statement about @xmath201 follows immediately . recall that if @xmath207}$ ] its * newton polytope * , denoted @xmath208 , is the subset of @xmath72 defined as the convex hull of the exponent vectors of the monomials which appear in @xmath75 . for any polytope @xmath209 , let @xmath210 be any upper bound on @xmath211 . in general , it is not easy to find a tight upper bound for this number . if one can compute the ehrhart polynomial of @xmath209 explicitly , than an easy upper bound is the sum of the positive coefficients . otherwise , it is possible to bound the coefficients of the ehrhart polynomial in terms of the volume of @xmath209 ( see @xcite ) . using these estimates , for each @xmath212 one can compute constants @xmath102 and @xmath213 such that @xmath214 . clearly we have @xmath215 . this gives us an upper bound on the number of terms that @xmath44 can have : [ prop : newtonpolytope ] let @xmath216 . then @xmath44 has at most @xmath217 terms . by proposition [ prop : powersofn ] , the number of terms in @xmath44 is at most the number of integral points in @xmath218 . this is less than or equal to @xmath217 . finally we need to know something about @xmath219 . let @xmath220 where @xmath221 denotes the @xmath222 coordinate function on @xmath72 . [ prop : projnewtonpolytope ] @xmath223 @xmath224 armed with these facts and lemmas [ lem : goodestimate ] and [ lem : onevariable ] we are now in a position to precisely state and prove our main result for amoebas of hypersurfaces . [ thm : hypersurface ] let @xmath188 . suppose @xmath225 is a point in the amoeba complement whose distance from @xmath11 is at least @xmath49 . let @xmath216 , and @xmath226 . 1 . if @xmath42 is large enough so that @xmath227 then @xmath189 is lopsided . 2 . if @xmath42 is large enough so that @xmath228 then @xmath189 is superlopsided . ( in fact it is @xmath229-superlopsided ) . the key to reducing to the one variable case is the following basic result from complex analysis . [ lem : averaging ] let @xmath230 be a laurent polynomial , and write @xmath231 , where @xmath232 . suppose for all @xmath233 we have @xmath234 . then for each @xmath235 , @xmath236 . we integrate the equations @xmath237 over the set @xmath238 . @xmath239 let @xmath240 , and let @xmath241 . take @xmath233 . the function @xmath242 has no roots in the annulus @xmath243 . if it did , say at @xmath244 , @xmath44 would have a corresponding root at @xmath245 . so @xmath246 , hence by assumption , @xmath247 . but , since @xmath248 , @xmath249 . we ll first prove statement 2 . assume that holds . by propositions [ prop : powersofn ] and [ prop : projnewtonpolytope ] , we are in a position to apply lemma [ lem : onevariable ] to the laurent polynomial @xmath201 , taking @xmath102 , and @xmath97 as above , and @xmath250 , @xmath251 , @xmath252 and @xmath253 . with these choices , @xmath42 is large enough so that @xmath254 this is true for any @xmath233 , and since @xmath255 is connected , we see that for any fixed @xmath58 , it must be the same monomial term which dominates in each @xmath242 , independent of the choice of @xmath256 . let this be the @xmath257-term , and let @xmath258 . write @xmath259 where @xmath260 is the monomial @xmath261 . let @xmath262 . note this does not depend on the particular choice of @xmath256 . let @xmath263 we wish to show that @xmath264 . since the number of terms in @xmath44 is at most @xmath265 , this will imply that @xmath44 is superlopsided . we consider the terms of @xmath201 , and for simplicity of notation , we will temporarily assume that @xmath266 . the @xmath267-term in @xmath268 is @xmath269 for @xmath270 , tells us that @xmath271 since this is true for all @xmath256 , by lemma [ lem : averaging ] , @xmath272 provided @xmath270 . but , this argument is valid for any @xmath58 , so in fact we have @xmath273 for all @xmath274 . in particular , since @xmath275 for some @xmath274 , we deduce @xmath276 as required . we now prove statement 1 . although the approach is essentially the same , it is slightly more difficult , and hence requires some additional lemmas ( calculations [ cal : lopbounds ] and [ cal : expderiv ] , which can be found in appendix [ app : calcdetails ] ) . the reason for this is that we can not get these bounds by appealing directly to lemma [ lem : onevariable ] . instead , we use lemma [ lem : goodestimate ] , which gives better estimates for the coefficients of @xmath242 . write @xmath277 , where @xmath278 . then for each @xmath58 , @xmath279 . so by lemma [ lem : goodestimate ] there is some @xmath280 such that @xmath281 and in fact it is the same @xmath280 for all choices of @xmath256 . as before , let @xmath262 , and let @xmath282 . we have for any @xmath256 that @xmath283 ; thus @xmath284 this is true for all @xmath233 , thus by lemma [ lem : averaging ] we actually have @xmath285 the inequality is valid for all @xmath58 . in particular we should take @xmath58 to be the value which maximises @xmath286 , as this will be the one which minimises the right hand side . as @xmath287 is a sum of terms @xmath288 , can use to get an upper bound on @xmath289 . to do this , note that the right hand side of depends on @xmath290 and @xmath235 only insomuch as it depends on the maximum value of @xmath286 . thus we need to count , for each @xmath291 , how many values of @xmath235 in this sum satisfy @xmath292 an upper bound for this answer , is the number of integral points in @xmath293 with @xmath294 . there are @xmath295 of these , and this number is in turn less than @xmath296 . using @xmath296 as our upper bound for the number of @xmath235 with @xmath297 , gives us @xmath298 and by calculation [ cal : expderiv ] ( see appendix [ app : calcdetails ] ) this becomes @xmath299 assume now that holds . by calculation [ cal : lopbounds ] ( see appendix [ app : calcdetails ] ) , @xmath42 is large enough so that the right hand side is less than @xmath300 . thus we have @xmath301 , or equivalently , @xmath302 , as required . as an immediate consequence of theorem [ thm : hypersurface ] we get : [ cor : uniform ] both the families @xmath303 and @xmath304 converge uniformly to @xmath11 . we have @xmath305 , and for @xmath42 sufficiently large , the distance from any point in @xmath306 to @xmath11 will be less than @xmath307 , which tends to @xmath29 as @xmath46 . although theorem [ thm : hypersurface ] only showed that @xmath189 is _ superlopsided _ for @xmath42 sufficiently large . we can easily deduce something stronger . [ cor : verylopsided ] let @xmath308 . if @xmath42 sufficiently large , then @xmath189 is @xmath309-superlopsided . equivalently , if we write @xmath310 , then for @xmath42 sufficiently large , some @xmath311 will dominate , and in fact @xmath312 to see this , one simply has to look at . as @xmath46 , the right hand side approaches @xmath29 , and thus so does @xmath313 . alternatively , one can note that any @xmath314 can be used in place of @xmath315 in theorem [ thm : hypersurface ] . thus we see that it suffices to take @xmath42 so that @xmath316 just as was the case in lemma [ lem : onevariable ] , the bounds on @xmath42 given in theorem [ thm : hypersurface ] are not quite optimal : there are a number of places in which the inequalities can obviously be made tighter . however , as @xmath317 , the bounds are at least asymptotically correct . to see this for the superlopsided case , we can consider the example @xmath318 the amoeba @xmath11 is the union of all coordinate hyperplanes in @xmath72 . we can easily compute @xmath319 if our point is @xmath320 , with each @xmath321 , the dominant term of @xmath44 , will be @xmath86 . the number of non - zero monomial terms in @xmath44 is greater than @xmath322 ; thus to have @xmath323 superlopsided , we certainly need @xmath324 for all @xmath58 . this implies that @xmath325 in contrast , says that in this example we should take @xmath42 so that @xmath326 our bound for lopsidedness appears to be slightly less satisfactory . in the above example , to guarantee lopsidedness , one needs @xmath42 large enough so that @xmath327 this will hold when @xmath328 noting that @xmath329 , and approximating @xmath330 , this condition becomes @xmath331 if we take @xmath332 then this simplifies to @xmath333 in contrast , theorem [ thm : hypersurface ] tells us that it is sufficient to take @xmath42 so that @xmath334 again , this shows that the bounds in theorem [ thm : hypersurface ] are asymptotically correct , at least for any fixed @xmath212 . we suspect , however , that the correct general answer does not have this last term , or any term which is linear in @xmath212 . instead of the family @xmath44 , one may wish to consider a more general family of cyclic resultants . let @xmath335 be positive integers , and consider @xmath336 unfortunately it is not true that the family @xmath337 converges uniformly to @xmath11 , as @xmath338 . trouble occurs if some of the @xmath339 are significantly larger than others . for example , consider the amoeba of @xmath340 at a point @xmath341 , with @xmath342 . then @xmath343 if @xmath344 , then the first two terms above will have the same order of magnitude . thus @xmath345 will not be superlopsided , even if @xmath346 ( and hence @xmath347 ) are large . it will not be even be lopsided . however , if we restrict ourselves to the situation in which each @xmath339 is bounded by some polynomial in each of the other @xmath348 , then an analogous statement to theorem [ thm : hypersurface ] will be true . for example , we could let @xmath339 be any polynomial function of a single parameter @xmath42 . we will not compute explicit bounds for approximating the amoeba to within @xmath49 , in this more general situation ; however , the answer will depend upon these polynomials . it will certainly still be true that @xmath337 converges uniformly to @xmath11 , as this argument really only depends upon the fact that degrees of @xmath349 are growing only polynomially , while the terms are becoming suitably sparse . theorem [ thm : hypersurface ] tells us that for @xmath42 sufficiently large , one term of @xmath44 dominates , but does not specify which one . the answer will depend upon in which component of the amoeba complement our point @xmath54 lies . since @xmath350 varies continuously with @xmath54 , it depends only on the component of the amoeba complement . now the number of components is relatively small compared to the number of terms of @xmath44 . there is a natural injective map @xmath351 ( c.f . this is called the * index * of the component a complete definition is given below . so only a few of the terms of @xmath44 can possibly be dominant terms . fortunately it is relatively simple to determine which these are . the newton polytope of @xmath44 is @xmath352 , and candidates for dominant term are in fact what one would expect them to be : namely , they are the images of the integral points of @xmath353 under this scaling . let @xmath354 , and and let @xmath355 be the corresponding point in @xmath353 . if @xmath189 is lopsided , then the term of @xmath190 which dominates has exponent vector @xmath356 , ( i.e. it is the @xmath357-term ) . in order to make complete sense of the statement we need to know a definition of the index @xmath290 . there are a number of equivalent definitions , but the simplest for our purposes is the following . let @xmath233 . for each @xmath358 , consider the polynomial , @xmath359 if @xmath10 is a polynomial , then @xmath280 is the number of roots ( with multiplicity ) of @xmath360 inside the open disc @xmath361 . since @xmath360 never has a root on the circle @xmath362 for any @xmath256 , this number is independent of @xmath256 . if @xmath10 is a laurent polynomial , then @xmath363 it suffices to prove this if @xmath10 is a polynomial . if @xmath10 is a laurent polynomial , we multiply by some monomial to make it a polynomial , and note that this shifts both @xmath290 and the dominant term of @xmath44 appropriately . now @xmath364 is a product of @xmath365 terms , each of which is of the form @xmath366 , for some @xmath367 , @xmath368 . thus the number of roots of @xmath201 inside the disc @xmath369 is @xmath370 . as we noted in remark [ rmk : domterm ] , the number of roots in this disc is actually the exponent of the dominant term . since the dominant term of @xmath44 must project to the dominant term of each @xmath242 , the result follows . fix @xmath42 , and let @xmath372 where @xmath373 and @xmath315 are defined as in theorem [ thm : hypersurface ] . suppose we wish to approximate the component of @xmath371 corresponding to @xmath374 . from theorem [ thm : hypersurface ] , we know that @xmath375 and approximates @xmath11 to within @xmath49 . thus each component of @xmath371 is approximated by a component of @xmath376 . write @xmath377 where @xmath378 is the candidate for the dominant term in this component , and @xmath379 . are the other monomials . the corresponding component of @xmath376 is the set @xmath380 where @xmath381 is the number terms in @xmath44 . equivalently , this is the set of @xmath382 such that @xmath383 for all @xmath384 . this is a system of linear inequalities in the variable @xmath76 , so the solutions to these equations are a convex polyhedron which approximates the component of the amoeba to within @xmath49 . if there is no component of the @xmath371 corresponding to @xmath290 then this system of equations will have no solutions . conversely , if this system of inequalities has no solutions , then this component of the amoeba ( if it exists ) is not large enough to contain a ball of radius @xmath49 . thus we can realise any component of the @xmath371 as an increasing union of convex polyhedra . this gives an independent proof of the basic fact @xcite that the the components of the @xmath371 are convex . we must admit , however , that there are simpler proofs of this fact . note that in theorem [ thm : hypersurface ] , we actually show that @xmath44 is @xmath385-superlopsided . thus we can in fact take @xmath386 in and the set of solutions to this system of inequalities will still approximate the component of @xmath371 to within @xmath49 . one of the primary tools for studying amoebas has been the ronkin function @xmath387 , defined in @xcite . for @xmath12}$ ] , @xmath387 is defined to be the pushforward of @xmath388 under the map @xmath18 : @xmath389 ronkin shows in @xcite that @xmath387 is a convex function , and it is affine - linear precisely on the components of @xmath371 . when restricted to a single component of @xmath390 of @xmath371 , @xmath391 . passare and rullgrd use this function to define the spine of the amoeba @xcite , as follows . for each component @xmath392 of @xmath371 , extend the locally affine - linear function of @xmath393 to an affine - linear function @xmath394 on all of @xmath72 . let @xmath395 this is a convex piecewise linear function on @xmath72 , superscribing @xmath387 . the * spine * of the amoeba @xmath11 is defined to be the set of points where @xmath396 is not differentiable , and is denoted @xmath397 . the spine of the amoeba @xmath397 is a strong deformation retract of @xmath11 @xcite . also note that @xmath397 is actually a tropical hypersurface , as defined in the introduction : i.e. it is the singular locus of the maximum of a finite set of linear functions , where the gradient of each linear function is a lattice vector . now observe that @xmath398 can be thought of as a riemann sum for @xmath387 . in particular , we might expect @xmath399 to converge pointwise to @xmath400 . this will certainly be true provided that @xmath401 , is bounded on @xmath402 , which is the case when @xmath403 . suppose @xmath404 is in the component of @xmath371 of index @xmath374 . assume that @xmath76 has distance at least @xmath405 from the amoeba , where @xmath406 is fixed . for any @xmath188 we can find @xmath42 sufficiently large so that @xmath407 where each @xmath408 is relatively small , i.e. @xmath409 ( see corollary [ cor : verylopsided ] ) . thus we have @xmath410 thus we see that as @xmath411 , the values of @xmath399 , @xmath412 and @xmath413 all converge upon @xmath414 . we can use this fact obtain good approximations for the spine of the amoeba . for each @xmath42 we consider the function @xmath415 given by @xmath416 where the maximum is taken over all components of @xmath371 . this is a piecewise linear function . we define the approximate spine of the amoeba @xmath417 to be the set of points where @xmath418 is not smooth . equivalently @xmath417 is the set of points where the maximum in equation is attained by two distinct values of @xmath290 . 1 . @xmath419 . 2 . @xmath420 . statement 1 is true because on the component of @xmath421 of index @xmath290 , @xmath422 for any other @xmath423 . thus the maximum value in equation can not be attained by two distinct @xmath290 if @xmath424 . statement 2 follows from the fact that @xmath425 is a constant function and is less than @xmath49 for @xmath42 large . let @xmath426 and @xmath427 be components of @xmath371 of index @xmath428 and @xmath429 respectively . consider the hyperplane @xmath430 where @xmath431 and @xmath432 coincide , and the hyperplane @xmath433 where @xmath434 and @xmath435 coincide . the two hyperplanes @xmath436 and @xmath433 are parallel , and their distance apart is most @xmath437 , where @xmath99 is some constant depending only on @xmath428 and @xmath429 . as there are only finitely many @xmath438 , these distances can be made uniformly small . for practical reasons , we may wish use an alternate definition of @xmath439 , in which one takes the maximum in equation over only those components which appear in @xmath376 . if we do , statement 1 may be false for small @xmath42 , and we must settle for saying that @xmath440 . one might hope to be able to simplify this construction by taking the maximum in equation over all @xmath441 , rather than just those which actually correspond to components . it appears , however , that this does not give the same answer . with this alternate definition of @xmath439 , the ` approximate spine ' will have false chambers for all @xmath42 : i.e. the complement of this approximate spine will have components which do not correspond to components of @xmath371 . we might still hope that these false chambers would shrink to zero volume as @xmath42 gets large . unfortunately , experimental evidence suggests that the limit of these false chambers , as @xmath46 , can sometimes contain a ball of positive radius , and so this method does not produce a good approximation of the spine . the higher codimension statement ( theorem [ thm : generalamoeba ] ) follows fairly quickly from the hypersurface statement . let @xmath442 be a variety , which is the zero locus of an ideal @xmath443}$ ] . [ prop : reduction ] for every @xmath13 , there exists @xmath444 , such that @xmath445 for any laurent polynomial @xmath446},\ ] ] let @xmath447 denote its complex conjugate @xmath448 we define @xmath449 to be @xmath450 clearly @xmath449 is a laurent polynomial and is in @xmath451 . moreover , if we restrict @xmath452 to @xmath255 , then @xmath453 , so @xmath454 thus @xmath455 if and only if @xmath456 for all @xmath58 . this result is also true for ideals in @xmath457 $ ] : one can find a suitable monomial @xmath458 such that @xmath459 is a polynomial for all @xmath58 , and a similar argument holds if @xmath460 as an immediate consequence of proposition [ prop : reduction ] we have the following . [ cor : intersection ] for any ideal @xmath3}$ ] , @xmath461 it is also now a simple task to prove our second main result . [ thm : generalamoeba ] let @xmath3}$ ] be an ideal . a point @xmath13 is in the amoeba @xmath8 if and only if @xmath52 is not ( super)lopsided for every @xmath53 . if @xmath462 , then @xmath15 can not be lopsided for any @xmath55 , since @xmath24 for every @xmath55 . on the other hand , suppose @xmath463 . then by proposition [ prop : reduction ] if we take @xmath464 , then @xmath465 . by theorem [ thm : hypersurface ] , if @xmath42 is sufficiently large then @xmath466 is ( super)lopsided , and @xmath467 . [ rmk : summary ] in summary , we have three coincident sets , any of which can be used to define the amoeba @xmath7 of a variety @xmath468 . 1 . @xmath469 . 2 . @xmath470 . @xmath471 is not lopsided , for all @xmath472 . in section [ sec : tropical ] we shall see that this is precisely analogous to a theorem for tropical algebraic varieties . if a point @xmath54 is in @xmath473 , the proof of theorem [ thm : generalamoeba ] also tells us where to look for a witness to this fact : namely , we should look at @xmath474 for all @xmath42 . for some sufficiently large @xmath42 , this list will be lopsided . one unfortunate misfeature of this proof is that it requires us to use a different @xmath75 for every point @xmath475 . thus this statement is purely local . it does not give any clues as to how to produce a global uniform approximation to @xmath8 . however , in general we can not expect there to be any finite set of elements @xmath476 such that if @xmath463 , then some @xmath477 is lopsided for @xmath42 sufficiently large . if it were so , this would imply that @xmath8 is always an intersection of finitely many hypersurface amoebas , and this is certainly not true for dimensional reasons if @xmath478 . the best one could for is that we could find some family of sets @xmath479 such that @xmath480 approximates @xmath8 uniformly to within @xmath49 for @xmath42 sufficiently large . we present it as an open problem to find such a family explicitly . if we endow @xmath51 with the symplectic form @xmath481 , then the action of @xmath482 is hamiltonian , and its moment map is @xmath18 . suppose @xmath483 is another torus with a homomorphism @xmath484 . then the induced action of @xmath483 on @xmath51 is also hamiltonian , and we denote its moment map by @xmath485 . in this section we describe how to compute the image of @xmath1 under the map @xmath485 . the image @xmath486 is always a linear transformation of @xmath7 . if @xmath487 denotes the adjoint map to @xmath488 , then @xmath489 ( see e.g. ( * ? ? ? 3.2.8 ) ) . more concretely if we identify @xmath483 with @xmath490 , we can write @xmath491 where @xmath492 are the integer entries of a matrix @xmath102 the matrix representation of @xmath493and @xmath494 we could also take a matrix @xmath102 with integer entries as our starting point , and construct @xmath483 , @xmath488 , and the map @xmath485 so that holds . let @xmath3}$ ] denote the ideal of v. let @xmath495 denote the variety of the ideal @xmath496},\ ] ] where @xmath497 . now consider the projection of @xmath498 onto the @xmath499-coordinates @xmath500 . the image of @xmath498 under this projection is a variety @xmath501 . standard techniques of elimination theory allows us to compute its ideal @xmath502 ( see e.g. @xcite ) . [ prop : lintransform ] @xmath503 . a point in @xmath498 is a simply pair @xmath504 where @xmath505 and @xmath506 . thus we have @xmath507 projecting onto the @xmath499-coordinates , we obtain @xmath508 it is interesting to note that this construction is closely related to the cyclic resultants used in the proof of theorem [ thm : hypersurface ] . suppose @xmath509 is a hypersurface , and @xmath510 is the map @xmath511 . in this case , @xmath512 is a multiple of the identity matrix , and the variety @xmath501 is the zero locus of the function @xmath44 . intuitively , we should of think that the linear transformation is zooming in on the amoeba @xmath513 ; as we zoom in , theorem [ thm : hypersurface ] tells us that we see more and more detail in the approximations @xmath514 and @xmath515 . the most natural generalisation of amoebas in the compact setting , is to subvarieties of projective toric varieties . each projective toric variety is a compactification of @xmath51 , with an @xmath516 action which extends the @xmath516 action on @xmath51 . it also carries a natural symplectic form @xmath517 , for which the @xmath516 action is hamiltonian . we may therefore use the moment map for this hamiltonian action to replace the map @xmath18 . our goal in this section is to give a concrete description of this more general setting , and observe that our results still hold . this follows fairly easily from the non - compact case . our construction of toric varieties and their moment maps roughly follows a combination of @xcite and @xcite . let @xmath518 be a lattice polytope , i.e. the vertices of @xmath209 have integral coordinates . to every such @xmath209 we can associate the following data : 1 . a set of lattice points @xmath519 . 2 . a semigroup ring @xmath520 if @xmath521 , this is defined to be the quotient ring @xmath522/j,\ ] ] where each @xmath523 has degree @xmath86 , and @xmath524 is generated by all ( homogeneous ) relations of the form @xmath525 note that @xmath520 $ ] carries an action of the complex torus @xmath526 , given by @xmath527 3 . a toric variety @xmath528)$ ] . 4 . a projective embedding @xmath529)$ ] , induced by the map on rings @xmath530 \to { \mathbb{c}}[a]$ ] given by @xmath531 . 5 . a symplectic form @xmath532 , where @xmath533 is the fubini - study symplectic form on @xmath534 . 6 . a moment map @xmath535 for the @xmath516 action on @xmath536 . we can in fact write down the moment map @xmath535 explicitly . @xmath537 to evaluate the right hand side , we must choose a lifting of @xmath538 to @xmath539)$ ] . however , since this expression is homogeneous of degree @xmath29 in the @xmath540 , it is in fact well defined . it is well known that @xmath541 , and that if @xmath542 is any other projective toric variety with @xmath543 , then @xmath544 as toric varieties . let @xmath545 $ ] be a homogeneous ideal , and @xmath546/i)$ ] its variety inside @xmath547 . the * compactified amoeba * of @xmath1 is @xmath548 . we denote the compactified amoeba of @xmath1 by either @xmath549 or @xmath550 ( or by @xmath551 if @xmath552 is principal ) . let @xmath553 $ ] be a homogeneous polynomial of degree @xmath499 . we can again decompose @xmath10 as a sum of monomials , i.e. write @xmath554 , where each @xmath17 is a @xmath555-weight vector in @xmath520 $ ] . each of these @xmath17 is a well defined function on @xmath556 . let @xmath557 . we define @xmath558 is any preimage of @xmath559 in the composite map @xmath560 . of course , @xmath561 will depend on the choice of lifting under @xmath562 but only up to rescaling . thus the notions of @xmath561 being lopsided or superlopsided are still well defined . we define @xmath563 and @xmath564 to the set of points @xmath557 such that @xmath561 is non - lopsided , and non - superlopsided respectively . let @xmath565 denote the intersection of @xmath1 with the open dense subset of @xmath547 on which @xmath555 acts freely . we can identify this open dense subset of @xmath51 , and therefore consider @xmath566 . as both @xmath18 and @xmath567 are both submersions with fibres @xmath516 , it follows that @xmath566 is diffeomorphic to @xmath568 , where @xmath569 denotes the interior of @xmath209 . let @xmath570 denote this diffeomorphism . moreover any face @xmath571 of @xmath209 , corresponds to a toric subvariety @xmath572 . and @xmath573 @xcite . thus for every point @xmath557 , we can determine whether @xmath559 is in the compactified amoeba @xmath549 as follows . first we determine the face @xmath574 for which @xmath575 . then @xmath576 identifies @xmath577 with @xmath578 in such a way that @xmath579 is identified with @xmath580 . we then have @xmath581 if and only if @xmath582 . [ lem : uniformtransform ] the map @xmath583 is uniformly continuous . the projective embedding @xmath584 induces a map @xmath585 , which is define by monomials . this induces a linear map from @xmath586 . we also have a map from the moment polytope @xmath587 of @xmath588 to @xmath209 which is the projection induced by the inclusion of tori @xmath589 . the composite @xmath590 is @xmath583 . since the first and last maps are uniformly continuous , it suffices to show that @xmath591 is uniformly continuous . this is fairly straightforward . write @xmath592 . for @xmath593 we may write @xmath594 , and @xmath595 . if @xmath596 then @xmath597 i.e. @xmath598 . so @xmath599 . and the result follows . it should therefore come as no surprise that theorems [ thm : hypersurface ] and [ thm : generalamoeba ] have analogues in the compact setting . [ cor : compacthypersurface ] the families @xmath600 and @xmath601 converge uniformly to @xmath551 . moreover , to approximate the @xmath551 , to within @xmath49 , one can to choose @xmath602 , where @xmath603 can be determined solely from @xmath209 and @xmath604 . one can easily verify that @xmath605 is ( super)lopsided if and only if @xmath606 is ( super)lopsided , where @xmath575 . thus , since the lopsidedness of @xmath607 can be used to test membership in @xmath608 , @xmath609 can be used to determine membership in @xmath551 . the fact that the convergence is uniform follows from lemma [ lem : uniformtransform ] . moreover , as the data @xmath209 and @xmath604 , are enough to determine an outer bound on the newton polytope of @xmath10 , the rate of convergence can be determined from these data . we also immediately have : [ cor : compactamoeba ] @xmath610 . in particular @xmath611 it should be noted that corollary [ cor : compactamoeba ] holds for all toric varieties with a moment map , not just the compact ones . however the statement of uniform convergence in corollary [ cor : compacthypersurface ] does not hold in general for non - compact toric varieties . for example , if one considers the toric variety @xmath612 , with the standard moment map @xmath613 , the convergence of the family @xmath600 will almost never be uniform . one can even see this in the simple example @xmath614 . the failure is that lemma [ lem : uniformtransform ] does not hold : the map @xmath615 is not uniformly continuous , and so the uniform convergence does not carry over . it is most unfortunate that proposition [ prop : lintransform ] does not easily carry over to the compact case . the use of elimination theory appears only to be well suited to the study of @xmath51 with its particular standard symplectic form . in this section we show that theorem [ thm : generalamoeba ] is the analytic counterpart to a theorem for tropical varieties . we have already seen examples of tropical hypersurfaces . tropical varieties in general , can be thought of as a generalisation of amoebas , where one replaces the norm @xmath616 by a valuation in some non - archimedian field . for this reason , tropical varieties are also known as non - archimedian amoebas . let @xmath99 be an algebraically closed field , with valuation @xmath617 . for our purposes , a valuation on @xmath99 is a map @xmath618 , which satisfies the following conditions : * @xmath619 * @xmath620 . this differs from the usual definition of a valuation in two purely cosmetic ways . first , a valuation is traditionally given as a map to @xmath621 ; we have simply translated into the operations of @xmath622 . second , this is @xmath623 times the usual notion of a valuation . our reasons for making these cosmetic changes will become abundantly clear by the end of this section . to every @xmath624}$ ] we can associate a tropical polynomial as follows . if @xmath625 , write @xmath626 and call it the * tropicalisation * of @xmath10 . we denote the tropical hypersurface associated to @xmath627 by @xmath628 . if @xmath629 , we can assign a weight to every monomial @xmath630}$ ] : define the weight of @xmath631 at @xmath54 to be @xmath632 if @xmath633 where @xmath17 are monomials , let @xmath634 recall that in @xmath622 , a list of numbers @xmath635 is ( tropically ) lopsided if the maximum of element of this list does not occur twice ( in which case the maximum element is greater than the tropical sum of all the other elements ) . thus @xmath636 is lopsided if and only if @xmath637 . let @xmath3}$ ] be an ideal , and @xmath638 be its affine variety . let @xmath639 be the map @xmath640 the following theorem , as stated , most closely resembles the formulation in @xcite , though variants of it have also appeared in @xcite . [ thm : tropical ] the following subsets of @xmath641 coincide : 1 . the closure of the set @xmath642 . the intersection of all tropical hypersurfaces @xmath643 3 . the set of points @xmath644 , such that @xmath636 is not lopsided for all @xmath55 . this set is called the * tropical variety * of the ideal @xmath451 . in fact a stronger result than theorem [ thm : tropical ] ( as stated here ) is shown in @xcite . let @xmath106 denote the residue field of @xmath99 . if @xmath645 $ ] , then one can construct an initial ideal of i , in @xmath646 $ ] , corresponding to any weight @xmath647 . one can equivalently describe the tropical variety of @xmath451 as the set of points @xmath647 such that the associated initial ideal contains no monomial . thus it is possible to determine membership in a tropical variety using grbner basis techniques . one can easily see that theorem [ thm : tropical ] is precisely analogous to the summary given in remark [ rmk : summary ] . the proofs of these results , however , are extremely different . an obvious question , therefore , is whether analogous statements can be made in other contexts . the following is a general context in which one might hope for such a theorem to be true . suppose that @xmath99 is an algebraically closed field , and let @xmath648 be a totally ordered semiring . suppose @xmath649 satisfies the following conditions : 1 . @xmath650 for all @xmath651 . 2 . for all @xmath652 we have @xmath653 in particular condition 2 implies that @xmath654 for all @xmath655 . thus @xmath656 is an @xmath657-valued norm . let @xmath658}$ ] , and write @xmath659 as a sum of monomials . for any point @xmath660 , let @xmath256 be such that @xmath661 . we define @xmath662 as @xmath656 is multiplicative , this is independent of the choice of @xmath256 . since @xmath657 is totally ordered , we can define a list to of elements of @xmath657 to be _ lopsided _ if and only if one number is greater than the sum of all the others . let @xmath663 be a variety defined by an ideal @xmath664}$ ] . we can consider the following sets : * the closure of @xmath665 * @xmath666 is not lopsided @xmath667 the question is whether these two sets are equal for a particular @xmath668 . in this paper , we primarily discussed the example in which @xmath669 , @xmath670 , and @xmath671 , and showed that they are equal . we have also just seen that this is true if @xmath99 is a non - archimedian field , with @xmath656 as its valuation and @xmath672 . many ( though not quite all ) of the elements of the proof of theorem [ thm : hypersurface ] are valid in a more general context . suppose that in addition to being a totally ordered semiring , @xmath657 is a @xmath673-module ( i.e. we can make sense of such things as @xmath674 for @xmath675 ) ; for example @xmath622 is a @xmath673-module with the trivial @xmath673-action . define a binary operation @xmath676 on @xmath657 by @xmath677 whenever this set is nonempty . ( we need not overly concern ourselves with the fact that a precise minimum may not exist : one can always get around this by treating this set as a dedekind cut . ) then the triangle inequality @xmath678 is valid ( assuming @xmath679 ) . to see this note that @xmath680 implies @xmath681 ; thus @xmath682 clearly @xmath683 which implies that @xmath684 a closer examination of the proofs of lemma [ lem : goodestimate ] and calculation [ cal : expderiv ] now reveal that they are also valid ( almost word for word ) for a general @xmath685 . we can also prove lemma [ lem : averaging ] in general , by replacing the integral over the torus @xmath686 by a discrete average over a finite subgroup of the torus @xmath687 if @xmath48 is suitably large this discrete average will have the same effect as the integral ( i.e. picking out a single term from the polynomial ) . in fact we can follow the proof statement 1 of theorem [ thm : hypersurface ] , up to and including the inequality . all that remains to show that the right hand side of becomes sufficiently small as @xmath42 gets large . unfortunately , in general , this is not always true for all @xmath688 ( e.g. if @xmath689 $ ] with @xmath690 if the leading coefficient of @xmath691 is positive , this fails for @xmath692 ) . it is true , however , for @xmath693 , so our proof of theorem [ thm : hypersurface ] actually works in the tropical setting as well . this suggests that it may be possible to prove a more general statement , with perhaps some additional restrictions on @xmath657 . on the other hand , our proof of corollary [ cor : intersection ] , which is the key to theorem [ thm : generalamoeba ] , is very highly dependent on the fact that our underlying field is the complex numbers . a very different proof is required in the tropical setting , and it would be interesting to try and unite these two . in this appendix we give the mundane , technical calculations which are used in the proofs of lemma [ lem : onevariable ] and theorem [ thm : hypersurface ] .	the amoeba of an affine algebraic variety @xmath0 is the image of @xmath1 under the map @xmath2 . we give a characterisation of the amoeba based on the triangle inequality , which we call ` testing for _ lopsidedness _ ' . we show that if a point is outside the amoeba of @xmath1 , there is an element of the defining ideal which witnesses this fact by being lopsided . this condition is necessary and sufficient for amoebas of arbitrary codimension , as well as for compactifications of amoebas inside any toric variety . our approach naturally leads to methods for approximating hypersurface amoebas and their spines by systems of linear inequalities . finally , we remark that our main result can be seen a precise analogue of a nullstellensatz statement for tropical varieties .
this paper is motivated by the study of the nonhomogeneous linear recursion @xmath6 where @xmath2 is a nonnegative random vector with @xmath7 , @xmath8 , @xmath9 , and @xmath10 is a sequence of iid random variables , independent of @xmath2 , having the same distribution as @xmath0 . this recursion appeared recently in the stochastic analysis of google s pagerank algorithm , see @xcite and the references therein for the latest work in the area . these types of weighted recursions , also studied in the literature on weighted branching processes @xcite and branching random walks @xcite , are found in the probabilistic analysis of other algorithms as well @xcite , e.g. , quicksort algorithm @xcite . in order to study the preceding recursion in its full generality we extend the implicit renewal theory of goldie @xcite to cover recursions on trees . the extension of goldie s theorem is presented in theorem [ t.newgoldie ] of section [ s.renewal ] . one of the observations that allows this extension is that an appropriately constructed measure on a weighted branching tree is a renewal measure , see lemma [ l.renewalmeasure ] and equation . in the remainder of the paper we apply the newly developed framework to analyze a number of linear and non - linear stochastic recursions on trees , starting with . note that the majority of the work in the rest of the paper goes into the application of the main theorem to specific problems . in this regard , in section [ s.linearrec ] , we first construct an explicit solution to on a weighted branching tree and then provide sufficient conditions for the finiteness of moments and the uniqueness of this solution in lemmas [ l.moments_r ] and [ l.convergence ] , respectively . furthermore , it is worth noting that our moment estimates are explicit , see lemma [ l.generalmoment ] , which may be of independent interest . then , the main result , which characterizes the power - tail behavior of @xmath0 is presented in theorem [ t.linearrecursion ] . in addition , for integer power exponent ( @xmath11 ) the asymptotic tail behavior can be explicitly computed as stated in corollary [ c.explicit ] . furthermore , for non integer @xmath12 , lemma [ l.alpha_moments ] yields an explicit bound on the tail behavior of @xmath0 . related work in the literature of weighted branching processes ( wbps ) for the case when @xmath13 and @xmath14 are nonnegative deterministic constants can be found in @xcite ( see theorem 5 ) , and more recently , for real valued constants , in @xcite . however , these deterministic assumptions fall outside of the scope of this paper ; for more details see the remarks after theorem [ t.linearrecursion ] in section [ ss.mainlinear ] . next , we show how our technique can be applied to study the tail asymptotics of the solution to the critical , @xmath15 = 1 $ ] , homogeneous linear equation @xmath16 where @xmath17 is a nonnegative random vector with @xmath7 and @xmath18 is a sequence of iid random variables independent of @xmath17 having the same distribution as @xmath0 . this type of recursion has been studied to a great extent under a variety of names , including branching random walks and multiplicative cascades . our work is more closely related to the results of @xcite and @xcite , where the conditions for power - tail asymptotics of the distribution of @xmath0 with power exponent @xmath19 were derived . in theorem [ t.linearhomog ] of section [ ss.mainlinear ] we provide an alternative derivation of theorem 2.2 in @xcite and proposition 7 in @xcite . furthermore , we note that our method yields a more explicit characterization of the power - tail proportionality constant , see corollary [ c.explicithom ] . for the full description of the set of solutions to see the very recent work in @xcite . for additional references on weighted branching processes and multiplicative cascades see @xcite and the references therein . for earlier historical references see @xcite . as an additional illustration of the newly developed framework , in section [ s.maxrec ] we study the recursion @xmath20 where @xmath2 is a nonnegative random vector with @xmath7 , @xmath21 and @xmath4 is a sequence of iid random variables independent of @xmath2 having the same distribution as @xmath0 . we characterize the tail behavior of @xmath22 in theorem [ t.maximumrecursion ] . similarly to the homogeneous linear case , this recursion was previously studied in @xcite under the assumption that @xmath23 , @xmath13 , and the @xmath24 are real valued deterministic constants . the more closely related case of @xmath23 and @xmath25 being random was studied earlier in @xcite . furthermore , these max - type stochastic recursions appear in a wide variety of applications , ranging from the average case analysis of algorithms to statistical physics ; see @xcite for a recent survey . we conclude the paper with a brief discussion of other non - linear recursions that could be studied using the developed techniques , including the solution to @xmath26 the majority of the proofs are postponed to section [ s.proofs ] . first we construct a random tree @xmath27 . we use the notation @xmath28 to denote the root node of @xmath27 , and @xmath29 , @xmath30 , to denote the set of all individuals in the @xmath31th generation of @xmath27 , @xmath32 . let @xmath33 be the number of individuals in the @xmath31th generation , that is , @xmath34 , where @xmath35 denotes the cardinality of a set ; in particular , @xmath36 . next , let @xmath37 be the set of positive integers and let @xmath38 be the set of all finite sequences @xmath39 , where by convention @xmath40 contains the null sequence @xmath28 . to ease the exposition , for a sequence @xmath41 we write @xmath42 , provided @xmath43 , and @xmath44 to denote the index truncation at level @xmath31 , @xmath30 . also , for @xmath45 we simply use the notation @xmath46 , that is , without the parenthesis . similarly , for @xmath47 we will use @xmath48 to denote the index concatenation operation , if @xmath49 , then @xmath50 . we iteratively construct the tree as follows . let @xmath51 be the number of individuals born to the root node @xmath28 , @xmath52 , and let @xmath53 be iid copies of @xmath51 . define now @xmath54 it follows that the number of individuals @xmath34 in the @xmath31th generation , @xmath55 , satisfies the branching recursion @xmath56 now , we construct the weighted branching tree @xmath57 as follows . the root node @xmath28 is assigned a vector @xmath58 with @xmath7 and @xmath59 ; @xmath51 determines the number of nodes in the first generation of @xmath27 according to . each node in the first generation is then assigned an iid copy @xmath60 of the root vector and the @xmath61 are used to define the second generation in @xmath27 according to . in general , for @xmath62 , to each node @xmath63 , we assign an iid copy @xmath64 of the root vector and construct @xmath65 ; the vectors @xmath66 , @xmath63 are chosen independently of all the previously assigned vectors @xmath67 , @xmath68 . for each node in @xmath57 we also define the weight @xmath69 via the recursion @xmath70 where @xmath71 is the weight of the root node . note that the weight @xmath72 is equal to the product of all the weights @xmath73 along the branch leading to node @xmath74 , as depicted in figure [ f.tree ] . in some places , e.g. in the following section , the value of @xmath75 may be of no importance , and thus we will consider a weighted branching tree defined by the smaller vector @xmath17 . this tree can be obtained form @xmath57 by simply disregarding the values for @xmath76 and is denoted by @xmath77 . ( 430,160)(0,0 ) ( 0,0 ) ( 125,150)@xmath78 ( 69,83)@xmath79 ( 131,83)@xmath80 ( 219,83)@xmath81 ( 22,17)@xmath82 ( 78,17)@xmath83 ( 126,17)@xmath84 ( 162,17)@xmath85 ( 213,17)@xmath86 ( 268,17)@xmath87 ( 350,150)@xmath36 ( 350,83)@xmath88 ( 350,17)@xmath89 studying the tail behavior of the solutions to recursions and fixed point equations embedded in this weighted branching tree is the objective of this paper . in this section we present an extension of goldie s implicit renewal theorem @xcite to weighted branching trees . the observation that facilitates this generalization is the following lemma which shows that a certain measure on a tree is actually a product measure ; a similar measure was used in a different context in @xcite . its proof is given in section [ ss.implicitproofs ] for completeness . throughout the paper we use the standard convention @xmath90 for all @xmath91 . [ l.renewalmeasure ] let @xmath92 be the weighted branching tree defined by the nonnegative vector @xmath17 , where @xmath7 . for any @xmath93 and @xmath94 , let @xmath95 . for @xmath91 define the measure @xmath96 , \quad n = 1 , 2 , \dots,\ ] ] and let @xmath97 . suppose that there exists @xmath98 with @xmath99 such that the measure @xmath100 is nonarithmetic , @xmath101 ~ < ~\infty$ ] and @xmath102 = 1 $ ] . then , @xmath103 is a nonarithmetic probability measure on @xmath104 that places no mass at @xmath105 and has mean @xmath106 .\ ] ] furthermore , @xmath107 , where @xmath108 denotes the @xmath31th convolution of @xmath109 with itself . we now present a generalization of goldie s implicit renewal theorem @xcite that will enable the analysis of recursions on weighted branching trees . note that except for the independence assumption , the random variable @xmath0 and the vector @xmath17 are arbitrary , and therefore the applicability of this theorem goes beyond the recursions that we study here . throughout the paper we use @xmath110 as @xmath111 to denote @xmath112 . [ t.newgoldie ] let @xmath17 be a nonnegative random vector , where @xmath7 . suppose that there exists @xmath98 with @xmath113 such that the measure @xmath100 is nonarithmetic . assume further that @xmath114 < \infty$ ] , @xmath115 = 1 $ ] , @xmath116 < \infty$ ] for some @xmath117 , and that @xmath0 is independent of @xmath17 with @xmath118 < \infty$ ] for any @xmath119 . if @xmath120 \right\| t^{\alpha-1 } dt < \infty,\ ] ] then @xmath121 where @xmath122 is given by @xmath123 } \int_{0}^\infty v^{\alpha-1 } \left ( p(r > v ) - e\left [ \sum_{j=1}^{n } \indicator(c_{j } r > v ) \right ] \right ) dv .\end{aligned}\ ] ] remarks : ( i ) as pointed out in @xcite , the statement of the theorem only has content when @xmath0 has infinite moment of order @xmath12 , since otherwise the constant @xmath124 is zero . ( ii ) similarly as in @xcite , this theorem can be generalized to incorporate negative weights @xmath24 at the expense of additional technical complications . however , when the @xmath125 and @xmath0 is real - valued , one can use exactly the same proof to derive the asymptotics of @xmath126 ; we omit the statement here since our applications do not require it . ( iii ) when the @xmath127 are lattice valued , a similar version of the theorem can be derived by using the corresponding renewal theorem for lattice random walks . ( iv ) it appears , as noted in @xcite , that some of the early ideas of applying renewal theory to study the power tail asymptotics of autoregressive processes ( perpetuities ) is due to @xcite and @xcite . the proof given below follows the corresponding proof in @xcite . let @xmath92 be the weighted branching tree defined by the nonnegative vector @xmath17 . for each @xmath94 and all @xmath128 define @xmath129 ; note that @xmath130 is independent of @xmath131 but not of @xmath132 for any @xmath133 . also note that @xmath134 since @xmath94 let @xmath135 , @xmath136 , denote the @xmath137-algebra generated by @xmath138 , and let @xmath139 , @xmath140 . assume also that @xmath0 is independent of the entire weighted tree , @xmath92 . then , for any @xmath141 , we can write @xmath142 via a telescoping sum as follows ( note that all the expectations in are finite by markov s inequality and ) @xmath143 - e\left [ \sum_{({\bf i}\|k+1 ) \in a_{k+1 } } \indicator({\pi}_{{\bf i}\|k+1 } r > e^t ) \right ] \right ) \label{eq : telescoping } \\ & \hspace{5 mm } + e\left [ \sum_{({\bf i}\|n ) \in a_n } \indicator({\pi}_{{\bf i}\|n } r > e^t ) \right ] \notag \\ & = \sum_{k=0}^{n-1 } e\left [ \sum_{({\bf i}\|k ) \in a_{k } } \left ( \indicator ( { \pi}_{{\bf i}\|k } r > e^t ) - \sum_{j=1}^{n_{{\bf i}\|k } } \indicator({\pi}_{{\bf i}\|k } c_{({\bf i}\|k , j ) } r > e^t ) \right ) \right ] \notag \\ & \hspace{5 mm } + e\left [ \sum_{({\bf i}\|n ) \in a_n } \indicator({\pi}_{{\bf i}\|n } r > e^t ) \right ] \notag \\ & = \sum_{k=0}^{n-1 } e\left [ \sum_{({\bf i}\|k ) \in a_{k } } e\left [ \left . \indicator ( r > e^{t - v_{{\bf i}\|k } } ) - \sum_{j=1}^{n_{{\bf i}\|k } } \indicator ( c_{({\bf i}\|k , j ) } r > e^{t - v_{{\bf i}\|k } } ) \right\| \mathcal{f}_k \right ] \right ] \notag \\ & \hspace{5 mm } + e\left [ \sum_{({\bf i}\|n ) \in a_n } \indicator({\pi}_{{\bf i}\|n } r > e^t ) \right ] .\end{aligned}\ ] ] now , define the measures @xmath144 according to lemma [ l.renewalmeasure ] and let @xmath145 \right),\ ] ] @xmath146.\ ] ] recall that @xmath0 and @xmath147 are independent of @xmath135 , from where it follows that @xmath148 = e^{\alpha(v_{{\bf i}\|k}-t ) } g\left ( t - v_{{\bf i}\|k } \right).\ ] ] then , for any @xmath141 and @xmath93 , @xmath149 + \delta_n(t ) = ( g\nu_{n-1})(t ) + \delta_n(t).\end{aligned}\ ] ] next , define the operator @xmath150 and note that @xmath151 now , we will show that one can let @xmath152 in the preceding identity . to this end , let @xmath153 , and note that by lemma [ l.renewalmeasure ] @xmath103 is a nonarithmetic probability measure on @xmath104 that places no mass at @xmath105 and has mean , @xmath154 > 0 .\ ] ] moreover , by lemma [ l.renewalmeasure ] , @xmath155 = \sum_{k=0}^\infty \eta^{k}(dt)\ ] ] is its renewal measure . since @xmath156 , then @xmath157 for all @xmath158 whenever @xmath159 is directly riemann integrable . by we know that @xmath160 , so by lemma 9.1 from @xcite , @xmath161 is directly riemann integrable , resulting in @xmath162 for all @xmath158 . thus , @xmath163 < \infty$ ] , which implies that @xmath164 $ ] exists and , by fubini s theorem , @xmath165 \\ & = \sum_{k=0}^\infty e\left [ \sum_{({\bf i}\|k ) \in a_{k } } e^{\alpha v_{{\bf i}\|k } } \breve{g}(t - v_{{\bf i}\|k } ) \right ] = \lim_{n\to \infty } ( \breve{g}\nu_n)(t).\end{aligned}\ ] ] to see that @xmath166 as @xmath152 for all fixed @xmath158 , note that from the assumptions @xmath167 < \infty$ ] , @xmath115 = 1 $ ] , and @xmath116 < \infty$ ] for some @xmath117 , there exists @xmath168 such that @xmath169 < 1 $ ] ( by convexity ) . then , for such @xmath170 , @xmath171 du \notag \\ & \leq e^{(\alpha-\beta)t } e\left [ \sum_{({\bf i}\|n ) \in a_n } \int_{-\infty}^t e^ { \beta u } \indicator\left({\pi}_{{\bf i}\|n } r > e^u \right ) du \ , \right ] \notag \\ & = e^{(\alpha-\beta ) t } e\left [ \sum_{({\bf i}\|n ) \in a_n } \int_{-\infty}^{\min\{t , \log({\pi}_{{\bf i}\|n } r)\ } } e^{\beta u } du \ , \right ] \notag \\ & \leq \frac{e^{(\alpha-\beta ) t}}{\beta } e\left [ \sum_{({\bf i}\|n ) \in a_n } ( { \pi}_{{\bf i}\|n } r)^{\beta } \right ] . \label{eq : delta_error}\end{aligned}\ ] ] it remains to show that the expectation in converges to zero as @xmath152 . first note that from the independence of @xmath0 and @xmath77 , @xmath172 = e[r^\beta ] e\left [ \sum_{({\bf i}\|n ) \in a_n } ( { \pi}_{{\bf i}\|n})^{\beta } \right],\ ] ] where @xmath118 < \infty$ ] , for @xmath173 . for the expectation involving @xmath174 condition on @xmath175 and use the independence of @xmath176 from @xmath175 as follows @xmath177 & = e\left [ \sum_{({\bf i}\|n-1 ) \in a_{n-1 } } e\left [ \left . \sum_{j=1}^{n_{{\bf i}\|n-1 } } ( { \pi}_{{\bf i}\|n-1})^{\beta } c_{({\bf i}\|n-1,j)}^{\beta } \right\| \mathcal{f}_{n-1 } \right ] \right ] \notag \\ & = e\left [ \sum_{({\bf i}\|n-1 ) \in a_{n-1 } } ( { \pi}_{{\bf i}\|n-1})^{\beta } e\left [ \left . \sum_{j=1}^{n_{{\bf i}\|n-1 } } c_{({\bf i}\|n-1,j)}^{\beta } \right\| \mathcal{f}_{n-1 } \right ] \right ] \notag \\ & = e\left [ \sum_{j=1}^n c_j^{\beta } \right ] e\left [ \sum_{({\bf i}\|n-1 ) \in a_{n-1 } } ( { \pi}_{{\bf i}\|n-1})^{\beta } \right ] \notag \\ & = \left ( e\left [ \sum_{j=1}^{n } c_{j}^{\beta } \right ] \right)^n \qquad \text{(iterating $ n-1 $ times)}. \label{eq : pimoments}\end{aligned}\ ] ] since @xmath178 < 1 $ ] , then the above converges to zero as @xmath152 . hence , the preceding arguments allow us to pass @xmath152 in , and obtain @xmath179 now , by the key renewal theorem for two - sided random walks , see theorem 4.2 in @xcite , @xmath180 clearly , @xmath181 since the left - hand side of the preceding equation is positive , and thus , by lemma 9.3 in @xcite , @xmath182 finally , @xmath183 \right ) dt \\ & = \frac{1 } { \mu } \int_{0}^\infty v^{\alpha-1 } \left ( p(r > v ) - e\left [ \sum_{j=1}^{n } \indicator(c_{j } r > v ) \right ] \right ) dv . \end{aligned}\ ] ] motivated by the information ranking problem on the internet , e.g. google s pagerank algorithm @xcite , in this section we apply the implicit renewal theory for trees developed in the previous section to the following linear recursion : @xmath185 where @xmath186 is a nonnegative random vector with @xmath7 , @xmath187 , and @xmath4 is a sequence of iid random variables independent of @xmath2 having the same distribution as @xmath0 . note that the power tail of @xmath0 in the critical homogeneous case @xmath188 was previously studied in @xcite and @xcite . in section [ ss.homogeneous ] we will give an alternative derivation of those results using our method and will provide pointers to the appropriate literature . as for the nonhomogeneous case , the first result we need to establish is the existence and finiteness of a solution to . for the purpose of existence we will provide an explicit construction of the solution @xmath0 to on a tree . note that such constructed @xmath0 will be the main object of study of this section . recall that throughout the paper the convention is to denote the random vector associated to the root node @xmath28 by @xmath189 . we now define the process @xmath190 on the weighted branching tree @xmath191 , as constructed in section [ s.modeldescription ] . define the process @xmath192 according to @xmath193 that is , @xmath194 is the sum of the weights of all the nodes on the tree up to the @xmath31th generation . it is not hard to see that @xmath194 satisfies the recursion @xmath195 where @xmath196 are independent copies of @xmath197 corresponding to the tree starting with individual @xmath198 in the first generation and ending on the @xmath31th generation ; note that @xmath199 . similarly , since the tree structure repeats itself after the first generation , @xmath200 satisfies @xmath201 where @xmath202 is a sequence of iid random variables independent of @xmath17 and having the same distribution as @xmath203 . next , define the random variable @xmath0 according to @xmath204 where the limit is properly defined by and monotonicity . hence , it is easy to verify , by applying monotone convergence in , that @xmath0 must solve @xmath205 where @xmath206 are iid , have the same distribution as @xmath0 , and are independent of @xmath2 . the derivation provided above implies in particular the existence of a solution in distribution to . moreover , under additional technical conditions , @xmath0 is the unique solution under iterations as we will define and show in the following section . the constructed @xmath0 , as defined in , is the main object of study in the remainder of this section . in this section we derive estimates for the moments of @xmath200 and @xmath0 . we start by stating a lemma about the moments of a sum of random variables . the proofs of lemmas [ l.alpha_moments ] , [ l.momentsmaller_1 ] and [ l.generalmoment ] are given in section [ ss.momentsproofs ] . [ l.alpha_moments ] for any @xmath207 let @xmath208 be a sequence of nonnegative random variables and let @xmath209 be a sequence of nonnegative iid random variables , independent of the @xmath24 , having the same distribution as @xmath210 . for @xmath211 set @xmath212 , and if @xmath213 assume that @xmath214 a.s . then , @xmath215 \leq \left ( e\left [ y^{p-1 } \right ] \right)^{\beta/(p-1 ) } e\left [ \left(\sum_{i=1}^k c_i \right)^\beta \right].\ ] ] remark : note that the preceding lemma does not exclude the case when @xmath216 = \infty$ ] but @xmath217 < \infty$ ] . we now give estimates for the @xmath170-moments of @xmath200 for @xmath218 $ ] and @xmath211 in lemmas [ l.momentsmaller_1 ] and [ l.generalmoment ] , respectively . throughout the rest of the paper define @xmath219 $ ] for any @xmath220 , and @xmath221 . [ l.momentsmaller_1 ] for @xmath222 and all @xmath30 , @xmath223 \leq e[q^\beta ] \rho_\beta^{n}.\ ] ] [ l.generalmoment ] for @xmath211 suppose @xmath224 < \infty$ ] , @xmath225 < \infty$ ] , and @xmath226 . then , there exists a constant @xmath227 such that for all @xmath30 , @xmath223 \leq k_\beta ( \rho \vee \rho_\beta ) ^{n}.\ ] ] now we are ready to establish the finiteness of moments of the solution @xmath0 given by in section [ s.linearrec ] . the proof of this lemma uses well known contraction arguments , but for completeness we provide the details below . [ l.moments_r ] assume that @xmath224 < \infty$ ] for some @xmath220 . in addition , suppose that either ( i ) @xmath228 if @xmath229 , or ( ii ) @xmath230 and @xmath231 < \infty$ ] if @xmath232 . then , @xmath233 < \infty$ ] for all @xmath234 , and in particular , @xmath235 a.s . moreover , if @xmath232 , @xmath236 , where @xmath237 stands for convergence in @xmath238 norm . remark : it is interesting to observe that for @xmath211 the conditions @xmath228 and @xmath225 < \infty$ ] are consistent with theorem 3.1 in @xcite , proposition 4 in @xcite and theorem 2.1 in @xcite , which give the conditions for the finiteness of the @xmath170-moment of the solution to the related critical ( @xmath239 ) homogeneous ( @xmath23 ) equation . let @xmath240 then by lemmas [ l.momentsmaller_1 ] and [ l.generalmoment ] , @xmath241 \leq k \eta^n\ ] ] for some @xmath242 . suppose @xmath232 , then , by monotone convergence and minkowski s inequality , @xmath243 & = e\left [ \lim_{n\to\infty } \left(\sum_{k=0}^n w_k \right)^\beta \right ] = \lim_{n\to \infty } e\left [ \left(\sum_{k=0}^n w_k\right)^\beta \right ] \\ & \leq \lim_{n\to\infty } \left ( \sum_{k=0}^n \left ( e[w_k^\beta ] \right)^{1/\beta } \right)^\beta \leq k \left ( \sum_{k=0}^\infty \eta^{k/\beta } \right)^\beta < \infty.\end{aligned}\ ] ] this implies that @xmath235 a.s . when @xmath244 use the inequality @xmath245 for any @xmath246 instead of minkowski s inequality . furthermore , for any @xmath247 , @xmath248 = e\left [ ( r^\beta)^{\gamma/\beta}\right ] \leq \left(e[r^\beta ] \right)^{\gamma/\beta } < \infty.\ ] ] that @xmath236 whenever @xmath232 follows from noting that @xmath249 $ ] @xmath250 $ ] and applying the same arguments used above to obtain the bound @xmath249 \leq k \eta^{n+1}/(1 - \eta^{1/\beta})^\beta$ ] . next , we show that under some technical conditions , the iteration of recursion results in a process that converges in distribution to @xmath0 for any initial condition @xmath251 . to this end , consider a weighted branching tree @xmath57 , as defined in section [ s.modeldescription ] . now , define @xmath252 where @xmath197 is given by , @xmath253 and @xmath254 are iid copies of an initial value @xmath251 , independent of the entire weighted tree @xmath57 . it follows from and that , for @xmath30 , @xmath255 where @xmath256 are independent copies of @xmath197 corresponding to the tree starting with individual @xmath198 in the first generation and ending on the @xmath31th generation , and @xmath257 is the set of all nodes in the @xmath258th generation that are descendants of individual @xmath198 in the first generation . it follows that @xmath259 where @xmath260 are the expressions inside the parenthesis in . clearly , @xmath260 are iid copies of @xmath261 , thus we show that @xmath261 is equal in distribution to the process derived by iterating with an initial condition @xmath251 . the following lemma shows that @xmath262 for any initial condition @xmath251 satisfying a moment assumption , where @xmath263 denotes convergence in distribution . [ l.convergence ] for any initial condition @xmath264 , if @xmath224 , e[(r_0^)^\beta ] < \infty$ ] and @xmath219 < 1 $ ] for some @xmath222 , then @xmath265 with @xmath118 < \infty$ ] . furthermore , under these assumptions , the distribution of @xmath0 is the unique solution with finite @xmath170-moment to recursion . since @xmath266 a.s . , the result will follow from slutsky s theorem ( see theorem 25.4 , p. 332 in @xcite ) once we show that @xmath267 . to this end , note that @xmath268 , as defined by , is the same as @xmath200 if we substitute the @xmath269 by the @xmath270 . then , for every @xmath271 we have that @xmath272 \\ & \leq \epsilon^{-\beta } \rho_\beta^n e[(r_0^)^\beta ] \qquad \text{(by lemma \ref{l.momentsmaller_1 } ) } .\end{aligned}\ ] ] since by assumption the right - hand side converges to zero as @xmath152 , then @xmath262 . furthermore , @xmath118 < \infty$ ] by lemma [ l.moments_r ] . clearly , under the assumptions , the distribution of @xmath0 represents the unique solution to , since any other possible solution with finite @xmath170-moment would have to converge to the same limit . remarks : ( i ) note that when @xmath273 < 1 $ ] the branching tree is a.s . finite and no conditions on the @xmath24 are necessary for @xmath235 a.s . this corresponds to the second condition in theorem 1 of @xcite . ( ii ) in view of the same theorem from @xcite , one could possibly establish the convergence of @xmath274 under milder conditions . however , since in this paper we only study the power tails of @xmath0 , the assumptions of lemma [ l.convergence ] are not restrictive . ( iii ) note that if @xmath102 = 1 $ ] with @xmath275 $ ] , then there might not be a @xmath173 for which @xmath276 < 1 $ ] , e.g. , the case of deterministic @xmath277 s that was studied in @xcite . we now characterize the tail behavior of the distribution of the solution @xmath0 to the nonhomogeneous equation , as defined by . [ t.linearrecursion ] let @xmath2 be a nonnegative random vector , with @xmath7 , @xmath59 and @xmath0 be the solution to given by . suppose that there exists @xmath98 with @xmath113 such that the measure @xmath278 is nonarithmetic , and that for some @xmath91 , @xmath279 < \infty$ ] , @xmath280 < \infty$ ] and @xmath281 = 1 $ ] . in addition , assume 1 . @xmath15 < 1 $ ] and @xmath282 < \infty$ ] , if @xmath283 ; or , 2 . @xmath284 < \infty$ ] for some @xmath285 , if @xmath286 . then , @xmath121 where @xmath122 is given by @xmath287 } \int_{0}^\infty v^{\alpha-1 } \left ( p(r > v ) - e\left [ \sum_{i=1}^{n } \indicator(c_{i } r > v ) \right ] \right ) dv \\ & = \frac{e\left [ \left ( \sum_{i=1}^n c_i r_i + q \right)^\alpha - \sum_{i=1}^n ( c_i r_i ) ^\alpha \right]}{\alpha e\left [ \sum_{i=1}^n c_i^\alpha \log c_i \right ] } .\end{aligned}\ ] ] remarks : ( i ) the nonhomogeneous equation has been previously studied for the special case when @xmath75 and the @xmath24 are deterministic constants . in particular , theorem 5 of @xcite analyzes the solutions to when @xmath75 and the @xmath24 are nonnegative deterministic constants , which , when @xmath288 , @xmath289 , implies that @xmath290 for all @xmath291 and @xmath292 , falling outside of the scope of this paper . the solutions to for the case when @xmath75 and the @xmath277 s are real valued deterministic constants were analyzed in @xcite . for the very recent work ( published on arxiv after the first draft of this paper ) that characterizes all the solutions to for @xmath75 and @xmath24 random see @xcite . ( ii ) when @xmath283 , the condition @xmath282 < \infty$ ] is needed to ensure that the tail of @xmath0 is not dominated by @xmath51 . in particular , if the @xmath24 are iid and independent of @xmath51 , the condition reduces to @xmath293 < \infty$ ] since @xmath294 < \infty$ ] is implied by the other conditions ; see theorems 4.2 and 5.4 in @xcite . furthermore , when @xmath286 the condition @xmath282 < \infty$ ] is redundant since @xmath282 \leq e\left [ \sum_{i=1}^n c_i^\alpha \right ] = 1 $ ] , and the additional condition @xmath295 < \infty$ ] is needed . when the @xmath296 are iid and independent of @xmath51 , the latter condition reduces to @xmath297 < \infty$ ] ( given the other assumptions ) , which is consistent with theorem 4.2 in @xcite . ( iii ) note that the second expression for @xmath124 is more suitable for actually computing it , especially in the case of @xmath12 being an integer , as will be stated in the forthcoming corollary [ c.explicit ] . when @xmath19 is not an integer , we can derive an explicit upper bound on @xmath124 by using lemma [ l.max_approx ] . regarding the lower bound , the elementary inequality @xmath298 for @xmath299 and @xmath300 , yields @xmath301}{\alpha e\left [ \sum_{i=1}^n c_i^\alpha \log c_i \right ] } > 0.\ ] ] similarly , for @xmath302 , using the corresponding inequality @xmath303 for @xmath304 , @xmath300 , we obtain @xmath305}/{\left(\alpha e\left [ \sum_{i=1}^n c_i^\alpha \log c_i \right ] \right)}.$ ] ( iv ) let us also observe that the solution @xmath0 , given by , to equation may be a constant ( non power law ) @xmath306 when @xmath307 . however , similarly as in remark ( i ) , such a solution is excluded from the theorem since @xmath308 implies @xmath309\le 0 , \alpha>0 $ ] . before proceeding with the proof of theorem [ t.linearrecursion ] , we need the following two technical results ; their proofs are given in section [ ss.linearproofs ] . lemma [ l.max_approx ] below will also be used in subsequent sections for other recursions . with some abuse of notation , we will use throughout the paper @xmath310 to denote @xmath311 in case @xmath13 . [ l.max_approx ] suppose @xmath17 is a nonnegative random vector , with @xmath7 and let @xmath10 be a sequence of iid nonnegative random variables independent of @xmath17 having the same distribution as @xmath0 . for @xmath312 , suppose that @xmath313 a.s . and @xmath118 < \infty$ ] for any @xmath173 . furthermore , assume that @xmath284 < \infty$ ] for some @xmath285 . then , @xmath314 - p\left ( \max_{1\leq i \leq n } c_i r_i > t \right ) \right ) t^{\alpha -1 } \ , dt \\ & = \frac{1}{\alpha } e \left [ \sum_{i=1}^n \left(c_i r_i \right)^\alpha - \left ( \max_{1\leq i \leq n } c_i r_i \right)^\alpha \right ] < \infty.\end{aligned}\ ] ] [ l.extraq ] let @xmath2 be a nonnegative vector with @xmath7 and let @xmath315 be a sequence of iid random variables , independent of @xmath2 . suppose that for some @xmath283 we have @xmath279 < \infty$ ] , @xmath282 < \infty$ ] , @xmath118 < \infty$ ] for any @xmath173 , and @xmath316 a.s . then @xmath317 < \infty.\ ] ] by lemma [ l.moments_r ] , we know that @xmath118 < \infty$ ] for any @xmath173 . to verify that @xmath318 < \infty$ ] for some @xmath117 note that if @xmath283 we have , by the assumptions of the theorem and jensen s inequality , @xmath319 \leq e\left [ \left ( \sum_{i=1}^n c_i \right)^\gamma \right ] \leq \left ( e\left [ \left(\sum_{i=1}^n c_i \right)^\alpha \right ] \right)^{\gamma/\alpha } < \infty\ ] ] for any @xmath320 . if @xmath286 , then for @xmath321 we have @xmath319 \leq e\left [ \left ( \sum_{i=1}^n c_i^{\alpha/(1+\epsilon ) } \right)^{1+\epsilon/2 } \right ] \leq \left ( e\left [ \left ( \sum_{i=1}^n c_i^{\alpha/(1+\epsilon ) } \right)^{1+\epsilon } \right ] \right)^\frac{1+\epsilon/2}{1+\epsilon } < \infty.\ ] ] the statement of the theorem with the first expression for @xmath124 will follow from theorem [ t.newgoldie ] once we prove that condition holds . to this end , define @xmath322 then , @xmath323 \right\| & \leq \left\| p(r > t ) - p\left ( \max_{1\leq i \leq n } c_i r_i > t \right ) \right\| \\ & \hspace{5 mm } + \left\| p\left ( \max_{1\leq i \leq n } c_i r_i > t \right ) - e\left [ \sum_{i=1}^n \indicator(c_ir_i > t ) \right ] \right\|.\end{aligned}\ ] ] since @xmath324 , the first absolute value disappears . for the second one , note that @xmath325 - p\left ( \max_{1\leq i \leq n } c_i r_i > t \right ) \\ & = e\left [ \sum_{i=1}^n \indicator(c_ir_i > t ) \right ] - e\left [ \indicator\left ( \max_{1\leq i \leq n } c_i r_i > t \right ) \right ] \geq 0.\end{aligned}\ ] ] now it follows that @xmath326 \right\| \notag \\ & \leq p(r > t ) - p\left ( \max_{1\leq i \leq n } c_i r_i > t \right ) \notag \\ & \hspace{5 mm } + e\left [ \sum_{i=1}^n \indicator(c_ir_i > t ) \right ] - p\left ( \max_{1\leq i \leq n } c_i r_i > t \right ) . \label{eq : term2}\end{aligned}\ ] ] note that the integral corresponding to is finite by lemma [ l.max_approx ] if we show that the assumptions of lemma [ l.max_approx ] are satisfied when @xmath283 . note that in this case we can choose @xmath271 such that @xmath327 and use the inequality @xmath328 for @xmath329 , @xmath300 , @xmath330 to obtain @xmath331 \leq e\left [ \left ( \sum_{i=1}^n c_i \right)^\alpha \right ] < \infty.\ ] ] therefore , it only remains to show that @xmath332 to see this , note that @xmath333 and @xmath334 , and thus , by fubini s theorem , we have @xmath335.\ ] ] if @xmath286 , we apply to obtain @xmath336 \leq e \left [ q^\alpha + \sum_{i=1}^n ( c_ir_i)^\alpha - \left ( \max_{1\leq i \leq n } c_i r_i \right)^\alpha \right],\ ] ] which is finite by lemma [ l.max_approx ] and the assumption @xmath279 < \infty$ ] . if @xmath283 , we have @xmath337 , @xmath300 , @xmath330 , implying that we can split the expectation as follows @xmath338 & = e \left [ ( r^)^\alpha - \sum_{i=1}^n \left(c_i r_i \right)^\alpha \right ] \\ & \hspace{5 mm } + e \left [ \sum_{i=1}^n \left(c_i r_i \right)^\alpha - \left ( \max_{1\leq i \leq n } c_i r_i \right)^\alpha \right],\end{aligned}\ ] ] which can be done since both expressions inside the expectations on the right - hand side are nonnegative . the first expectation is finite by lemma [ l.extraq ] and the second expectation is again finite by lemma [ l.max_approx ] . finally , applying theorem [ t.newgoldie ] gives @xmath339 where @xmath340 \right)^{-1 } \int_{0}^\infty v^{\alpha-1 } \left ( p(r > v ) - e\left [ \sum_{j=1}^{n } \indicator(c_{j } r > v ) \right ] \right ) dv$ ] . to obtain the second expression for @xmath124 note that @xmath341 \right ) dv \notag \\ & = \int_0^\infty v^{\alpha-1 } e\left[\indicator\left(\sum_{i=1}^n c_i r_i + q > v \right ) - \sum_{i=1}^n \indicator(c_ir_i > v ) \right ] \ , dv \notag \\ & = e \left [ \int_0^\infty v^{\alpha-1 } \left ( \indicator\left(\sum_{i=1}^n c_i r_i + q > v \right ) - \sum_{i=1}^n \indicator(c_ir_i > v ) \right ) dv \right ] \label{eq : fubini } \\ & = e \left [ \int_0^{\sum_{i=1}^n c_i r_i + q } v^{\alpha-1 } dv - \sum_{i=1}^n \int_0^{c_i r_i } v^{\alpha-1 } dv \right ] \label{eq : diffintegrals } \\ & = \frac{1}{\alpha } e\left [ \left ( \sum_{i=1}^n c_i r_i + q \right)^\alpha - \sum_{i=1}^n ( c_i r_i)^\alpha \right ] , \notag\end{aligned}\ ] ] where is justified by fubini s theorem and the integrability of @xmath342 which is a consequence of and lemma [ l.max_approx ] ; and follows from the observation that @xmath343 are each almost surely absolutely integrable with respect to @xmath344 as well . this completes the proof . as indicated earlier , when @xmath345 is an integer , we can obtain the following explicit expression for @xmath124 . [ c.explicit ] for integer @xmath346 , and under the same assumptions of theorem [ t.linearrecursion ] , the constant @xmath124 can be explicitly computed as a function of @xmath347 , e[c^k ] , e[q^k]$ ] , @xmath348 . in particular , for @xmath349 , @xmath350}{e\left [ \sum_{i=1}^n c_i \log c_i \right ] } , \ ] ] and for @xmath351 , @xmath352 + 2 e[r ] e\left [ q \sum_{i=1}^n c_i \right ] + 2 ( e[r])^2 e\left [ \sum_{i=1}^n \sum_{j = i+1}^n c_i c_j \right ] } { 2 e\left [ \sum_{i=1}^n c_i^2 \log c_i \right ] } , \\ & e[r ] = \frac{e[q]}{1-e\left [ \sum_{i=1}^n c_i \right]}.\end{aligned}\ ] ] the proof follows directly from multinomial expansions of the second expression for @xmath124 in theorem [ t.linearrecursion ] . in this section we briefly describe how the methodology developed in the previous sections can be applied to study the critical , @xmath15 = 1 $ ] , homogeneous linear recursion @xmath353 where @xmath17 is a nonnegative random vector with @xmath7 and @xmath354 is a sequence of iid random variables independent of @xmath17 having the same distribution as @xmath0 . this equation has been studied extensively in the literature under various different assumptions ; for recent results see @xcite and the references therein . based on the model from section [ s.linearrec ] we can construct a solution to as follows . consider the process @xmath355 defined by with @xmath356 . then , the @xmath357 satisfy in distribution the homogeneous recursion in and , given that @xmath15 = 1 $ ] , we have @xmath358=1 $ ] . hence , @xmath355 is a nonnegative martingale and by the martingale convergence theorem @xmath359 a.s . with @xmath360\leq 1 $ ] . next , provided that @xmath361 < 0 \quad\text { and } \quad e\left [ \left ( \sum_{i=1}^n c_i \right ) \log^+ \left ( \sum_{i=1}^n c_i \right ) \right ] < \infty\ ] ] it can be shown that @xmath360=1 $ ] , see theorem 1.1(d ) in @xcite ( see also theorem 2 in @xcite ) ; @xmath362 . furthermore , as argued in equation ( 1.9 ) of @xcite , it can easily be shown that this @xmath0 is a solution to . note that the same construction of the solution @xmath0 on a branching tree was given in @xcite and @xcite . since the solutions to are scale invariant , this construction also shows that for any @xmath363 there is a solution @xmath0 with mean @xmath364 ; or equivalently , it is enough to study the solutions with mean @xmath365 . moreover , under additional assumptions it can be shown that this constructed @xmath0 is the only solution with mean @xmath365 , e.g. see @xcite . however , it is not the objective of this section to study the uniqueness of this solution , rather we focus on studying the tail behavior of any such possible solution ( since our theorem [ t.newgoldie ] does not require the uniqueness of @xmath0 ) . as a side note , we point out that can have solutions if @xmath276=1 $ ] for some @xmath366 , as studied in @xcite . a version of the following theorem , with a possibly less explicit constant , was previously proved in theorem 2.2 in @xcite and proposition 7 in @xcite ; they also study the lattice case . regarding the lattice case , as pointed out earlier in the remark after theorem [ t.newgoldie ] , all the results in this paper can be developed for this case as well by using the corresponding renewal theorem . [ t.linearhomog ] suppose that there exists @xmath98 with @xmath113 such that the measure @xmath100 is nonarithmetic . suppose further that for some @xmath283 , @xmath282 < \infty$ ] , @xmath367 < \infty$ ] and @xmath15 = e \left [ \sum_{i=1}^n c_i^\alpha \right ] = 1 $ ] . then , equation has a solution @xmath0 with @xmath368 < \infty$ ] such that @xmath121 where @xmath122 is given by @xmath287 } \int_{0}^\infty v^{\alpha-1 } \left ( p(r > v ) - e\left [ \sum_{i=1}^{n } \indicator(c_{i } r > v ) \right ] \right ) dv \\ & = \frac{e\left [ \left ( \sum_{i=1}^n c_i r_i \right)^\alpha - \sum_{i=1}^n ( c_i r_i ) ^\alpha \right]}{\alpha e\left [ \sum_{i=1}^n c_i^\alpha \log c_i \right ] } .\end{aligned}\ ] ] furthermore , if @xmath369 , @xmath370 , then @xmath371 . by the assumptions , the function @xmath372 $ ] is convex , finite , and continuous on @xmath373 $ ] , since @xmath374 . furthermore , by standard arguments , it can be shown that both @xmath375 and @xmath376 exist on the open interval @xmath377 and , in particular , @xmath378.\ ] ] clearly , @xmath379 provided that @xmath380 . to see that this is indeed the case , note that @xmath15 = 1 $ ] implies that @xmath381 , which combined with the nonarithmetic assumption yields @xmath380 . hence , there exists @xmath382 such that @xmath383 and @xmath384 , implying by the monotonicity of @xmath385 and monotone convergence that @xmath386 \leq e\left [ \sum_{i=1}^n c_i^{\alpha } \log^+ c_i \right ] < \infty \qquad \text{and}\ ] ] @xmath387 < 0 .\ ] ] the last expression and the observation @xmath388 < \infty$ ] ( implied by @xmath282 < \infty$ ] ) yields , as argued at the beginning of this section , that recursion has a solution with finite positive mean , see theorem 1.1(d ) and equation ( 1.9 ) in @xcite ( see also theorem 2 in @xcite ) . next , in order to apply theorem [ t.newgoldie ] , we use and @xmath389 < \infty$ ] for any @xmath173 ; the latter follows from theorem 3.1 in @xcite and the strict convexity of @xmath390 argued above ( see also , proposition 4 in @xcite and theorem 2.1 in @xcite ) . the rest of the proof , except for the @xmath371 part , proceeds exactly as that of theorem [ t.linearrecursion ] by setting @xmath23 . regarding the @xmath371 statement , the assumption @xmath369 implies that there exist @xmath391 and @xmath392 such that @xmath393 , which further yields , for some @xmath394 , @xmath395 next , by using the inequality @xmath396 for @xmath397 and @xmath283 , the second expressions for @xmath124 in the theorem can be bounded from below by @xmath398}{\alpha e\left [ \sum_{i=1}^n c_i^\alpha \log c_i \right ] } .\ ] ] to further bound the numerator in we define the function @xmath399 where @xmath400 , @xmath401 and @xmath402 . it can be shown that @xmath403 for @xmath404 $ ] , since @xmath405 and @xmath406 on @xmath407 $ ] . hence , by using the inequality @xmath403 , we derive for @xmath408 , @xmath409 and @xmath410 @xmath411 the inequality clearly holds even if @xmath412 since both of its sides are zero . thus , by applying this last inequality to and using , we obtain @xmath413}{\alpha e\left [ \sum_{i=1}^n c_i^\alpha \log c_i \right ] } \\ & \geq \frac{c \delta^\alpha p(n\ge i_2 , c_{i_1}>\delta , c_{i_2}>\delta ) e[\left(\min \{r_{i_1},r_{i_2}\}\right)^\alpha ] } { \alpha e\left [ \sum_{i=1}^n c_i^\alpha \log c_i \right ] } > 0.\end{aligned}\ ] ] this completes the proof . remarks : ( i ) note that the assumptions of theorem [ t.linearhomog ] differ slightly from those of theorem [ t.linearrecursion ] in the condition @xmath414 < ~\infty$ ] , which is implied by @xmath415 < \infty$ ] , the strict convexity of @xmath416 $ ] and the hypothesis that @xmath374 , as argued in the preceding proof . ( ii ) the assumption @xmath369 is the minimal one to ensure the existence of a nontrivial solution , see conditions ( h0 ) in @xcite and ( c4 ) in @xcite . otherwise , when @xmath417 , @xmath200 is a simple multiplicative random walk with no branching ; clearly , in this case our expression for @xmath124 reduces to zero . also , if @xmath418 , @xmath0 can only be a constant ; see the remark above theorem 2.1 in @xcite . however , this last case is excluded from the theorem since @xmath418 implies @xmath419 a.s . , which , in conjunction with @xmath420 , yields @xmath421 , but this can not happen due to the nonarithmetic assumption . ( iii ) note also that condition ( c3 ) in @xcite ( equivalent to @xmath380 in our notation ) is implied by the nonarithmetic assumption of our theorem . interestingly enough , if this last condition fails , lemma 1.1 of @xcite shows that equation has no nontrivial solutions . ( iv ) as stated earlier , a version of this theorem was proved in theorem 2.2 of @xcite , by transforming recursion into a first order difference ( autoregressive / perpetuity ) equation on a different probability space , see lemma 4.1 in @xcite . however , it appears that the method from @xcite does not extend to the nonhomogeneous and non - linear problems that we cover here , since the proof of lemma 4.1 in @xcite critically depends on having both @xmath360 = 1 $ ] and @xmath15 = 1 $ ] . similarly as in corollary [ c.explicit ] , the constant @xmath124 can be computed explicitly for integer @xmath422 . [ c.explicithom ] for integer @xmath422 , and under the same assumptions of theorem [ t.linearhomog ] , the constant @xmath124 can be explicitly computed as a function of @xmath347 , e[c^k]$ ] , @xmath423 . in particular , for @xmath351 , @xmath424 } { e\left [ \sum_{i=1}^n c_i^2 \log c_i \right ] } .\ ] ] the proof follows directly from multinomial expansions of the second expression for @xmath124 in theorem [ t.linearhomog ] . we also want to point out that for non - integer general @xmath283 we can use lemma [ l.alpha_moments ] to obtain the following bound for @xmath124 , @xmath425 \right)^{\alpha/(p-1 ) } e\left [ \left ( \sum_{i=1}^n c_i \right)^\alpha\right]}{\alpha e\left [ \sum_{i=1}^n c_i^\alpha \log c_i \right ] } , \ ] ] where @xmath426 . in order to show the general applicability of the implicit renewal theorem , we study in this section the following non - linear recursion : @xmath428 where @xmath2 is a nonnegative random vector with @xmath7 , @xmath59 and @xmath4 is a sequence of iid random variables independent of @xmath2 having the same distribution as @xmath0 . note that in the case of page ranking applications , where the @xmath315 represent the ranks of the neighboring pages , the potential ranking algorithm defined by the preceding recursion , determines the rank of a page as a weighted version of the most highly ranked neighboring page . in other words , the highest ranked reference has the dominant impact . similarly to the homogeneous linear case , this recursion was previously studied in @xcite under the assumption that @xmath23 , @xmath13 , and the @xmath24 are real valued deterministic constants . the more closely related case of @xmath23 and @xmath25 being random was studied earlier in @xcite . furthermore , these max - type stochastic recursions appear in a wide variety of applications , ranging from the average case analysis of algorithms to statistical physics ; see @xcite for a recent survey . using standard arguments , we start by constructing a solution to on a tree and then we show that this solution is finite a.s . and unique under iterations and some moment conditions , similar to what was done for the nonhomogeneous linear recursion in section [ s.linearrec ] . our main result of this section is stated in theorem [ t.maximumrecursion ] . following the same notation as in section [ s.linearrec ] , define the process @xmath429 on the weighted branching tree @xmath191 , as constructed in section [ s.modeldescription ] . recall that the convention is that @xmath430 denotes the random vector corresponding to the root node . with a possible abuse of notation relative to section [ s.linearrec ] , define the process @xmath192 according to @xmath431 just as with the linear recursion from section [ s.linearrec ] , it is not hard to see that @xmath194 satisfies the recursion @xmath432 where @xmath433 are independent copies of @xmath197 corresponding to the tree starting with individual @xmath198 in the first generation and ending on the @xmath31th generation . one can also verify that @xmath434 where @xmath435 is a sequence of iid random variables independent of @xmath17 and having the same distribution as @xmath436 . we now define the random variable @xmath0 according to @xmath437 note that @xmath194 is monotone increasing sample - pathwise , so @xmath0 is well defined . also , by monotonicity of @xmath194 , and monotone convergence , we obtain that @xmath0 solves @xmath438 where @xmath439 are iid copies of @xmath0 , independent of @xmath2 . clearly this implies that @xmath0 , as defined by , is a solution in distribution to . however , this solution might be @xmath440 . now , we establish the finiteness of the moments of @xmath0 , and in particular that @xmath235 a.s . , in the following lemma ; its proof uses standard contraction arguments but is included for completeness . [ l.moments_r_max ] assume that @xmath219<1 $ ] and @xmath224 < \infty$ ] for some @xmath441 . then , @xmath233 < \infty$ ] for all @xmath234 , and in particular , @xmath235 a.s . moreover , if @xmath442 , @xmath236 , where @xmath237 stands for convergence in @xmath238 norm . by following the same steps leading to , we obtain that for any @xmath443 , @xmath444 = e\left [ \bigvee_{{\bf i } \in a_k } q_{\bf i}^\beta { \pi}_{\bf i}^\beta \right ] \leq e\left [ \sum_{{\bf i } \in a_k } q_{\bf i}^\beta { \pi}_{\bf i}^\beta \right ] = e[q^\beta ] \rho_\beta^k.\ ] ] hence , @xmath445 = e\left [ \bigvee_{k=0}^\infty v_k^\beta \right ] \leq e\left [ \sum_{k=0}^\infty v_k^\beta \right ] \leq \frac{e[q^\beta]}{1-\rho_\beta } < \infty,\ ] ] implying that @xmath233 < \infty$ ] for all @xmath247 . that @xmath236 whenever @xmath446 follows from noting that @xmath249 \le e\left [ \left ( \bigvee_{k = n+1}^\infty v_k \right)^\beta \right ] \leq e\left [ \sum_{k = n+1}^\infty v_k^\beta \right]$ ] and applying the preceding geometric bound for @xmath447 $ ] . just as with the linear recursion from section [ s.linearrec ] , we can define the process @xmath448 as @xmath449 where @xmath450 and @xmath254 are iid copies of an initial value @xmath251 , independent of the entire weighted tree @xmath57 . it follows from and that @xmath451 where @xmath256 are independent copies of @xmath197 corresponding to the tree starting with individual @xmath198 in the first generation and ending on the @xmath31th generation , and @xmath257 is the set of all nodes in the @xmath258th generation that are descendants of individual @xmath198 in the first generation . moreover , @xmath260 are iid copies of @xmath261 , and thus , @xmath261 is equal in distribution to the process obtained by iterating with an initial condition @xmath251 . this process can be shown to converge in distribution to @xmath0 for any initial condition @xmath251 satisfying the following moment condition . [ l.convergencemax ] for any @xmath264 , if @xmath224 , e[(r_0^)^\beta ] < \infty$ ] and @xmath228 for some @xmath452 , then @xmath265 with @xmath118 < \infty$ ] . furthermore , under these assumptions , the distribution of @xmath0 is the unique solution with finite @xmath170-moment to recursion . the result will follow from slutsky s theorem ( see theorem 25.4 , p. 332 in @xcite ) once we show that @xmath453 . to this end , recall that @xmath454 is the same as @xmath455 if we substitute the @xmath269 by the @xmath270 . then , for every @xmath271 we have that @xmath456 \\ & \leq \epsilon^{-\beta } \rho_\beta^n e[(r_0^)^\beta ] \qquad \text{(by \eqref{eq : vmoment } ) } .\end{aligned}\ ] ] since by assumption the right - hand side converges to zero as @xmath152 , then @xmath262 . furthermore , @xmath118 < \infty$ ] by lemma [ l.moments_r_max ] . clearly , under the assumptions , the distribution of @xmath0 represents the unique solution to , since any other possible solution with finite @xmath170-moment would have to converge to the same limit . now we are ready to state the main result of this section . [ t.maximumrecursion ] let @xmath2 be a nonnegative random vector , with @xmath7 , @xmath59 and @xmath0 be the solution to given by . suppose that there exists @xmath98 with @xmath113 such that the measure @xmath100 is nonarithmetic , and that for some @xmath91 , @xmath279 < \infty$ ] , @xmath280 < \infty$ ] and @xmath281 = 1 $ ] . in addition , assume 1 . @xmath282 < \infty$ ] , , if @xmath283 ; or , 2 . @xmath284 < \infty$ ] for some @xmath285 , if @xmath286 . then , @xmath121 where @xmath122 is given by @xmath287 } \int_{0}^\infty v^{\alpha-1 } \left ( p(r > v ) - e\left [ \sum_{i=1}^{n } \indicator(c_{i } r > v ) \right ] \right ) dv \\ & = \frac{e\left [ \left ( \bigvee_{i=1}^n c_i r_i \right)^\alpha \vee q^\alpha - \sum_{i=1}^n ( c_i r_i ) ^\alpha \right]}{\alpha e\left [ \sum_{i=1}^n c_i^\alpha \log c_i \right ] } .\end{aligned}\ ] ] by lemma [ l.moments_r_max ] we know that @xmath118 < \infty$ ] for any @xmath173 . the same arguments used in the proof of theorem [ t.linearrecursion ] give that @xmath318 < \infty$ ] for some @xmath457 . the statement of the theorem with the first expression for @xmath124 will follow from theorem [ t.newgoldie ] once we prove that condition holds . define @xmath458 then , @xmath323 \right\| & \leq \left\| p(r > t ) - p\left ( \max_{1\leq i \leq n } c_i r_i > t \right ) \right\| \\ & \hspace{5 mm } + \left\| p\left ( \max_{1\leq i \leq n } c_i r_i > t \right ) - e\left [ \sum_{i=1}^n \indicator(c_ir_i > t ) \right ] \right\|.\end{aligned}\ ] ] since @xmath324 , the first absolute value disappears . the integral corresponding to the second term is finite by lemma [ l.max_approx ] , just as in the proof of theorem [ t.linearrecursion ] . to see that the integral corresponding to the first term , @xmath459 is finite we proceed as in the proof of theorem [ t.linearrecursion ] . first we use fubini s theorem to obtain that @xmath460 \\ & = \frac{1}{\alpha } e \left [ \left ( \bigvee_{i=1}^n c_i r_i \right)^\alpha \vee q^\alpha - \left ( \bigvee_{i=1}^n c_i r_i \right)^\alpha \right ] \\ & \leq \frac{e[q^\alpha]}{\alpha}.\end{aligned}\ ] ] now , applying theorem [ t.newgoldie ] gives @xmath339 where @xmath340 \right)^{-1 } \int_{0}^\infty v^{\alpha-1 } \left ( p(r > v ) - e\left [ \sum_{j=1}^{n } \indicator(c_{j } r > v ) \right ] \right ) dv$ ] . the same steps used in the proof of theorem [ t.linearrecursion ] give the second expression for @xmath124 . as an illustration of the generality of the methodology presented in this paper , we mention in this section other recursions that fall within its scope . one example that is closely related to the recursions from sections [ s.linearrec ] and [ s.maxrec ] is the following @xmath461 where @xmath2 is a nonnegative vector with @xmath7 , @xmath59 , and @xmath462 is a sequence of iid random variables independent of @xmath2 having the same distribution as @xmath0 . recursion was termed discounted tree sums " in @xcite ; for additional details on the existence and uniqueness of its solution see section 4.4 in @xcite . similarly as in @xcite , it appears that one could study other non - linear recursions on trees using implicit renewal theory . for example , one could analyze the solution to the equation @xmath463 where @xmath2 is a nonnegative vector with @xmath7 , @xmath59 , and @xmath464 is a sequence of iid random variables independent of @xmath2 . here , the primary difficulty would be in establishing the existence and uniqueness of the solution as well as the finiteness of moments . we give in this section the proof of lemma [ l.renewalmeasure ] . observe that the measure @xmath465 $ ] is nonarithmetic ( nonlattice ) by our assumption since , if we assume to the contrary that it is lattice on a lattice set @xmath466 , then on the complement @xmath467 of this set we have @xmath468\ge p(\log c_j \in l^c , c_j>0,n \ge j)>0,\ ] ] which is a contradiction . hence , @xmath109 is nonarithmetic as well , and it places no mass at @xmath105 due to the exponential term @xmath469 . to see that it is a probability measure note that @xmath470 \\ & = e\left [ \sum_{j=1}^n \int_{-\infty}^\infty e^{\alpha u } \indicator(\log c_j \in du ) \right ] \qquad \text{(by fubini 's theorem ) } \\ & = e\left [ \sum_{j=1}^n c_j^\alpha \right ] = 1.\end{aligned}\ ] ] similarly , its mean is given by @xmath471 .\ ] ] to show that @xmath472 we proceed by induction . let @xmath473 denote the @xmath137-algebra generated by @xmath474 , @xmath139 , and for each @xmath94 set @xmath475 . hence , using this notation we derive @xmath476 ) & = \int_{-\infty}^t e^{\alpha u } e\left [ \sum_{{\bf i } \in a_{n } } \sum_{j=1}^{n_{{\bf i } } } \indicator(v_{{\bf i } } + \log c_{({\bf i},j ) } \in du ) \right ] \\ & = \int_{-\infty}^t e^{\alpha u } e\left [ \sum_{{\bf i } \in a_{n } } e\left [ \left . \sum_{j=1}^{n_{{\bf i } } } \indicator(v_{{\bf i } } + \log c_{({\bf i},j ) } \in du ) \right\| \mathcal{f}_n \right ] \right ] \\ & = e\left [ \sum_{{\bf i } \in a_{n } } e^{\alpha v_{\bf i } } \int_{-\infty}^t e^{\alpha ( u- v_{\bf i } ) } e\left [ \left . \sum_{j=1}^{n_{{\bf i } } } \indicator ( \log c_{({\bf i},j ) } \in du - v_{\bf i } ) \right\| \mathcal{f}_n \right ] \right ] \\ & = e\left [ \sum_{{\bf i } \in a_{n } } e^{\alpha v_{\bf i } } \eta((-\infty , t - v_{{\bf i } } ] ) \right ] \\ & = \int_{-\infty}^\infty \eta((-\infty , t - v ] ) \mu_n(dv),\end{aligned}\ ] ] where in the fourth equality we used the independence of @xmath477 from @xmath473 . therefore @xmath478 and the induction hypothesis gives the result . in this section we prove lemmas [ l.alpha_moments ] , [ l.momentsmaller_1 ] and [ l.generalmoment ] . we also include a result that provides bounds for @xmath479 $ ] for integer @xmath480 , which will be used in the proof of lemma [ l.generalmoment ] . let @xmath481 and @xmath482 $ ] . suppose first that @xmath483 and define @xmath484 . then , for any sequence of nonnegative numbers @xmath485 we have @xmath486 where for the last step we used the well known inequality @xmath487 for @xmath488 and @xmath300 . we now use the conditional jensen s inequality to obtain @xmath489 \\ & \leq e\left [ \left ( \sum_{(j_1,\dots , j_k ) \in a_p(k ) } \binom{p}{j_1,\dots , j_k } ( c_1y_1)^{j_1 } \cdots ( c_k y_k)^{j_k } \right)^\gamma \right ] \qquad \text{(by \eqref{eq : scalarbound } ) } \\ & \leq e\left [ \left ( e\left [ \left . \sum_{(j_1,\dots , j_k ) \in a_p(k ) } \binom{p}{j_1,\dots , j_k } ( c_1y_1)^{j_1 } \cdots ( c_k y_k)^{j_k } \right\| c_1,\dots , c_k \right ] \right)^\gamma \right ] \\ & = e \left [ \left ( \sum_{(j_1,\dots , j_k ) \in a_p(k ) } \binom{p}{j_1,\dots , j_k } c_1^{j_1 } \cdots c_k^{j_k } e\left [ \left . y_1^{j_1 } \cdots y_k^{j_k } \right\| c_1,\dots , c_k \right ] \right)^\gamma \right].\end{aligned}\ ] ] since @xmath490 is a sequence of iid random variables having the same distribution as @xmath210 , independent of the @xmath277 s we have that @xmath491 = e\left [ y_1^{j_1 } \cdots y_k^{j_k } \right ] = \|\| y \|\|_{j_1}^{j_1 } \cdots \|\| y \|\|_{j_k}^{j_k},\ ] ] where @xmath492 \right)^{1/\kappa}$ ] for @xmath493 and @xmath494 . since @xmath495 is increasing for @xmath493 it follows that @xmath496 . hence @xmath497 which in turn implies that @xmath498 & \leq e \left [ \left ( \sum_{(j_1,\dots , j_k ) \in a_p(k ) } \binom{p}{j_1,\dots , j_k } c_1^{j_1 } \cdots c_k^{j_k } \|\| y \|\|_{p-1}^p \right)^\gamma \right ] \\ & = \|\| y \|\|_{p-1}^\beta e\left [ \left ( \left(\sum_{i=1}^k c_i \right)^p - \sum_{i=1}^k c_i^p \right)^\gamma \right ] \\ & \leq \|\| y \|\|_{p-1}^\beta e\left [ \left(\sum_{i=1}^k c_i \right)^\beta \right].\end{aligned}\ ] ] this completes the proof for @xmath499 finite . when @xmath213 , first note that from the well known inequality @xmath500 for @xmath397 and @xmath211 we obtain the monotonicity in @xmath499 of the following difference @xmath501 hence , @xmath502 \notag \\ & = \lim_{k\to \infty } e\left [ \left ( \left ( \sum_{i=1}^k c_iy_i \right)^\beta - \sum_{i=1}^k ( c_i y_i)^{\beta } \right ) \right ] \label{eq : exchange1 } \\ & \leq \lim_{k\to \infty } e\left [ \left ( \sum_{(j_1,\dots , j_k ) \in a_p(k ) } \binom{p}{j_1,\dots , j_k } ( c_1y_1)^{j_1 } \cdots ( c_k y_k)^{j_k } \right)^\gamma \right ] \notag \\ & \leq \lim_{k \to \infty } \|\| y \|\|_{p-1}^\beta e\left [ \left(\sum_{i=1}^k c_i \right)^\beta \right ] \notag \\ & = \|\| y \|\|_{p-1}^\beta e\left [ \left(\sum_{i=1}^\infty c_i \right)^\beta \right ] , \label{eq : exchange2}\end{aligned}\ ] ] where and are justified by monotone convergence . we use the well known inequality @xmath503 for @xmath222 , @xmath246 , @xmath330 , to obtain @xmath504 & = e\left [ \left ( \sum_{i=1}^n c_i w_{(n-1),i } \right)^\beta \right ] \notag \\ & \leq e\left [ \sum_{i=1}^n c_i^\beta w_{(n-1),i}^\beta \right ] \notag \\ & = e[w_{n-1}^\beta ] \rho_\beta \qquad \text{(by conditioning on $ n , c_i$ and fubini 's theorem ) } \notag \\ & \leq \rho_\beta^{n } e[w_0^\beta ] \qquad \text{(iterating $ n$ times ) } \notag \\ & = \rho_\beta^{n } e[q^\beta ] . \end{aligned}\ ] ] before proving the moment inequality for general @xmath211 , we will show first the corresponding result for integer moments . [ l.integermoment ] let @xmath505 and recall that @xmath506 $ ] , @xmath221 . < \infty$ ] , @xmath508 < \infty$ ] , and @xmath509 . then , there exists a constant @xmath510 such that @xmath511 \leq k_p \left ( \rho \vee \rho_p \right)^n\ ] ] for all @xmath30 . we will give an induction proof in @xmath480 . for @xmath512 we have @xmath513 & = e\left [ \left ( \sum_{i=1}^n c_i w_{(n-1),i } \right)^2 \right ] \\ & = e\left [ \sum_{i=1}^n c_i^2 w_{(n-1),i}^2 + \sum_{i \neq j } c_i w_{(n-1),i } c_j w_{(n-1),j } \right ] \\ & = e[w_{n-1}^2 ] e\left [ \sum_{i=1}^n c_i^2 \right ] + \left ( e[w_{n-1 } ] \right)^2 e\left [ \sum_{i \neq j } c_i c_j \right ] \\ & \hspace{4.5 cm } \text{(by conditioning on $ n , c_i$ and fubini 's theorem ) } \\ & \leq \rho_2 e[w_{n-1}^2 ] + e\left [ \left ( \sum_{i=1}^n c_i \right)^2 \right ] \left ( e[w_{n-1 } ] \right)^2 .\end{aligned}\ ] ] using the preceding recursion and noting that , @xmath514 = \rho^{n-1 } e[q],\ ] ] we obtain @xmath515 \leq \rho_2 e[w_{n-1}^2 ] + k \rho^{2(n-1)},\ ] ] where @xmath516 \left ( e[q ] \right)^2 $ ] . now , iterating gives @xmath513 & \leq \rho_2 \left ( \rho_2 e[w_{n-2}^2 ] + k \rho^{2(n-2 ) } \right ) + k \rho^{2(n-1 ) } \\ & \leq \rho_2^{n-1 } \left ( \rho_2 e[w_{0}^2 ] + k \right ) + k \sum_{i=0}^{n-2 } \rho_2^i \ , \rho^{2(n-1-i ) } \\ & = \rho_2^n e[q^2 ] + k \sum_{i=0}^{n-1 } \rho_2^i \ , \rho^{2(n-1-i ) } \\ & \leq ( \rho_2 \vee \rho)^n e[q^2 ] + k ( \rho_2 \vee \rho)^n \sum_{i=0}^{n-1 } ( \rho_2 \vee \rho)^{n-2 - i } \\ & \leq \left ( e[q^2 ] + \frac{k}{\rho_2 \vee \rho } \sum_{j=0}^{\infty } ( \rho_2 \vee \rho)^{j } \right ) ( \rho_2 \vee \rho)^n \\ & = k_2 ( \rho_2 \vee \rho)^n , \end{aligned}\ ] ] which completes the case @xmath512 . suppose now that there exists a constant @xmath517 such that @xmath518 \leq k_{p-1 } \left ( \rho_{p-1 } \vee \rho \right)^n\ ] ] for all @xmath30 . then , by lemmas [ l.alpha_moments ] and [ l.momentsmaller_1 ] , we have @xmath519 & = e\left [ \left ( \sum_{i=1}^n c_i w_{(n-1),i } \right)^p - \sum_{i=1}^n c_i^p w_{(n-1),i}^p \right ] + e\left [ \sum_{i=1}^n c_i^p w_{(n-1),i}^p \right ] \\ & \leq \left ( e\left [ w_{n-1}^{p-1 } \right ] \right)^{p/(p-1 ) } e\left [ \left ( \sum_{i=1}^n c_i \right)^p \right ] + e\left [ \sum_{i=1}^n c_i^p w_{(n-1),i}^p \right ] \\ & = \left ( e\left [ w_{n-1}^{p-1 } \right ] \right)^{p/(p-1 ) } e\left [ \left ( \sum_{i=1}^n c_i \right)^p \right ] + \rho_p e\left [ w_{n-1}^{p } \right ] \\ & \leq e\left [ \left ( \sum_{i=1}^n c_i \right)^p \right ] ( k_{p-1})^{p/(p-1 ) } ( \rho_{p-1 } \vee \rho)^{(n-1)p/(p-1 ) } + \rho_p e[w_{n-1}^p],\end{aligned}\ ] ] where in the second equality we conditioned on @xmath520 and used fubini s theorem , and the last inequality corresponds to the induction hypothesis . we then obtain the recursion @xmath521 \leq \rho_p e[w_{n-1}^p ] + k ( \rho_{p-1 } \vee \rho)^{\frac{(n-1)p}{p-1}},\ ] ] where @xmath522 ( k_{p-1})^{p/(p-1)}$ ] . iterating as for the case @xmath523 gives @xmath519 & \leq \rho_p^n e[q^p ] + k \sum_{i=0}^{n-1 } \rho_p^i \ , ( \rho_{p-1 } \vee \rho)^{(n-1-i)p/(p-1 ) } \notag \\ & \leq ( \rho_p \vee \rho)^n e[q^p ] + k \sum_{i=0}^{n-1 } ( \rho_p \vee \rho)^{((n-1)p -i)/(p-1 ) } \label{eq : mgfconvexity } \\ \phantom{e[w_n^p]}&= ( \rho_p \vee \rho)^n e[q^p ] + k ( \rho_p \vee \rho)^{n-1 } \sum_{i=0}^{n-1 } ( \rho_p \vee \rho)^{(n- i - 1)/(p-1 ) } \notag \\ & \leq \left ( e[q^p ] + k ( \rho_p \vee \rho)^{-1 } \sum_{j=0}^{\infty } ( \rho_p \vee \rho)^{\frac{j}{p-1 } } \right ) ( \rho_p \vee \rho)^n \notag \\ & = k_p ( \rho_p \vee \rho)^n , \notag\end{aligned}\ ] ] where in we used the convexity of @xmath524 , i.e. @xmath525 . the proof for the general @xmath170-moment , @xmath211 , is given below . set @xmath526 . then , by lemmas [ l.alpha_moments ] and [ l.momentsmaller_1 ] , @xmath504 & = e\left [ \left ( \sum_{i=1}^n c_i w_{(n-1),i } \right)^\beta - \sum_{i=1}^n c_i^\beta w_{(n-1),i}^\beta \right ] + e\left [ \sum_{i=1}^n c_i^\beta w_{(n-1),i}^\beta \right ] \\ & \leq \left ( e\left [ w_{n-1}^{p-1 } \right ] \right)^{\beta/(p-1 ) } e\left [ \left ( \sum_{i=1}^n c_i \right)^\beta \right ] + e\left [ \sum_{i=1}^n c_i^\beta w_{(n-1),i}^\beta \right ] \\ & = \left ( e\left [ w_{n-1}^{p-1 } \right ] \right)^{\beta/(p-1 ) } e\left [ \left ( \sum_{i=1}^n c_i \right)^\beta \right ] + \rho_\beta e\left [ w_{n-1}^{\beta } \right ] .\end{aligned}\ ] ] by lemma [ l.integermoment ] , @xmath504 & \leq \rho_\beta e[w_{n-1}^\beta ] + e\left [ \left ( \sum_{i=1}^n c_i \right)^\beta \right ] ( k_{p-1})^{\beta/(p-1 ) } ( \rho_{p-1 } \vee \rho)^{(n-1)\beta/(p-1 ) } \\ & = \rho_\beta e [ w_{n-1}^\beta ] + k ( \rho_{p-1 } \vee \rho)^{(n-1)\gamma},\end{aligned}\ ] ] where @xmath527 . finally , iterating the preceding bound @xmath528 times gives @xmath504 & \leq \rho_\beta^n e[w_0^\beta ] + k \sum_{i=0}^{n-1 } \rho_\beta^i ( \rho \vee \rho_{p-1})^{\gamma(n-1-i ) } \\ & \leq e[w_0^\beta ] ( \rho \vee \rho_\beta)^n + k \sum_{i=0}^{n-1 } ( \rho \vee \rho_\beta)^{\gamma(n-1-i ) + i } \qquad \text{(by convexity of $ \varphi(\beta ) = \rho_\beta$ ) } \\ & = e[q^\beta ] ( \rho \vee \rho_\beta)^n + k ( \rho \vee \rho_\beta)^{n-1 } \sum_{i=0}^{n-1 } ( \rho \vee \rho_\beta)^{(\gamma-1 ) i } \\ & \leq k_\beta ( \rho \vee \rho_\beta)^n .\end{aligned}\ ] ] this completes the proof . in this section we give the proofs of the technical lemmas [ l.max_approx ] and [ l.extraq ] for the linear recursion . note that the integral is positive since @xmath529 & \leq e\left [ \sum_{i=1}^n \indicator\left ( c_i r_i > t \right ) \right ] .\end{aligned}\ ] ] to see that the integral is equal to the expectation involving the @xmath12-moments note that @xmath530 - p\left ( \max_{1\leq i \leq n } c_i r_i > t \right ) \right ) t^{\alpha -1 } \ , dt \\ & = \int_0^\infty e\left [ \sum_{i=1}^n \indicator(c_i r_i > t ) - \indicator\left(\max_{1\leq i \leq n } c_i r_i > t \right ) \right ] t^{\alpha -1 } \ , dt \\ & = e\left [ \int_0^\infty \left ( \sum_{i=1}^n \indicator(c_i r_i > t ) - \indicator\left ( \max_{1\leq i \leq n } c_i r_i > t \right ) \right ) t^{\alpha -1 } \ , dt \right ] \qquad \text{(by fubini 's theorem ) } \\ & = e\left [ \sum_{i=1}^n \frac{1}{\alpha } ( c_i r_i)^{\alpha } - \frac{1}{\alpha } \left(\max_{1\leq i \leq n } c_i r_i \right)^{\alpha } \right ] , \end{aligned}\ ] ] where the last equality is justified by the assumption that @xmath313 a.s . it now remains to show that the integral ( expectation ) is finite . to do this let @xmath531 . similar arguments to those used above give @xmath532 - p\left ( \max_{1\leq i \leq n } c_i r_i > t \right ) \right ) t^{\alpha -1 } \ , dt \\ & = \int_0^\infty e\left [ e\left [ \left . \sum_{i=1}^n \indicator(c_i r_i > t ) - \indicator\left(\max_{1\leq i \leq n } c_i r_i > t \right ) \right\| { \bf x } \right ] \right ] t^{\alpha -1 } \ , dt \\ & = e\left [ \int_0^\infty e\left [ \left . \sum_{i=1}^n \indicator(c_i r_i > t ) - \indicator\left(\max_{1\leq i \leq n } c_i r_i > t \right ) \right\| { \bf x } \right ] t^{\alpha -1 } \ , dt \right ] , \end{aligned}\ ] ] where in the last step we used fubini s theorem . furthermore , @xmath533 \\ & = e\left [ \left . \indicator\left ( \max_{1\leq i \leq n } c_i r_i \leq t \right ) \right\| { \bf x } \right ] - 1 + \sum_{i=1}^n e\left [ \indicator(c_i r_i > t ) \| { \bf x}\right ] . \end{aligned}\ ] ] note that given @xmath534 , the random variables @xmath535 are independent ( since the @xmath0 s are ) , so if we let @xmath536 , then @xmath537 = \prod_{i=1}^n e\left [ \left . \indicator\left ( c_i r_i \leq t \right ) \right\| { \bf x } \right ] = \prod_{i=1}^n \left(1 - \overline{f}(t / c_i ) \right).\ ] ] we now use the inequality @xmath538 for @xmath539 to obtain @xmath540 next , let @xmath541 and set @xmath542 . by markov s inequality , @xmath543 t^{-\beta } = t^{-\beta } e[r^\beta ] \sum_{i=1}^n c_i^\beta.\ ] ] now , define the function @xmath544 and note that @xmath545 is increasing for @xmath539 . therefore , @xmath546 \sum_{i=1}^n c_i^\beta \right).\ ] ] this observation combined with the previous derivations gives @xmath547 t^{\alpha -1 } \ , dt \\ & \leq \int_0^\infty \left ( e^{-r s_\beta t^{-\beta } } - 1 + r s_\beta t^{-\beta } \right ) t^{\alpha-1 } dt,\end{aligned}\ ] ] where @xmath548 and @xmath549 < \infty$ ] . hence , using the change of variables @xmath550 gives @xmath551 our choice of @xmath542 guarantees that @xmath552 . to see that the ( non - random ) integral is finite note that @xmath553 and @xmath554 for any @xmath539 , implying @xmath555 now , it follows that @xmath530 - p\left ( \max_{1\leq i \leq n } c_i r_i > t \right ) \right ) t^{\alpha -1 } \ , dt \\ & \leq e\left [ k ( r s_\beta)^{\alpha/\beta } \right ] = k r^{\alpha/\beta } e\left [ \left ( \sum_{i=1}^n c_i^\beta \right)^{\alpha/\beta } \right ] .\end{aligned}\ ] ] the last expectation is finite by assumption ( @xmath556 ) , which completes the proof . let @xmath557 a.s . , @xmath426 and note that @xmath558 . then , since @xmath559 and @xmath560 , we can break the expectation as follows @xmath561 & = e\left [ ( s+q)^\alpha - s^\alpha \right ] + e\left [ \left ( \sum_{i=1}^n c_i r_i \right)^\alpha - \sum_{i=1}^n ( c_ir_i)^\alpha \right ] \\ & \leq e\left [ ( s+q)^\alpha - s^\alpha \right ] + \left ( e [ r^{p-1 } ] \right)^{\alpha/(p-1 ) } e\left [ \left(\sum_{i=1}^n c_i \right)^\alpha \right ] , \end{aligned}\ ] ] where the inequality is justified by lemma [ l.alpha_moments ] . the second expectation is finite since by assumption @xmath118 < \infty$ ] for any @xmath562 . for the first expectation we use the inequality @xmath563 for any @xmath564 . we apply the second inequality @xmath565 times and then the first one to obtain @xmath566 hence , it follows that @xmath567 \leq \alpha^p e[q^\alpha ] + \alpha^p\sum_{i=1}^{p-1 } e[s^{\alpha - i } q^i].\ ] ] to see that each of the expectations involving a product of @xmath568 and @xmath75 is finite let @xmath569 and note that for @xmath570 , @xmath571 \\ & = e\left [ e\left [ \left . \left ( q^{(p-1)/(\alpha - p+1 ) } \sum_{j=1}^n c_j r_j \right)^{\alpha - p+1 } \right\| { \bf x } \right ] \right ] \notag \\ & \leq e\left [ \left ( e\left [ \left . q^{(p-1)/(\alpha - p+1 ) } \sum_{j=1}^n c_j r_j \right\| { \bf x } \right ] \right)^{\alpha - p+1 } \right ] \quad \text{(by jensen 's inequality ) } \notag \\ & = ( e [ r ] ) ^{\alpha - p+1 } e\left [ q^{p-1 } \left ( \sum_{j=1}^n c_j \right)^{\alpha - p+1 } \right ] , \label{eq : prodmoments_1}\end{aligned}\ ] ] where the last equality was obtained by using the independence of @xmath572 and @xmath534 . for @xmath573 let @xmath574 and condition on @xmath75 and @xmath534 , respectively , to obtain @xmath575 & = e\left [ \left ( s^{\alpha - i } - \sum_{j=1}^n ( c_j r_j)^{\alpha - i } \right ) q^i \right ] + e\left [ q^i \sum_{j=1}^n ( c_j r_j)^{\alpha - i } \right ] \notag \\ & = e\left [ q^i e\left [ \left . s^{\alpha - i } - \sum_{j=1}^n ( c_j r_j)^{\alpha - i } \right\| q \right ] \right ] + e [ r^{\alpha - i } ] e\left [ q^i \sum_{j=1}^n c_j^{\alpha - i } \right ] \notag \end{aligned}\ ] ] @xmath576 \right)^{\frac{\alpha - i}{q_i - 1 } } e\left [ \left . \left ( \sum_{j=1}^n c_j \right)^{\alpha - i } \right\| q \right ] \right ] \\ & \hspace{5 mm } + e [ r^{\alpha - i } ] e\left [ q^i \left ( \sum_{j=1}^n c_j \right)^{\alpha - i } \right ] \notag \\ & = \left ( \left ( e[r^{q_i - 1 } ] \right)^{\frac{\alpha - i}{q_i - 1 } } + e[r^{\alpha - i } ] \right ) e\left [ q^i \left ( \sum_{j=1}^n c_j \right)^{\alpha - i } \right ] , \label{eq : prodmoments_2}\end{aligned}\ ] ] where for the inequality we used lemma [ l.alpha_moments ] ( @xmath577 ) and the inequality @xmath578 for any @xmath232 and @xmath246 . note that all the expectations involving @xmath0 in and are finite since @xmath118 < \infty$ ] for all @xmath173 by assumption . next , observe that all the other expectations are of the form @xmath579 $ ] for @xmath580 . to see that these are finite use hlder s inequality with @xmath581 and @xmath582 to obtain @xmath583 & \leq \left\|\left\|\left ( \sum_{j=1}^n c_j \right)^{\alpha - i } \right\|\right\|_q \|\|q^i\|\|_r \\ & = \left ( e\left [ \left ( \sum_{j=1}^n c_j \right)^{\alpha } \right ] \right)^{1/q } \left ( e \left [ q^\alpha \right ] \right)^{1/r } < \infty.\end{aligned}\ ] ] we would like to thank an anonymous referee and matthias meiners for their helpful comments . p. jagers and u. rsler . stochastic fixed points involving the maximum . in _ mathematics and computer science _ * iii * ( m. drmota , p. flajolet , d. gardy and b. gittenberger , eds . ) , 325 - 338 . birkhuser , basel , 2004 .	we extend goldie s ( 1991 ) implicit renewal theorem to enable the analysis of recursions on weighted branching trees . we illustrate the developed method by deriving the power tail asymptotics of the distributions of the solutions @xmath0 to @xmath1 and similar recursions , where @xmath2 is a nonnegative random vector with @xmath3 , and @xmath4 are iid copies of @xmath0 , independent of @xmath2 ; here @xmath5 denotes the maximum operator .
time - resolved faraday rotation ( trfr ) measurements have proved to be a powerful tool for the study of dynamical spin phenomena in semiconductors @xcite . in its most basic form , trfr is an optical pump - probe technique that can be used to monitor the temporal evolution of an optically - injected spin polarization and determine the electron g - factor and spin lifetime in materials with advantageous selection rules@xcite . variations on trfr measurements , including spin - drag and larmor magnetometry , enable the study of spin transport properties @xcite , spin - orbit fields@xcite , dynamic nuclear polarization@xcite , and many other phenomena . when the spin polarization lifetime @xmath0 is on the order of or larger than the laser repetition time , the spins excited by previous pump pulses make non - negligible contributions to the faraday rotation signal , resulting in resonant spin amplification ( rsa)@xcite . in samples with long lifetimes , rsa can provide a more accurate way to measure @xmath0 than trfr@xcite . in this paper , we investigate another effect of previous pump pulses which results in an effective phase shift that depends on the applied magnetic field . this phase shift must be accounted for in any situation in which the contribution to trfr from previous pulses is expected to be important , and can provide another method for determining the spin lifetime . in a typical trfr measurement , a circularly polarized laser pulse excites a spin polarization parallel to the direction of laser propagation , which we take to be @xmath1 . after some time delay , a linearly polarized probe pulse interrogates the system by measuring faraday rotation , which is proportional to the spin polarization along @xmath1 . typically , an ultrafast laser is used to generate the pulses and the time delay between pulses is controlled with a mechanical delay line . in this way , the shortest accessible time scale is set by the pulse time and the precision with which the delay line can be positioned . the spin polarization lifetime can be found by measuring the timescale of the decay of faraday rotation as @xmath2 , the delay between pump and probe pulses , is swept . for a measurement of the g - factor , a magnetic field is applied in the voigt geometry along @xmath3 . spins will then precess in the @xmath4-@xmath1 plane at their larmor frequency @xmath5 , where g is the electron g factor , @xmath6 is the bohr magneton , and @xmath7 the reduced planck constant . in this geometry , a single pump pulse gives rise to a spin polarization of : @xmath8\theta(\delta t)\ ] ] where a is the initial spin polarization and @xmath9 is the heaviside step function which enforces causality in the expression . when the spin lifetime is such that an appreciable spin polarization remains after the laser repetition time , spins due to previous pump pulses must be taken into account as well . it is convenient at this point to use phasor notation , @xmath10 where @xmath11 is a complex spin phasor . in this notation , @xmath12 is understood to be the real part of @xmath11 , re@xmath13 and @xmath14 becomes the imaginary part of @xmath11 , im@xmath13 . in this case , each pump pulse contributes : @xmath15\theta(\delta t)\ ] ] to the total spin . summing over previous pulses , the total @xmath11 becomes : @xmath16\nonumber\\ & \times&\theta(\delta t + nt_{rep})\end{aligned}\ ] ] where @xmath17 is the laser repetition time . figure 1 ( a ) shows experimental data taken on a commercially purchased sample consisting of a 2 @xmath18 m layer of n - gaas doped at @xmath19 @xmath20 atop a semi - insulating gaas substrate . measurements were performed at 10 k as a function of external magnetic field @xmath21 and pump - probe delay time @xmath2 . a mode - locked ti : sapphire laser tuned to @xmath22 nm with a pulse time of approximately 3 ps and a repetition time @xmath17 of 13.2 ns was used . pump and probe beam powers were measured to be 500 @xmath18w and 75 @xmath18w , respectively , with beam diameters of approximately 35 @xmath18 m each . here , the magnitude of the signal shown is in millivolts as measured on a lock - in amplifier , which is proportional to @xmath23 , the faraday rotation angle , but taken to be an arbitrary unit . the data set was then fit simultaneously to extract the electron g factor and spin lifetime . the data are fit to the real part of eq . ( 3 ) , with a term added which takes into account the contribution from optically excited carriers before recombination , which decays rapidly . in total , the fit equation used is : @xmath24\\ & \times\theta(\delta t + nt_{rep})\big\}\\ + a_o\cos&(\omega_l\delta t)\exp[\delta t / t_o]\times\theta(\delta t ) \end{split}\ ] ] the fit finds @xmath25@xcite , @xmath26 ns , and optical spin lifetime @xmath27 ps . the ratio of faraday rotation from long - lived spins to spins from optical carriers , @xmath28 , is found to be @xmath29 . figure 1 ( b ) shows computed data using the results of the fit . both plates in fig . 1 have been normalized using the same normalization factor . measurements which rely on resonant spin amplification take advantage of the collective effect of spin polarization resulting from previous pulses . typically , the external magnetic field is scanned instead of the delay time . when the magnetic field reaches a value at which spins precess an integer number of times in the laser repetition time , spins resulting from all previous pump pulses are aligned along @xmath1 when the next pulse arrives , resulting in a large spin polarization at @xmath30 . when a half - integer number of precessions occur , spins antialign and a smaller polarization is seen . this is shown diagramatically in fig . 2 plates ( a ) and ( b ) . these figures show the @xmath4 and @xmath1 components of spin polarization at @xmath30 , immediately after a pump pulse arrives , normalized to the initial spin polarization of a single pulse @xmath31 . contributions to the total polarization from five additional pump pulses are shown in blue , along with the total spin polarization in red . here , data was synthetically generated using @xmath32 , @xmath33 ps and @xmath34 . the degree to which the spin polarization at @xmath35 exceeds a when a resonance condition is met is determined by the lifetime of the spins ; longer lifetimes result in higher and sharper resonance peaks as measured by scanning the external magnetic field . the g - factor and lifetime can be found by fitting this data to eq . 3 . this method can give better fits of the lifetime than standard trfr scans when it is large compared to the accessible delay times@xcite . by pulling terms out of the sum in eq . ( 3 ) which do not depend on @xmath36 and enforcing @xmath37 , @xmath11 becomes : @xmath38\nonumber\\ \times&&\sum_{n\geq 0}\exp[(i\omega_l - 1/t_2^)nt_{rep}]\end{aligned}\ ] ] where now the sum is in the form of a converging geometric series with common ratio @xmath39 $ ] . the sum can therefore be evaluated : @xmath40 with @xmath41 and @xmath42 . this can be rewritten as a complex number @xmath43 whose amplitude @xmath44 and phase @xmath45 are functions of the external magnetic field and spin lifetime : @xmath46 by taking the real part of eq . ( 5 ) using these expressions , @xmath12 becomes : @xmath47 these expressions are functionally equivalent to the closed - form expression found in ref . 17 , but are valid for @xmath48 instead of @xmath49 , and are written in a form that makes explicit the fact that previous pulses contribute an effective phase shift @xmath45 to trfr measurements in addition to modifying the amplitude of the spin polarization . this phase shift is equivalent to @xmath50 . in an effect analogous to peaks in rsa amplitude , there exist values of the external field where the sum of the spin polarization due to previous pulses will result in a maximum effective phase . at these field values , the spin polarization immediately preceding the arrival of a pump pulse @xmath51 , will lay in the sample plane along @xmath4@xcite . this condition is shown in fig . 2 plates ( c ) and ( d ) . in these figures , the green vector indicates the sum of previous pulses excluding the most recent pulse . again , the red and blue arrows show the total spin polarization and the first five terms in the sum , respectively . at field values which satisfy this condition , the effective phase @xmath45 in eq . ( 9 ) is maximized . figure 3 shows plots of @xmath44 and @xmath45 versus external magnetic field calculated using equations ( 7 ) and ( 8) with a range of lifetimes . individual delay scans taken from the experimental data in fig . 1 ( a ) were fit to eq . ( 9 ) to extract @xmath44 and @xmath45 , and the results are plotted as black squares . calculated values of @xmath44 and @xmath45 using the experimentally determined lifetime and g factor from the simultaneous fit above are shown as a black line to highlight the close agreement with the model . as the spin lifetime increases , the maximum and minimum observed phase angles increase in magnitude and are found closer to the external magnetic field values corresponding to an rsa peak . as the magnetic field is swept , the total spin polarization at @xmath30 , depicted by the red arrows in fig . 2 , traces out a circular path in the @xmath4-@xmath1 plane . this path is shown in fig . 4 for the lifetimes used previously in fig . 3 . as the lifetime grows , the radius of the circle increases , and in the limit that @xmath0 goes to infinity the radius of this circle diverges . accordingly , the maximum and minimum phase angles approach @xmath52 and @xmath53 , respectively . the rate at which @xmath54 moves around the circle as the magnetic field is swept is not constant , but rather exhibits a rapid shift in phase as the field value moves past an rsa peak and then a slow shift in phase through the rsa minimum . since arctangent grows monotonically with its argument , the extrema of @xmath45 can be found by extremizing the argument of the arctangent in eq . setting its derivative equal to zero and solving gives the values of @xmath55 which extremize @xmath45 : @xmath56 as @xmath0 grows ( causing @xmath57 to shrink with constant @xmath17 ) , the values of @xmath58 grow closer to @xmath59 , which correspond to peaks in @xmath44 . substituting eq . ( 10 ) with @xmath36 = 0 into eq . ( 8) gives the maximum phase angle as a function of @xmath57 , given by : @xmath60 which is plotted in fig . @xmath61 has a monotonic dependence on the spin lifetime , and can therefore be used to measure @xmath0 when it is sufficiently large compared to the laser repetition time . such a measurement could be performed by first setting @xmath2 to a small negative value and tuning the magnetic field so that the faraday rotation disappears . at this magnetic field , the polarization due to previous pulses will lay in the sample plane , and correspond to a condition of maximum or minimum phase . then , a delay scan would be performed and fit to extract @xmath45 , which , by inverting eq . ( 11 ) , can be used to find @xmath0 . in the experimental data shown in fig . 3 , the maximum observed phase angle was @xmath62 , which gives a lifetime of 10.4 ns , close to the value obtained above . trfr can be used to measure the local magnetic field when the g factor is known by a process dubbed larmor magnetometry@xcite . traditionally , a delay scan is performed over at least a few periods of precession and a fit to the precession frequency extracts the magnetic field present . the speed with which such a measurement can be performed is set by the time it takes to complete the delay scan , typically on the order of 10 to 100 seconds . measurements have also been performed by maintaining a fixed delay and continuously monitoring @xmath23 @xcite . this technique does not generally allow for a good determination of the magnetic field because @xmath23 is not single - valued when the magnetic field changes , and when near a local extrema , @xmath23 changes slowly with magnetic field , reducing or eliminating the ability to measure changes in the magnetic field . additionally , shifts in laser power or slow drifts in pump - probe overlap are indistinguishable from changes in magnetic field . these limitations can be avoided if instead the delay time is continuously adjusted to follow a zero crossing of @xmath23 of known order . the position at which a zero crossing occurs is unaffected by changes in laser power or pump - probe overlap , and , with care , a zero crossing can be picked whose delay time has a one - to - one correspondence with the magnetic field strength over a range of values . a measurement can then be completed in the time it takes to adjust the position of the mechanical delay line once and take a measurement of @xmath23 , which can take place in approximately one second . a zero crossing will occur when @xmath63 , or from eq . ( 9 ) , the @xmath64 crossing occurs when @xmath65 . the presence of @xmath45 in this expression complicates this measurement . figure 5 ( b ) shows the delay time at which the @xmath66 crossing occurs versus magnetic field for a range of lifetimes . again , colors refer to the legend in fig . when the spin lifetime is long , the change in phase due to previous pulses is large , and results in a series of sharp steps in the delay time at which a zero crossing occurs that correspond with rsa peaks as the magnetic field strength is scanned . it can be shown that if a zero crossing occurs at a delay time that satisfies @xmath67 , there will always be a one - to - one relationship between the local magnetic field strength and the delay time at which the crossing occurs . the inset in fig . 5 ( b ) shows what happens when this relation is violated for a subset of the spin lifetimes defined in fig . , the @xmath68 crossing has been chosen . in this situation , for long spin lifetimes , the local magnetic field can no longer be uniquely determined from the delay time of the zero crossing . in order to measure the local magnetic field using this technique , the spin lifetime must now be known , which adds uncertainty to the measurement of the magnetic field . additionally , it must be known exactly which crossing is being followed . this could be ascertained by performing a full delay scan at a fixed , known magnetic field . in the preceding analysis , it is assumed that the contributions from previous pulses decay exponentially with the time elapsed since the pump pulse arrived at the sample . in situations where there is electron drift or diffusion , each term in the sum in eq . ( 3 ) must now include a factor which takes this into account . in general , this will depend on many parameters , including but not limited to beam size and shape , mobility , in - plane electric fields , and diffusion constant . equation ( 5 ) becomes : @xmath69\nonumber\\ \times & & f_n(\delta t)\theta(\delta t + nt_{rep})\big\}\end{aligned}\ ] ] where @xmath70 takes into account changes in overlap . since @xmath70 depends on @xmath36 and @xmath2 , it can not be factored out of the sum . as a result , this expression can not be simplified to the form of eq . ( 9 ) , and the results of a fit to that functional form will no longer directly give the precession frequency or lifetime . care must be taken to characterize and correct for this effect whenever the in - plane motion of electrons is expected to be important . also , we assume that spins resulting from each pump pulse are not affected by subsequent pulses . if the optical power is sufficient that the number of optical carriers is on the order of or greater than the number of dopant carriers within the beam area , this is no longer a good assumption@xcite , and a factor which accounts for this must be included in the sum . this factor will not depend on @xmath2 , however , so that eq . ( 9 ) may still be valid , though the expressions for @xmath44 and @xmath45 must be revised . this effect can be minimized by decreasing the pump and probe powers . in conclusion , we have recast expressions used in the context of rsa measurements in a form that highlights an effective phase shift due to the collective action of spin polarization from previous pulses in trfr measurements . we experimentally measure this phase shift in n - gaas and find that it is in good agreement with the model . in situations with long spin lifetimes , the phase shift can become a significant factor in trfr measurements . we suggest a measurement which utilizes the phase to provide another means of accessing the spin lifetime when it is long compared to the laser repetition time in which @xmath2 is scanned instead of the external magnetic field . this work was supported by the national science foundation materials research science and engineering center program dmr-1120923 and under grant no . eccs-0844908 , the office of naval research , the air force office of scientific research , and the defense threat reduction agency basic research award no . hdtra1 - 13 - 1 - 0013 . b. m. norman , c. j. trowbridge , j. stephens , a. c. gossard , d. d. awschalom , and v. sih , phys . b 82 * , 081304(r ) ( 2010 ) . s. kuhlen , k. schmalbuch , m. hagedorn , p. schlammes , m. patt , m. lepsa , g. gntherodt , and b. beschoten , phys . rev 109 * , 146603 ( 2012 ) . b. beschoten , in _ lecture manuscripts of the 36th spring school of the institute of solid state researchmagnetism goes nano . " _ , ed . by s. blgel , t. brckel , and c. m. schneider , reihe matter and materials ( forschungszentrum juelich gmbh , 2005 ) , vol . 26 , p. e.7.1 .	time - resolved faraday rotation measurements have proved transformative in the investigation of spin dynamics in semiconductors . in materials with spin lifetimes which are on the order of , or greater than , the laser repetition time , the collective effect of spin polarization due to the whole pump pulse train becomes important . here , we discuss a relative phase shift which results from these spins . we derive and experimentally validate a closed - form expression which describes this phase shift , and characterize it throughout parameter space . a spin lifetime measurement based on this phase shift is described , and we discuss situations in which the model used must be augmented to be applicable .
on deep pictures of the magellanic clouds ( mcs ) taken in light , meaburn and collaborators ( cf . meaburn 1980 ) recognized ten huge ` bubbles ' with diameters of 600@xmath5@xmath7pc , the so - called supergiant shells ( sgss or supershells ) . these huge shells , according to goudis & meaburn ( 1978 ) , build a new group different from the smaller ( @xmath8pc ) giant shells ( gss or superbubbles ) . these structures need very effective creation mechanisms and collisions of high velocity clouds ( hvcs ) with the disk of the galaxy or stochastic self - propagating star formation ( sspsf ) have been proposed . both might explain the ring of regions and the hole in the layer , as observed in the case of lmc@xmath54 . according to the sspsf model , star formation would ` eat ' its way from the initial point to all directions through the interstellar medium , creating a big cavity and a thick outer shell of neutral hydrogen ionized at the inner edge by the early - type stars ( o - b2 ) . thus one should see a clear age gradient from the centre to the rim . this contribution will give first results for the sgs smc@xmath51 , summarize the study at the east side of the lmc with a focus on lmc@xmath52 and n@xmath5171 , and ends with a short update ( cf . braun 1998 ; braun et al . 1997 ) of knowledge about the stellar content of lmc@xmath54 and its implications on the creation mechanism . smc@xmath51 ( @xmath9 dem@xmath5s@xmath5167 ) is the only supergiant shell ( sgs ) detected by meaburn and collaborators ( see e.g. meaburn 1980 , fig . 3 ) in the small magellanic cloud . it has a diameter of about 600 pc , as indicated by the rough dashed boundaries in fig . 1 . its appearance is not as regular as that of lmc@xmath54 , with pronounced emission from south ( incl . n@xmath10 @xmath9 dem@xmath5s@xmath5166 ) to east , in the northwest , and with a double ` rim ' in the north . from the east rim the emission extends toward ne via n@xmath589 ( @xmath9 dem@xmath5s@xmath5164 ) to n@xmath588 ( @xmath9 dem@xmath5s@xmath5161 ) . the presented photometry has been taken at the @xmath11 m danish telescope at eso observatory on cerro la silla with dfosc and the 2k@xmath12pix@xmath13 loral ccd ( w7 chip ; @xmath14 field of view ) by me in january 1998 ( c , s ) , and by martin altmann in november 1999 ( n , w ; only @xmath15 ) . the cmds with fitted isochrones in fig . 2 reveal a small young population of @xmath16myr and a second population of at least @xmath17myr ( see braun 2001 ; braun et al . 2001b for more details ) . * fig . 1 . * mosaic of the central smc wing region ( shapley 1940 ) out of @xmath18 charts of the hodge & wright ( 1977 ) atlas . the four fields ( north , centre , south , west ) of the presented dataset in this supergiant shell are outlined , with the rim of smc@xmath51 roughly indicated by the dashes . * * cmds of the entire smc@xmath51 region ( see fig . 1 ) with the isochrones of the geneva group ( charbonnel et al . 1993 ) for smc metallicity ( @xmath19 ) of logarithmic ages @xmath20 ) \in \ { 6.8 , 7.0 , 7.3 , 7.7 \}$ ] and applied colour excess of @xmath21mag . the @xmath22 diagram contains @xmath23 data points , the @xmath24 a total of @xmath25 . in 1996 i observed five regions located at the lmc east side with 10 ccd positions in @xmath4 passbands with the @xmath26 m danish telescope . these regions ( see fig . 3 ) are from north to south : the giant shell n@xmath570 ( region a@xmath27 , cf . braun 1998 ) , a strip inside sgs lmc@xmath52 ( b@xmath28 ) , an outer field east of lmc@xmath52 ( c ) , the region around n@xmath5171 ( d ) , and the southern part of n@xmath5214 ( e@xmath27 ) . here i will concentrate on lmc@xmath52 and n@xmath5171 data ( see braun 2001 ; braun et al . 2001a for further information ) . * sketch of the large - scale features ( see mizuno et al . 2001 for a detailed map of the dark cloud , dc ) and ob associations at the lmc east side . additionally , the 5 regions ( a - e ) out of 10 ccd positions are indicated . the position of the rotation centre , rc , and the movement of the lmc in the galactic halo are indicated by the direction of rotation and proper motion . the resulting bow - shock , driven by the sum of motion and rotation , is @xmath29 km s@xmath30 ( see de boer et al . 1998 ; de boer 1998 ) . * * cmds of the entire strip inside lmc@xmath52 ( b@xmath28 ) with geneva isochrones ( see schaerer et al . 1993 ; @xmath31 ) of logarithmic ages @xmath20 ) \in \ { 6.9 , 7.2 , 7.5 \}$ ] and applied of 0.68@xmath32mag . * fig . 5 . * cmds of the outer field east of lmc@xmath52 ( c ) with geneva isochrones ( schaerer et al . 1993 ; @xmath31 ) of logarithmic ages @xmath20 ) = 7.7 $ ] and applied of 0.65@xmath32mag . * fig . 6 . * cmds of the region around n@xmath5171 ( d ) with geneva isochrones ( schaerer et al . 1993 ; @xmath31 ) of logarithmic ages @xmath20 ) = 7.25 $ ] and applied of 0.96@xmath32mag . one should note that the isochrone has only been fitted to the youngest and less reddened stars of the @xmath22 diagram . 4 and 5 show the cmds of the entire strip inside lmc@xmath52 and of the outer field , respectively . while the cmds of the outer field only show a stellar population of about @xmath33myr and older , the cmds of the strip containing n@xmath5164 , ngc@xmath52100 , and ngc@xmath52102 reveal a mixture of young populations ( present in all four fields ) , covering the interval from 8 to 32 myr . while the two southern fields are similar , the two northern show differences in the age structure by a component younger than 8 myr concentrated on n@xmath5164 ( b@xmath34 ) and a 16 myr component of ngc@xmath52100 . for n@xmath5171 , the youngest population discovered in the cmds ( fig . 6 ) has an age of about 18 myr . this region was selected at the locus of an x - ray shadowing region ( cf . blondiau et al . 1997 ) and shows extraordinary reddening with a large spread . the colour extinctions ( @xmath35/[mag ] ) derived from the reddening free parameter q ( peak of the distribution with fwhm in (@xmath5) ) and from the isochrone fit ( in [ ] ) are : 0.075 ( 0.08 ) [ 0.11 ] for n@xmath570 ( a@xmath34 ) , 0.14 ( 0.08 ) [ 0.21 ] for the outer field east of lmc@xmath52 , 0.17 ( 0.11 ) [ 0.19 ] for n@xmath5214 , 0.185 ( 0.16 ) [ 0.22 ] for the lmc@xmath52 strip , and * 0.29 ( 0.24 ) [ 0.31 ] * for the * n@xmath5171 * region . one should note that the peak of the distribution is shifted from the mean and that the isochrone fit is based on the youngest and for n@xmath5214 not numerous population , which may have a different reddening than most other stars ( maybe due to unrevealed 3d structure ) . our first @xmath15 photometry inside lmc@xmath54 ( a j-shaped region , see fig . 7 ) taken in 1993 with the @xmath36 m dutch telescope ( braun et al . 1997 ; braun 1998 ) has shown , as given in fig . 8b , that the central superassociation , lh@xmath577 , has an age of @xmath37myr . this population is present in the entire interior of this supergiant shell and at the inner rim region ( e.g. lh@xmath563 ) . the associations at the rim tend to show age gradients with younger ages for associations more distant from the inner rim of filaments ( see gouliermis et al . 2001 ; table 6 of braun et al . 1997 ) . thus , this huge coeval population gave rise to the need for a large - scale trigger , like an infalling cloud or a large shock front created by the movement of the lmc through the galactic halo ( de boer et al . 1998 ; de boer 1998 ) , while models with a central trigger or a propagation could be ruled out . sspsf is therefore only valid on small scales ( up to @xmath38pc ) as it is visible on the rim , while the creation mechanism of lmc@xmath54 has its origins on larger ( @xmath39kpc ) scales . * * a ) image of supergiant shell lmc@xmath54 [ left panel ] made from a scan of a photographic plate taken with the curtis schmidt telescope at cerro tololo ( kennicutt & hodge 1986 ) . the locations of some stellar associations and the borders of the two datasets ( j and c , respectively ) are marked . b ) sketch of selected structures [ right panel ] near lmc@xmath54 ( adapted from fig . 2 of efremov & elmegreen 1998 ) . * cmds of lmc@xmath54 with geneva isochrones ( schaerer et al . 1993 ; @xmath31 ) of logarithmic ages @xmath20 ) \in \ { 7.05 , 7.5 , 8.0 \}$ ] and extinction correction of @xmath40mag . + a ) cmd with @xmath41 data points of the central data set ( c , see fig . 7a ) of lmc@xmath54 [ left panel ] . + b ) cmd with @xmath42 d. p. of the lmc@xmath54 j dataset ( braun et al . 1997 ) without fields 11 - 16 [ right panel ] . + in panel a ) the crosses near @xmath43mag are six a supergiants also marked in fig . nevertheless , after these results further models with central trigger had been proposed by efremov & elmegreen ( 1998 ) , describing three stellar arcs as triggered by central clusters ( as given by the bows and boxes in fig . additionally , efremov & fargion ( 2000 ) presented a scenario , by which cluster arcs ( now even four in the lmc@xmath54 region and vicinity ) are each pressurized by the collimated and precessing beam of a @xmath44-ray burst / soft gamma repeater ( originating in ngc@xmath51978 ) . to further constrain the triggering mechanism , we observed three central fields ( c ) of lmc@xmath54 ( see fig . 7a and braun et al . 2000 ) in @xmath45 and @xmath18 . the resulting colour - magnitude diagram ( fig . 8a ) has the same morphology as the one of the j dataset excluding the overlapping region ( fig . the two clusters ( hs@xmath5343 and kmhk@xmath51000 ) , which may have triggered star formation after efremov & elmegreen , are much too old ( 0.1 gyr and 0.3 gyr , respectively ) to have stimulated further star formation in the birth cloud of lh@xmath577 . the triggering of different arcs does neither fit to a homogeneous coeval population nor to the morphology of lmc@xmath54 in and . with the stringent results of the good - quality ) shows good agreement . all three datasets are available electronically , see cds j / a+a/328/167 , j / aj/116/1275 , and aiub ftp account , linked from the www pages given above ( braun et al . ] photometry available , it seems to be inevitable that lmc@xmath54 has been driven and ionized by the massive stars of the coeval population inside the sgs , with about 5000 stars already exploded in supernovae of type ii . thus , the gas had been driven out of the central region in a turbulent fashion , causing an outbreak at the rear side and a component approaching with 11 km / s ( domgrgen et al . the star burst 12 myr ago has been most probably driven by the bow - shock , resulting in a compression zone . its location is currently , due to the lmc rotation ( see fig . 3 ) , at the south eastern edge and thus may have caused the co complex south of 30@xmath5doradus ( see mizuno et al . 2001 ) . braun , j.m . 1998 , `` the magellanic clouds and other dwarf galaxies '' , eds t. richtler , j.m . braun , shaker verlag , aachen , isbn 3 - 8265 - 4457 - 9 , pp . 115 - 120 + ( see http://www.astro.uni-bonn.de/@xmath46webstw/ws98/jmb_pt.html ) mizuno , n. , yamaguchi , r. , onishi , t. , mizuno , a. , fukui , y. , 2001 , `` dwarf galaxies and their environment '' , eds k.s . de boer , r.j . dettmar , u. klein , shaker verlag , aachen , in press ( these proceedings )	in this contribution i will present the current status of our project of stellar population analyses and spatial information of both magellanic clouds ( mcs ) . the magellanic clouds are suitable laboratories and testing ground for theoretical models of star formation . with distance moduli of 18.5 and @xmath0mag for the lmc and smc , respectively , and small galactic extinction , their stellar content can be studied in detail from the most massive stars of the youngest populations ( @xmath1myr ) connected to emission down to the low mass end of about @xmath2 of a solar mass . especially the lmc with its large size and small depth ( @xmath3pc ) is a prefered target to constrain star formation mechanisms . based on broad - band photometry ( @xmath4 ) i present results for the supergiant shell ( sgs ) smc@xmath51 , some regions at the lmc east side incl . lmc@xmath52 showing different overlapping young populations and the region around n@xmath5171 with its large and varying colour excess , and lmc@xmath54 . this best studied sgs shows a coeval population aged about @xmath6myr with little age spread and no correlation to distance from lmc@xmath54 s centre . i will show that the available data are not compatible with many of the proposed scenarios like sspsf or a central trigger ( like a cluster or grb ) , while a large - scale trigger like the bow - shock of the rotating lmc can do the job . ps . @plainmkbothoddheadoddfoot``dwarf galaxies and their environment '' ; bad honnef , germany , 23 - 27 january 2001 ; eds . k.s . de boer , r .- j.dettmar , u. klein ; shaker verlagevenheadevenfootoddfoot
we first derive the ekpyrotic spectrum in a very naive ` time delay ' approach which totally neglects cosmological expansion , gravity and all moduli other than @xmath0 . we also ignore the crucial element of reversal from contraction to expansion , which will be a critical aspect of the discussion in sections v and vi . as we shall see , despite the fact that the time delay argument ignores these very important features , it nevertheless comes close to matching our final answer . thus , as is the case in inflation , the time delay argument turns out to be a convenient heuristic even though it is neither physically rigorous nor numerically accurate . we assume that at large positive @xmath0 the effective potential governing the evolution of @xmath0 takes the form v= - v_0 e^-c , [ eq : pot ] where @xmath11 and @xmath12 are positive constants . the ekpyrotic scenario starts at large positive @xmath0 , in a state of nearly zero energy , and rolls towards negative values . as @xmath0 becomes increasingly negative and the branes come close together , @xmath7 must turn upwards and approach zero . as discussed in ref . [ nonsing ] , for example , this is the behavior to be expected in m - theory because the string coupling constant vanishes as the outer branes collide . ( a similar constraint applies in the bulk brane collision , as discussed in ref . 1 . ) generally , in brane world models , the separation of the two boundary branes can be described by a canonically normalized , minimally coupled scalar field @xmath0 ( the ` radion ' ) . as the brane separation @xmath13 goes to zero , one can ignore the effect of the bulk ` warp factor ' and one obtains the kaluza - klein result in which @xmath14 as @xmath15 . ( here and below , @xmath16 is the four dimensional reduced planck mass . ) in standard kaluza - klein theory , the same dependence holds at large @xmath0 , so the range of @xmath0 is @xmath17 . but the presence of a bulk warp factor alters this . in many models , when the outer - brane distance @xmath13 tends to infinity , @xmath0 tends to a finite value , which may be taken to be zero . for example , for a pair of positive and negative tension boundary branes with a bulk anti - de sitter space , the inter - brane distance is @xmath18 ln(coth(@xmath19 where @xmath20 is the ads radius @xcite . a suitable ekpyrotic potential would then behave as @xmath21 , at small @xmath0 , and we would be interested in starting the system in a state such that @xmath22 at large negative times . in models with a bulk brane , such as those considered in ref . 1 , the location of the bulk brane would again be described by a scalar field , but this time its range would be finite . the exponential model ( [ eq : pot ] ) is useful , since it is mathematically tractable . however , one should remember , in general it should only be expected to apply over some finite range of @xmath0 . we now review how the time delay formalism can be applied to our example . the classical solution is given by solving @xmath23 . for @xmath7 given in ( [ eq : pot ] ) , we obtain @xmath24 where the time @xmath25 is large and negative at large @xmath0 . we follow the evolution to some finite , small negative @xmath25 , at which point the modes of interest are ` frozen in ' . for small fluctuations , expanding in plane waves @xmath26 , one finds @xmath27 ( using the solution eq . ( [ eq1 ] ) above ) . curiously , even though the background here is nearly static , this is the same equation as that governing a massless field @xmath28 in a de sitter background ( with @xmath29 , with @xmath6 the scale factor and @xmath25 conformal time ) . as in that case , starting from an initial minkowski vacuum one generates a long wavelength spectrum of scale invariant fluctuations . we assume that the quantum fluctuation @xmath30 starts in the minkowski vacuum , @xmath31 , so the amplitude of normalized modes with wavenumber @xmath32 is @xmath33 . consider a mode where @xmath34 , where @xmath35 is the initial time . the mode oscillates at fixed amplitude @xmath36 until @xmath25 increases to @xmath37 , when the destabilizing @xmath38 term in ( [ eq2 ] ) begins to dominate over the stabilizing @xmath39 term . one then has @xmath40 , with growing solution @xmath41 . hence , the final amplitude for the fluctuation , as @xmath25 approaches zero , is @xmath42 . the precise result for the long wavelength spectrum is \|_\|^2= 12 k^3 t^2 , -kt < < 1 , [ eq : noexp ] which is scale - invariant as claimed . a naive estimate of the final density perturbation amplitude runs as follows . the scalar field fluctuations give rise to a time delay in the moment of collision . neglecting gravitational effects , the time delay between collisions relative to the mean collision time is @xmath43 . the net density perturbation after collision is then just @xmath44 where @xmath45 is the hubble parameter at the time of collision . despite the deficiencies of this time delay approach , this answer is quite close to the result we shall eventually derive , including gravitational back - reaction and matching across the bounce , which we give in equation ( [ eq : finalampf ] ) below . the first step in improving the above treatment is to include gravitational backreaction in the initial , contracting phase . for this purpose , we focus on the gravitational backreaction corrections to eq . ( [ eq : noexp ] ) . for exponential potentials , there exist analytic scaling solutions to the friedmann - robertson - walker equations which at large negative @xmath25 approximate the assumed initial conditions in the ekpyrotic setup , and which allow an analytic treatment @xcite . for the exponential potential considered as an approximation over some range of @xmath0 in the ekpyrotic scenario , one has the background solution a(t)= ( -t)^p , _ 0(t)=2c ln ( -m t ) , [ eq : back ] where @xmath46 is the scale factor , @xmath25 is proper frw time , taking negative values , and p= 2c^2 m_pl^2 , m^2= v_0 m_pl^2 p(1 - 3p ) , [ eq : params ] and @xmath47 is the reduced planck mass . some useful quantities are : h = pt , _ 0= , v =- p(1 - 3p ) m_pl^2 t^2 , v _ , = -2t^2 ( 1- 3p ) , [ eq : quants ] where dot denotes @xmath48 . in ekpyrotic models , we are interested in potentials in which the evolution is very slow or , equivalently , for which @xmath49 . ( parenthetically , note that the same solution , for @xmath50 , describes a universe undergoing power law inflation ) . the conformal time is = - _ t^0 dt a(t ) = - ( -t)^1-p ( 1-p ) = t a ( 1-p ) , [ eq : conft ] running from @xmath51 to @xmath52 as @xmath25 does , although as mentioned we shall have to alter the form of the potential as @xmath25 and @xmath53 approach zero , to avoid its diverging to minus infinity . derivatives with respect to @xmath25 shall , as above , be denoted by dots and henceforth , derivatives with respect to @xmath53 shall be denoted by primes . mathematically , newtonian gauge is convenient for studying scalar field perturbations since the linearized equations reduce to a single second order differential equation for the newtonian potential @xmath3 ( for a derivation see e.g. ref . [ bgt ] ) , + 2 aa + k^2 + 2 _ 0(h _0) = 0 , [ eq : phieq ] where @xmath54 , and @xmath55 . here , and below , the wavenumber dependence of all perturbation quantities is not shown explicitly . for @xmath3 one should read @xmath56 and so on . we now eliminate the first derivative term by setting @xmath57 , obtaining u-aa u + k^2 u + 2 _ 0(h_0) u=0 . [ eq : neweq ] substituting the above scaling solution , we obtain u=-k^2 u + p(1-p)^2 ^2 u , [ eq : feq ] which is a form of bessel s equation , with corresponding order @xmath58 . note that our notation for the variable @xmath59 , and for the variable @xmath60 introduced later , matches mukhanov s original notation @xcite . the initial conditions are that the scalar field fluctuations should be in the minkowski vacuum state as @xmath61 : in newtonian gauge , two constraints determine @xmath3 and @xmath62 in terms of @xmath63 and @xmath64 : & = & 2 m_pl^2 _ 0 , & = & - 2 m_pl^2 _ 0 . [ eq : phicons ] at large @xmath65 , for @xmath66 , which is the incoming minkowski vacuum , one finds ~i 2 m_pl k^32 t e^- i k [ eq : phiin ] which vanishes at early times , consistent with our ekpyrotic initial condition that the space is asymptotically minkowski in the past . neglecting an irrelevant phase factor , the solution for @xmath59 is then u= ( -k ) ^12 h_^1(-k ) , [ eq : fsol ] where @xmath67 is a hankel function and @xmath68 was given above . we follow this solution forward to conformal times at which @xmath69 , when the modes become frozen in . converting back to @xmath3 , and using the small argument expansion for the hankel function , again neglecting phase factors we find ~ 2m_pl k^-(1 + ) ( -)^-2(1+o(p ) ) . [ eq : phisol ] since we are interested in @xmath70 , we henceforth neglect corrections of order @xmath71 to the numerical coefficient . but we keep the @xmath71 dependence in the scaling with wavelength and time for comparison with later results . now we can convert back to the scalar field @xmath63 , using the formulae ( [ eq : phicons ] ) given above , and obtain : = 2 m_pl^2 _ 0 ~2 ^ -1/2 k^-(1 + ) ( -t)^-(1+p ) . [ eq : dps ] where we used @xmath72 . the resulting power spectrum of fluctuations is : \|_\|^2 ~ k^-2(1 + ) ( -t)^-2(1+p ) = \|_\|^2^(0 ) 1 k^2p/(1-p ) ( -t)^2p , [ eq : dpsa ] where @xmath73 is the result obtained in eq . ( [ eq : noexp ] ) ignoring gravitational backreaction . recalling that the regime of interest for the ekpyrotic scenario is small @xmath71 , we see that we obtain the same answer as in eq . ( [ eq : noexp ] ) up to small corrections . the backreaction produces a slight reddening of the power spectrum , presumably because long wavelength perturbations were generated sooner and had longer to self - gravitate . however , since in the ekpyrotic scenario we have @xmath70 , these corrections to the spectral index are actually smaller than corrections arising from the nontrivial kinetic term for @xmath0 , which were computed in ref . the result shows that backreaction and metric fluctuations have an insignificant effect on the density perturbations developed during the initial , contracting phase of the ekpyrotic universe . rather than solve for the evolution of the newtonian potential in eq . ( [ eq : phieq ] ) , a common procedure used for inflationary models is to track the curvature perturbation on spatial slices which are comoving with the matter . as we shall discuss , this variable is in fact insensitive to the growing mode density perturbation in the contracting phase . following the notation of mukhanov _ @xcite , we denote the curvature perturbation on comoving slices by @xmath1 , defined by + = h + h^2 4 g _ 0^2 + , where @xmath55 and @xmath74 is the ratio of pressure to density in the background universe . the variable @xmath1 was introduced by bardeen@xcite in his classic paper , in which it was termed @xmath75 . it was employed in the context of inflation by bardeen , steinhardt , and turner @xcite and its use later elaborated upon by many authors @xcite . @xcite re - express @xmath1 as @xmath76 ( for spatially flat hypersurfaces ) and thereby derive the ` @xmath60-equation ' : @xmath77 where @xmath78 . for the power law solution given in equation ( [ eq : quants ] ) , this reads : @xmath79 since the last term is negative , it follows that there is no classical instability in the variable @xmath60 , a point emphasized by ref . ( note that ref . 3 relabels mukhanov s @xmath60 variable as @xmath59 ; here we have kept to the original notation . ) however , this does not at all imply the absence of a growing mode density perturbation : the variable @xmath3 _ does _ exhibit an instability , as we have discussed above . given a solution @xmath60 , the potential @xmath3 can be obtained by solving @xmath80 the general solution to the @xmath60-equation to order @xmath39 is : v = c_1 z ( 1- k^2 ^_- ^_- z^2 d ) + c_2 z ( ^_- - k^2 ^_- ^_- z^2 d ) , where @xmath81 and @xmath82 are arbitrary constants . substituting into eq . ( [ eq : sol1 ] ) , we obtain @xmath83 where @xmath84 , and @xmath85 and @xmath86 are constants . for the power law solutions studied here , @xmath87 . for an expanding universe , the first term dominates the second as @xmath88 . however , for a contracting universe , the second term is the proper growing mode as @xmath89 . the solution is the same as that found in eq . ( [ eq : phisol ] ) , from which we derived the scale - invariant spectrum . various subtleties are worthy of further comment . first note that @xmath84 is an _ exact _ solution of the perturbation equation ( [ eq : phieq ] ) for @xmath90 . in fact , for @xmath90 , the perturbation @xmath91 just represents a coordinate transformation of the the time @xmath25 , and is therefore unphysical . however , for nonzero @xmath32 the solutions tending towards @xmath91 are not gauge modes and therefore represent a real gravitational instability . second , note that @xmath1 is insensitive to any component of @xmath3 tending towards @xmath91 , as may be seen from the expression @xmath92 for inflationary models , tracking @xmath1 , as is done in many analyses , is useful because the components that are projected out in this formula are harmless decaying modes . however , for the ekpyrotic model , the growing mode tends to @xmath91 . one can in fact recover the the growing mode solution from @xmath1 or @xmath60 , but special care has to be taken to do so , by keeping the next order terms in @xmath39 as we have done above . in this section we show that the variables @xmath59 in eq . ( [ eq : neweq ] ) and @xmath60 in eq . ( [ eq : ueq ] ) at the heart of the analysis of the previous sections are in fact canonically conjugate . the importance of this is that in the @xmath93 phase space plane , the trajectories of the system are focused towards certain lines - for example the line @xmath94 in the case of the upside - down harmonic oscillator . these lines are then the classically amplified directions in phase space , and from liouville s theorem the phase space density is necessarily squeezed in the orthogonal directions . while the calculation of reference 5 is indeed correct , it does not by itself indicate the absence of a growing mode density perturbation , because the curvature perturbation which was computed is actually _ orthogonal _ to the direction in which fluctuations are amplified in the ekpyrotic setup . ( we note that references [ lyth2 ] and [ branden ] agree with this conclusion ) . our starting point is the action for gravity plus a scalar field in canonical ( first order ) form . the relevant formulae may be found for example in ref . the newtonian potential @xmath3 is written in terms of @xmath59 , and @xmath1 is written in terms of @xmath60 as above . one finds the quadratic action for perturbations reduces to = 2 m_pl^2 p d ( v f + v f -12 f^2 - k^22 v^2 ) , [ eq : act ] where @xmath95 , @xmath96 , and we have written the laplacian @xmath97 as @xmath98 . the sum over all fourier modes is implicit . the action ( [ eq : act ] ) is in canonical form @xmath99 , with @xmath45 the hamiltonian . therefore , up to normalization , @xmath60 and @xmath100 are canonically conjugate . the equations of motion are f=-^(v ^) k^2 v = ^- ( f ^) , [ eq : eoms ] which are just the relations above : equation ( [ eq : sol1 ] ) , which expresses @xmath3 in terms of @xmath1 and @xmath101 , and equation ( [ eq : zintp ] ) , which expresses @xmath1 in terms of @xmath3 and @xmath102 . these imply the equations of motion ( [ eq : feq ] ) and ( [ eq : ueq ] ) used in the previous sections . in the inflationary case , which corresponds to @xmath50 in the above analysis , both the variable @xmath60 and its canonical conjugate momentum @xmath100 ( or @xmath59 ) obey equations exhibiting a classical instability : both are amplified as @xmath53 approaches zero . the system starts near @xmath103 , with only small quantum fluctuations . as time proceeds the system evolves away from the origin along a certain line , @xmath104 in phase space . by liouville s theorem , the phase space density is ` squeezed ' in the direction orthogonal to this line . but in the ekpyrotic case , @xmath70 , the @xmath60 ( or @xmath1 ) variable is instead stable and classical trajectories instead run out along the @xmath59 ( or @xmath3 ) axis . the @xmath60 ( or @xmath1 ) direction must in consequence be ` squeezed ' . thus there is no classical growth in @xmath1 , consistent with the finding of reference [ lyth0 ] . having established that there is indeed a growing mode scale - invariant spectrum of density perturbations in the collapsing phase , we now turn to the more challenging question of whether these perturbations survive the reversal to expansion . we have recently addressed the issue of reversal in a paper with n. seiberg@xcite , and we shall employ and extend the analysis of that paper here . as we have discussed , the variable @xmath1 is not amplified in the collapsing phase , since it is insensitive to the growing mode perturbation . however , @xmath1 _ does _ yield the amplitude of the growing mode perturbation in the expanding phase . therefore , if @xmath1 was in fact the correct variable to match at @xmath105 , the growing mode perturbation which developed in the collapsing phase would just match onto a pure decaying mode perturbation in the expanding phase . this has indeed been suggested by lyth in a follow - up paper following our communication of the above calculations to him @xcite . the new argument is essentially that perturbation evolution may reverse at collision . that is , @xmath3 began negligibly small , grew to be very large at the bounce but then , after reversal to expansion , @xmath3 shrank precisely back to zero . in this section , we wish to explain why matching @xmath1 across the bounce is not appropriate in the scenarios we consider . instead , we claim , the scale invariant growing mode spectrum generated in the contracting phase matches onto a linear combination of growing and decaying modes at reversal . consequently , the density perturbations survive the reversal of the scale factor to seed structure formation in the hot big bang . in order to analyze matching at the bounce , gauge invariant variables must be identified which are well - behaved as @xmath6 and @xmath53 approach zero . we have already seen that the growing mode solution for @xmath3 is proportional to @xmath106 , which is divergent as @xmath53 tends to zero . likewise , as discussed below , in the situation of interest , @xmath1 is logarithmically divergent . since linear perturbation theory can only be valid when the perturbation variables are small , it follows that @xmath1 is not a good matching variable . nevertheless , with careful treatment to exclude the logarithmic divergence , @xmath1 shall be very useful in the analysis after @xmath105 , just because it s ` long wavelength ' component is nearly constant in the expanding phase , and gives the amplitude of the growing mode linear density perturbation . as mentioned in the introduction , the potential @xmath7 in ( [ eq : pot ] ) can not hold as @xmath15 . in the main example we treat here , we instead assume @xmath7 bends upward so that @xmath107 as @xmath15 . in this case , just before collision the system is completely described as a massless scalar field coupled to gravity , in a collapsing universe . as discussed in ref . 4 , one can in this case define regular variables , in which the evolution through @xmath108 becomes well defined . for the perturbation analysis , it is very important that , as @xmath109 , the universe is dominated by scalar field kinetic energy . in this situation , a good perturbation variable is the fractional energy density perturbation on spatial slices which are comoving with the matter , termed @xmath110 by bardeen in his discussion of gauge invariant cosmological perturbation theory @xcite . this variable ( defined in his equation 3.13 ) is a linear combination of the energy density perturbation @xmath111 , and the velocity perturbation @xmath112 , which is invariant under linearized coordinate transformations . it equals the energy density perturbation in any gauge in which the matter worldlines are orthogonal to the @xmath113constant hypersurfaces , and as bardeen emphasized it is the natural choice of perturbation amplitude from the point of view of the matter . the equations of motion for the matter , @xmath114 , lead to the following equations of motion , ( a^3 _ m)&=&-(+p)a^3 k v_s^(0)+h v_s^(0 ) & = & k + k ( 1+w ) ( c_s^2 _ m + w ) , [ eq : neweeq ] where @xmath115 is the gauge - invariant velocity perturbation defined by bardeen . here we introduce @xmath116 parameterizing the equation of state for pressure @xmath71 and energy density @xmath117 , @xmath118 being the sound speed , and @xmath119 being the ` entropy perturbation ' , defined as the difference between the pressure perturbation and that expected from the density perturbation and the background pressure - density relation . in most of the analysis below we shall assume @xmath119 is zero , which is to say that the density perturbations are ` adiabatic ' . for the types of matter we consider ( perfect fluids and scalar fields ) , the anisotropic stress is zero , and bardeen s potentials are related to our potential @xmath3 by @xmath120 . with these simplifications , and for the flat universe we consider , equations ( 4.5 ) and ( 4.8 ) of bardeen s paper@xcite reduce to ( [ eq : neweeq ] ) above . remarkably , scalar field kinetic energy domination implies that @xmath110 is actually finite at @xmath105 . this can be seen from the einstein constraint equation , which reads ^2 = 4 g a^2 _ m , [ eq : poisson ] where @xmath117 is the total background energy density . for scalar field kinetic domination , @xmath121 as @xmath6 tends to zero . but the growing mode perturbation @xmath122 and , from eq . ( [ eq : poisson ] ) , @xmath110 is finite as @xmath6 tends to zero . having identified a non - singular matching variable , we still need to decide which time slice to match on . in the present situation , this is unambiguously defined as time slice where @xmath0 tends to @xmath51 , or , re - phrased in terms of the brane separation , the time when @xmath123 . if we assume that no radiation is present in the contracting phase , so that the only energy density present is that in @xmath0 , then the surfaces of constant @xmath0 are also the comoving surfaces ( because the perturbation to the momentum is proportional to @xmath63 ) , and @xmath110 is the fractional energy density perturbation evaluated upon these surfaces . at the bounce , we assume that there is a change in the internal state of the branes so that @xmath7 switches off and the boundary branes no longer attract . after the bounce , @xmath0 is therefore a massless , free field . we also assume some radiation is produced on the branes . again the matching surface is defined in terms of @xmath0 , but with radiation present it is not obvious that the surfaces of constant @xmath0 are still comoving with the matter . however this is indeed the case for adiabatic perturbations as the following calculation reveals . we perform the calculation in conformal newtonian gauge . the equation governing the evolution of the radiation perturbation is given in this gauge by ( _ = 43 k v_r , [ eq : radeq ] where @xmath124 is the fractional energy density perturbation in the radiation and @xmath125 the velocity perturbation . we now make the assumption of adiabaticity , equivalent to the statement that the ratio of radiation to scalar field energy density is spatially uniform being determined locally by the microphysics of brane collision . the condition for the ` entropy perturbation ' @xmath119 , defined above , to be zero at early times is that @xmath126 , where @xmath127 is the fractional density perturbation in the scalar field . in newtonian gauge we have _ = 2 ( _0- ) . [ eq : dep ] the final ingredient in calculating @xmath125 is the equation for perturbations in the massless scalar field , + 2 h = 4 _ 0-k^2 . [ eq : scalaeom ] > from the adiabaticity condition and equations ( [ eq : radeq ] ) , ( [ eq : dep ] ) and ( [ eq : scalaeom ] ) , as well as the background equation @xmath128 , one determines v_r= - k _ 0. [ eq : vrad ] this is the velocity of the radiation in newtonian gauge . we wish to transform to the time - orthogonal gauge in which @xmath63 is zero . the transformation required is a shift in conformal time by @xmath129 accompanied by a shift in spatial coordinates with a potential @xmath20 such that @xmath130 , to ensure the absence of time - space components of the metric ( in the notation of section iii in bardeen s paper@xcite ) . the result of these transformations is that in the @xmath131 gauge , the velocity potential for the radiation equals @xmath132 , which is zero . therefore , for adiabatic initial conditions , the radiation fluid is actually comoving with the massless scalar field . therefore the comoving slices correspond to constant scalar field slices , both before and after the bounce . so we may therefore focus on the evolution of the energy density perturbation on comoving slices , @xmath110 , and attempt to match it across the bounce at @xmath105 . the equation of motion for @xmath110 is straightforwardly derived from ( [ eq : neweeq ] ) by eliminating @xmath115 , to obtain & & _ m+ f ( ) _ m + g ( ) _ m = - k^2 w , [ eq : perteq ] where & & f()=h ( 1 + 3 c_s^2 - 6w ) , & & g()= h^2(9 c_s^2 + 92 w^2 -12 w -32 ) + k^2 c_s^2 . as an aside , we now infer the behavior of the variable @xmath1 at small @xmath53 . one has = - ( a^2 h _ m ) ( 1+w ) k^2 a^2 . [ eq : zetanew ] employing the background relations =-12(1 + 3w ) h^2,w=-3h ( c_s^2-w)(1+w ) , [ eq : relat ] one obtains = h c_s^2 _ m 1+w . [ eq : zetaeq ] since @xmath101 is down by @xmath39 relative to @xmath1 , the rate of change of @xmath1 is typically much smaller than @xmath133 for the low @xmath32 modes of interest . so one might imagine @xmath1 would be a good matching variable . however , as noted above , @xmath110 is finite as @xmath53 tends to zero . from eq . ( [ eq : zetaeq ] ) and the fact that @xmath134 for small @xmath53 , one observes that @xmath1 diverges logarithmically at small @xmath53 . hence , as stated above , in linear perturbation theory @xmath1 is not a good variable for establishing a matching condition at @xmath105 . nevertheless , @xmath1 does have some utility . if one separates out the short wavelength piece which is divergent , the remaining long wavelength component of @xmath1 is finite and nearly constant . this finite piece is then very useful , since it gives the amplitude of the final growing mode linear density perturbation in the expanding universe . we now turn to an analysis of eq . ( [ eq : perteq ] ) . the first thing to note is that in the setup considered here , @xmath135 , @xmath74 and @xmath136 all have expansions in terms of simple powers of @xmath53 , for either @xmath137 or @xmath138 : & & h = 12 ^-1 + h_0 + h_1 + , & & w = 1 + w_1 + w_2 ^ 2 + , & & c_s^2 = 1 + c_1 + c_2 ^ 2 + . [ eq : expa ] this follows from computing & = & 12 a^2a^2 , w&=&12 ^2 -a^2v ^2 + a^2v = 3 m_pl^2 q^2 -a^2 u 3m_pl^2 q^2 + a^2 u , c_s^2&= & 1 + 23 v _ , a^5 a a^2 = 1 + 23 a_1 u_,0 + a_0 u_,1 m_pl q , [ eq : occont ] where @xmath139 , @xmath140 and @xmath141 . as discussed in ref . 4 , the variables @xmath142 , and @xmath143 are both finite at @xmath105 . these variables were introduced in ref . [ kost ] , with motivation from the five dimensional geometry . in the ads case , @xmath144 and @xmath145 are the scale factors on the positive and negative tension boundary branes respectively . although the four dimensional einstein frame scale factor @xmath6 tends to zero at an outer - boundary collision , @xmath0 tends to minus infinity in just such a manner as to leave @xmath144 and @xmath145 finite at collision . the physics of bounce and reversal are discussed in ref . [ nonsing ] , the results of which are assumed here . the key point for our analysis is that the equations of motion for @xmath144 and @xmath145 , m_pl^2 a_0= ua_0 m_pl^2 a_1= - u a_1 , [ eq : apmeqs ] are actually regular at @xmath105 . it follows that all the variables in eq . ( [ eq : occont ] ) have simple power series expansions around @xmath105 , as shown in eq . ( [ eq : expa ] ) , for either positive or negative @xmath53 . it will be important below to include a component of radiation after the bounce . as discussed in ref . [ nonsing ] , the density and temperature of radiation on the branes is actually finite at @xmath108 , as long as the radiation couples to the scale factors @xmath144 or @xmath145 , which are both finite ( and equal ) as @xmath15 . if , for example , @xmath146 is zero after the collision but there is a density @xmath147 of radiation produced on the brane with scale factor @xmath148 ( in the simplest models , we have @xmath149 or @xmath145 for the positive or negative tension branes respectively ) . then the expression for @xmath74 in eq . ( [ eq : occont ] ) is replaced by w&= & 3 m_pl^2 q^2 + 13 ( \|_r \|a^4 ) a^2 3 m_pl^2 q^2 + ( \|_r \|a^4 ) a^2 . [ eq : occontrad ] before proceeding to analyze eq . ( [ eq : perteq ] ) , it is important to recognize that the expansion coefficients in eq . ( [ eq : expa ] ) are not independent , but are related by eq . ( [ eq : relat ] ) . these imply , for example , that @xmath150 and @xmath151 . the coefficient functions in eq . ( [ eq : perteq ] ) have the following expansions f ( ) = -1 + f_0+f_1 + , g ( ) = g_-1 ^-1 + g_0 + g_1 + , [ eq : expafg ] with @xmath152 , @xmath153 , and @xmath154 . now the general solution of the perturbation equation ( eq . ( [ eq : perteq ] ) ) is a sum of the two linearly independent solutions _ m&= & _ 0 d ( ) + _ 2 e ( ) , [ eq : sol ] where @xmath155 and @xmath156 are arbitrary constants , and d ( ) & = & 1 + d_1 + d_2 ^2 ln \|\| + d_3 ^3 ln \|\| + d_3 ^3 + , e ( ) & = & ^2 + e_3 ^3 + , the equation of motion determines all the coefficients @xmath157 , @xmath158 , @xmath159 , @xmath160 and @xmath161 , @xmath162 . one finds @xmath163 , @xmath164 , and so on . since the coefficients @xmath165 , @xmath166 , @xmath167 change across @xmath105 , the two series expansions are different for @xmath137 and @xmath138 . we denote the coefficients in ( [ eq : sol ] ) , for @xmath137 and @xmath138 respectively , as @xmath168 , @xmath169 , and @xmath170 , @xmath171 . the matching rule we seek should determine the latter two constants in terms of the former . it seems clear that we should match the amplitude of the finite perturbation variable @xmath110 across @xmath105 , and this fixes @xmath172 . but how should we determine @xmath171 ? the simplest prescription is just to set @xmath173 . this amounts to matching the amplitude of the linearly independent solution which vanishes at @xmath105 , as well as that which is finite at @xmath105 . this prescription is invariant under redefining the independent solutions , e.g. by adding an arbitrary amount of the solution @xmath174 to @xmath175 . matching any other non - singular perturbation variable , defined to be an arbitrary linear combination of @xmath110 and @xmath176 with coefficients which are non - singular background variables ( defined to possess power series expansions in @xmath53 , as above ) will , with the same prescription of matching the amplitudes of both linearly independent solutions , also yield precisely the same result . this prescription is simple , but it is certainly not unique , and we emphasize that a proper understanding of the correct matching condition must ultimately rest on a better understanding of the singularity , either directly from string or m theory or from a well defined regularization and renormalization procedure . note that any such matching rule applied at @xmath105 can not match both @xmath110 and @xmath176 , as would be appropriate at a regular point of the differential equation . this is because @xmath110 and @xmath176 _ can not _ be independently specified at @xmath105 , just because @xmath105 is a singular point of the differential equation . the value of @xmath156 is fixed by the limit as @xmath53 tends to zero of @xmath177 , so one needs to know @xmath178 as @xmath53 tends to zero , in order to determine @xmath156 . recall that since @xmath1 diverges logarithmically as @xmath179 , it is not a good matching variable . however , this divergence only affects the short wavelength part of @xmath1 , and the long wavelength part is still very useful since it yields the amplitude of the growing mode perturbations in the expanding phase . as we now see , the above prescription for matching @xmath110 actually implies that the long wavelength part of @xmath1 has a jump across @xmath105 . using eq . ( [ eq : relat ] ) , we re - express eq . ( [ eq : zetanew ] ) as = - h ( 1+w ) k^2 ( _ m + 32 h ( 1-w ) _ m ) . [ eq : zetajump ] substituting the above expansions , one finds the leading order behavior ~_04 ( ln \|\| -12 ) + k^-2(12 _ 2 -316 _ 0 w^(2 ) ) , [ eq : zetaj ] where @xmath180 , plus terms which vanish as @xmath53 tends to zero . the first term is logarithmically divergent at @xmath105 . however , since it is down by a factor of @xmath39 , it rapidly becomes irrelevant as @xmath53 increases away from zero . the second term , the long wavelength piece @xmath181 , is the quantity we are actually interested in . this constant , long wavelength piece is accurately conserved after the bounce as long as the matter evolution remains adiabatic , and yields the amplitude of the growing mode adiabatic density perturbation in the late universe . as we have discussed above , @xmath110 is finite at @xmath105 and from section iii and iv , there is no long wavelength contribution to @xmath1 in the collapsing phase . it follows that @xmath182 in that phase . one situation of sepcial interest is where @xmath183 in the collapsing phase , where the potential is irrelevant at @xmath184 and there is no radiation in the incoming state . in this case , @xmath185 . in this case , we would obtain the same final result from any matching rule which set @xmath186 , with any constant @xmath187 . the key point is that generically _ @xmath188 jumps _ across @xmath105 , since the background equation of state changes at the brane collision . matching @xmath155 and @xmath156 we obtain a long wavelength contribution to @xmath1 in the expanding phase , ^lw ( > 0 ) ~- 316 k^-2 _ 0 ( w^(2)<-w^(2 ) > ) , [ eq : zetajexp ] where @xmath189 and @xmath190 are the values of @xmath188 for @xmath137 and @xmath138 respectively . since @xmath191 and , as discussed earlier , @xmath3 has a nearly scale invariant power spectrum , it follows from eq . ( [ eq : zetajexp ] ) that if @xmath188 undergoes a jump , then @xmath1 will inherit a scale invariant long wavelength piece . this is our main result . in the following sections we will study a simple example of a situation where @xmath188 is discontinuous . we shall need explicit formulae for the jump in @xmath188 , and for @xmath155 . the formulae for @xmath188 are obtained from eqs . ( [ eq : occont ] ) and ( [ eq : occontrad ] ) above . if we assume that prior to @xmath105 there is no radiation , so that in addition to scalar kinetic energy we have only the potential @xmath146 , then taking the limit as @xmath53 tends to zero from below one finds that w^(2)<= 2 u_0 ^ 2 + 43 m_pl^2 q u_1 m_pl^4 q^2 , u_1= u\|_0 = a_0u a_0 + a_1u a_1 , [ eq : befo ] where @xmath192 . likewise , for the expanding phase , if we assume @xmath193 but radiation is now present , we obtain w^(2)>= 827 ( \|_r \|a^4)^2m_pl^4 q^2 . [ eq : afte ] since in general @xmath194 , we infer that generically , a scale invariant spectrum of perturbations will , with our prescription above , propagate across @xmath105 into the expanding hot big bang phase . finally , to compute the perturbation amplitude given in eq . ( [ eq : zetajexp ] ) we need @xmath155 . this can be read off from the expression eq . ( [ eq : phisol ] ) above , by translating the @xmath53 dependence into @xmath106 and employing the fact that the latter gives the exact dependence for the long wavelength modes of interest even when the potential breaks away from the pure exponential form used in the first half of this paper . we find that at @xmath105 , _ 0= 4 k^1- 3 m_pl a_0 ( a_0-a_1 ) . [ eq : epsres ] to summarize the results of this section , we have elaborated the conditions under which a scale invariant spectrum of perturbations survives the passage through @xmath108 . basically this requires that the equation of state , as parametrized by @xmath195 , have a discontinuous first or second derivative with respect to @xmath53 at @xmath105 . this condition would appear to be quite generically fulfilled , in any situation where entropy is generated at a brane collision . a key assumption in the calculation was that the energy density is dominated by scalar field kinetic energy as we approach @xmath108 . this assumption is natural in the ekpyrotic scenario , where the scalar field represents the separation of the boundary branes . we shall explore one particular , simplified model in the next section . in this section we review and extend the description of reversal from contraction to expansion , as elaborated in reference 4 . as discussed there , the background evolution is described by the variables @xmath144 and @xmath145 , which for the simplest brane model ( i.e. branes in ads ) represent the scale factors on the positive and negative tension boundary branes . the four dimensional effective scale factor @xmath6 is given by @xmath196 , and this vanishes at the bounce . if the potential @xmath7 vanishes as @xmath0 runs off to @xmath51 ( more precisely , if the quantity @xmath197 vanishes ) , then the friedmann constraint equation implies that trajectory in the @xmath198-plane intersects the boundary of moduli space @xmath199 along a light - like direction @xcite . then , if no radiation is produced on the branes , the trajectory simply reverses , corresponding to the matching condition @xmath200 . we describe this as an _ elastic _ collision , since the internal states of the two branes are unchanged by the collision . however , at any finite velocity , the boundary brane collision must result in the production of radiation on the branes , since it is a non - adiabatic process . in the m theory context the corresponding string theory is weakly coupled near the collision , and this production of radiation should be computable once the correct matching conditions are understood . let us consider the case where the incoming state has no radiation , and the potential @xmath7 vanishes at @xmath5 and thereafter . this requires that the potential is turned off at collision , requiring a sudden and permanent change in the internal state of the branes . associated with this change , we assume that a small amount of radiation , with density @xmath201 is generated , on a brane with scale factor @xmath148 . the collision is therefore inelastic . we parameterize the inelasticity as follows . the friedmann constraint after collision reads @xmath202 since right hand side is positive , the outgoing trajectory must be time - like in the @xmath203-plane . since radiation redshifts as @xmath204 , the expression @xmath205 is a constant . if radiation is generated on both branes , this term can be taken to represent the sum of the corresponding terms for both branes . we then define the the efficiency @xmath206 with which radiation is produced by @xmath207 which with ( [ eq : fc ] ) yields a single equation for the two velocities @xmath208 and @xmath209 . we need another equation to fix both . in the special case of dimensional reduction from five to four dimensions , there is a natural candidate for an approximately conserved quantity , analogous to the total momentum for an inelastic particle collision . as mentioned above , at small brane separations one has the standard kaluza - klein result that the size of the extra dimension is proportional to @xmath210 . as stated above , we are assuming that the potential @xmath7 vanishes as @xmath0 tends to @xmath51 . other terms in the lagrangian describing matter on the branes may in principle acquire @xmath0-dependence upon dimensional reduction . however , the terms describing massless gauge fields and fermions do not obtain any such @xmath0-dependence due to their conformal invariance in four dimensions . ( to see this , note that the four dimensional einstein - frame metric @xmath211 is conformally related to the four dimensional components of the five dimensional metric . ) therefore at the classical level , in the @xmath184 limit , the lagrangian describing gravity , @xmath0 and four dimensional radiation possesses a global symmetry @xmath212 , and it is plausible that the corresponding noether charge @xmath213 , @xmath214 tends to a constant as the collision approaches . in this limit the classical equation of motion of @xmath0 is just @xmath215 . however , the sign of @xmath213 must flip at the bounce . as argued in ref . 4 , this is essential in order that the trajectory remains in the physical region of the @xmath198-plane . the reversal of @xmath213 may be also understood by the following higher - dimensional argument . the value of @xmath213 at collision is proportional to the time derivative of log@xmath216 . the latter quantity is the distance , rather than the vector displacement , between the boundary branes . hence , if the branes come together and then draw apart , @xmath213 must change sign although its magnitude remains constant . it is therefore natural to impose the @xmath217 symmetry at @xmath5 , @xmath218 , which should become exact in the limit that the collision velocity approaches zero . these arguments suggest that we parameterize @xmath213-violation at the bounce ( brane collision ) using @xmath219 where @xmath220 is expected to be small . equations ( [ eq : fc ] ) , ( [ eq : efficiency ] ) and ( [ eq : chidef ] ) together uniquely parameterize the final values of @xmath221 and @xmath222 after the bounce : @xmath223 an example of an ekpyrotic two - brane collision is shown in figure 1 , for a specific choice of the inter - brane potential . both branes initially expand under the influence of the attractive potential . but when they get close , and the potential rises to zero , this decelerates @xmath144 so that it begins to contract . at collision , @xmath224 . immediately afterwards , @xmath221 is positive and , for small @xmath206 , @xmath222 is negative . assuming that the potential remains zero after collision , and that only radiation is present , we have @xmath144 and @xmath145 flying apart linearly in @xmath25 after collision , corresponding to an expanding @xmath6 . according to ( [ eq : nullb ] ) , @xmath145 is _ decreasing _ after collision , and it tends to zero . this would lead to the separation between the branes going to infinity . however , it is easily avoided , essentially because the @xmath145 modulus has a _ positive _ kinetic term . if there is one or more massless moduli fields @xmath28 coupling to @xmath145 as @xmath225 , they produce an effective potential in the @xmath145 equation which is proportional to @xmath226 . this repels the @xmath145 modulus from @xmath227 . figure 1 shows an example of the full evolution , including the brane collision , reversal and turn - around of @xmath145 so that @xmath144 and @xmath145 are both expanding at late times . the final evolution of @xmath145 is insensitive to the value of @xmath228 , which only affects the evolution at small @xmath145 . in the long time limit , the scalar field @xmath0 also tends to a constant and therefore so does the inter - brane separation . one can generalize these considerations to examples where the potential @xmath229 is finite and negative at brane collision . this requires rather special potentials @xmath7 , which diverge at large @xmath0 , but not too strongly . in this case , the trajectories are space - like at collision . having specified the background evolution for the ` nearly light - like bounce ' , we can now consider the matching of perturbations . we have assumed that @xmath146 vanishes at collision . for simplicity we shall assume that its first derivatives @xmath230 and @xmath231 also vanish there . now we can read off from eq . ( [ eq : befo ] ) that @xmath232 , but we have from eq . ( [ eq : afte ] ) and the formulae of this section that w^(2)>= 23 ^2 a_1(in)^2a_1 ^ 2 . [ eq : nbr ] we can also read off from eq . ( [ eq : epsres ] ) that for the ` nearly light - like bounce ' , _ 0= -2 k^1- 3m_pl a_1 a_1(in ) , [ eq : epsresn ] where we assumed @xmath233 . putting these together in ( [ eq : zetajexp ] ) , we find the final density perturbation amplitude is ^lw ~112 ^2 k^-1- m_pl a_1(in)a_1 ^ 3 . [ eq : finalamp ] it remains to compute @xmath145 and @xmath234 at collision using the detailed behavior of the potential @xmath7 for negative @xmath0 . as discussed in ref . 4 , as @xmath15 , the string coupling constant tends to zero . it is natural to expect that the potential @xmath7 goes to zero in this limit . we shall adopt a very simplified model here , in which the potential jumps to zero at some particular , negative value of @xmath235 . in this case it is straightforward to compute @xmath145 and @xmath222 as @xmath53 tends to zero from below . first , assume the jump happens at some time @xmath236 , before which equations ( [ eq : quants ] ) are valid . the total energy in the scalar field is @xmath237 , and this is equal to the kinetic energy in the scalar field after the potential jump . so , just after the jump we have = - . [ eq : velaft ] now , from @xmath238 we have a_0= a(ha_0- m_pl a_1 ) [ eq : avels ] and similarly for @xmath222 . since the potential vanishes , and no radiation is present , @xmath221 and @xmath222 are both constant up to collision . then equations ( [ eq : back ] ) , ( [ eq : quants ] ) and ( [ eq : avels ] ) imply that a_0(in)=-2 p ( -t_j)^2p-1 e^_j/ ( m_pl)= -a_1(in ) . [ eq : velsj ] after the potential jump , @xmath144 and @xmath145 evolve linearly in @xmath53 . to leading order in @xmath206 and @xmath220 we have a_0&= & ( -t_j)^p ( 2 cosh(_j/ m_pl)+a_1(in)(-_j))a_1&= & ( -t_j)^p ( 2 sinh(-_j/ m_pl)-a_1(in)(-_j ) ) . [ eq : linear ] setting these equal determines the time of collision and brane scale factors at collision , a_0=a_1= ( -t_j)^p e^-_j/ ( m_pl ) . [ eq : branesc ] now we have all we need to determine the final fluctuation spectrum . from equations ( [ eq : finalamp ] ) , ( [ eq : velsj ] ) and ( [ eq : branesc ] ) , we have ^lw ~16 m_pl k^1 + ( -t_j)^1+p ^2 e^4_j/ ( m_pl ) . [ eq : finalampf ] the dependence on @xmath32 , @xmath236 and @xmath71 is the same as that one obtains from the naive ` time delay ' formula mentioned in section ii . using equation ( [ eq : dps ] ) for @xmath63 , the time delay method yields a perturbation amplitude @xmath239 , in agreement with the dependence upon these quantities in equation ( [ eq : finalampf ] ) . in the time delay argument , however , one uses the hubble constant on the branes at collision , which is close to but not quite the same as the factor occurring in ( [ eq : finalampf ] ) . let us now translate the dependence of the last factor in equation ( [ eq : finalampf ] ) into quantities determined by observations in the final expanding , radiation dominated universe . after collision , we assume the potential @xmath7 is zero ( because the internal state of the branes has changed so they no longer attract each other ) , so that the @xmath0 modulus describing the inter - brane separation is a free massless field . we also assume radiation is present , in an abundance parameterized by @xmath206 . it is straightforward to analytically solve the equations of motion post - collision , assuming the presence of the additional modulus needed to keep @xmath145 away from zero . this modulus becomes irrelevant at late times . in the long time limit , one finds @xmath144 and @xmath145 increasing linearly in conformal time @xmath53 , with @xmath240 , neglecting the dependence on @xmath220 which is reasonable if @xmath220 is small . equating this to coth@xmath241 , we find that the _ final _ resting value for @xmath0 is given by e^(2 _ f / m_pl ) = 4 . [ eq : finalphi ] however , what enters equation ( [ eq : finalampf ] ) is not @xmath242 , but @xmath235 , the value of @xmath0 at which the potential @xmath7 switches off . we can translate both values of @xmath0 into the corresponding string coupling constants , using the relation @xmath243 , which follows from m theory with the assumption that the six calabi - yau dimensions are fixed in the 11 dimensional metric @xcite . for @xmath71 close to zero , we find the final result k^32 ^lw_k ~ ( g_s(j ) g_s(f))^43 . [ eq : finalampfin ] the right hand side is the amplitude of the growing mode adiabatic density perturbation relevant to structure formation in the late universe . ( for an accurate calculation one should of course retain the @xmath71 dependence , since over the many orders of magnitude of @xmath32 involved , this can significantly affect the final normalization . we leave this complication for future work . ) our result for @xmath244 depends on the square root of the potential energy @xmath245 at its minimum , in planck units . this is reminiscent of the usual inflationary result . however , additional suppression factors arise . first , the numerical coefficient is small . second , the factor @xmath246 is small if the efficiency of production of radiation at collision is small . finally , the string coupling constant where the potential turns off , which we have crudely parameterized as @xmath247 , would be expected to be substantially smaller than the value of the string coupling constant in the asymptotic outgoing state . translating the formulae relevant to horava - witten theory , given by witten@xcite , one finds for today s value of the string coupling constant g_s(f)= ( m_plm_gut)^3 g_gut^4 v_0 ^ 12 , [ eq : gsf ] where the volume of the calabi - yau manifold is @xmath248 before discussing numerical values , it is important to make the following caveats . first , the resting value of the scalar field @xmath0 determined during the radiation era following the bounce is not necessarily that measured in today s universe . if there is a stabilizing potential for @xmath0 , that will instead determine the final resting value nevertheless it is conceivable that the resting value early in the hot big bang phase is closely related to the final value ( as for example if @xmath0 develops a potential with many closely spaced degenerate minima ) . second , the presence of additional moduli ( such as are found in horava - witten theory ) could have important consequences on the dynamical evolution of @xmath0 in these early stages . in the above calculation we have limited ourselves to only one modulus , translating that directly into the string coupling constant . so the final numerical result can only be suggestive . for example , plausible values of the gut coupling are @xmath249 , and the gut mass @xmath250 gev , giving @xmath251 . the turn - off of non - perturbative potentials might plausibly occur at @xmath252 , if instanton effects produce factors of the form exp(@xmath253 ) . the last factor in ( [ eq : finalampfin ] ) then yields @xmath254 . for @xmath255 , and @xmath256 , we can obtain an amplitude @xmath10 , as required by observations . we conclude that the ekpyrotic scenario may offer a natural explanation for the smallness of the observed density perturbations . as we have emphasized , this is only suggestive at this stage , and will remain so in the absence of : @xmath257 a microscopic check of the matching condition used for @xmath110 , within the context of m theory and string theory , @xmath257 a computation of the efficiency parameter @xmath206 describing the production of radiation on the branes , @xmath257 a check that the parameter @xmath258 is indeed small , as was assumed , @xmath257 and , a full calculation of a realistic inter - brane potential @xmath7 and a numerical solution of the equations improving the jump approximation used above . although we have focused here on matching scalar perturbations , in principle one also needs a matching condition for tensor and vector modes . neither acquire long wavelength power in the contracting phase of the ekpyrotic model , and it seems unlikely they will be generated at the brane collision . nevertheless one can attempt to study possible matching conditions . it is not hard to see that the tensor amplitude @xmath259 exhibits the same logarithmic divergence as the perturbation in the three - curvature of comoving slices , @xmath1 . however , the canonically conjugate momentum @xmath260 does tend to a finite constant at @xmath105 , suggesting it provides a possible matching variable . again , establishing this will probably require a satisfactory microscopic theory . we have shown that inclusion of gravitational backreaction has a negligible effect on the density perturbations produced in the ekpyrotic universe , during the initial phase which is slowly contracting from the point - of - view of the four dimensional effective theory . we have proposed what we believe is a physically sensible matching condition at the bounce , based upon identifying physical perturbation variables which are well behaved ( i.e. small ) at @xmath105 , and matching on surfaces defined by the scalar field @xmath0 . the variables we used are well - behaved , we should emphasize , provided certain conditions are met as the 4d effective scale factor approaches zero , namely ( 1 ) the energy density is dominated by the kinetic energy of a scalar field ( the modulus @xmath0 in our case ) and ( 2 ) the inter - brane potential @xmath7 does not diverge as @xmath0 tends to @xmath51 . we showed that with the simplest matching prescription , namely matching the two linearly independent solutions at @xmath105 , the scale invariant spectrum of perturbations developed in @xmath3 early on in the contracting phase is generically passed on to the variable @xmath1 representing the amplitude of the long wavelength growing mode density perturbation in the expanding phase . we also identified examples where no density perturbations are generated in the final universe . if no radiation is generated at the brane collision , and if the potential vanishes sufficiently smoothly there , the perturbations ` time - reverse ' at @xmath108 and the amplitude of the growing mode perturbation is precisely zero in the final expanding universe . such examples are not realistic since there is no entropy generation at the brane collision , and in any case seem highly unlikely given that the outer - brane collision is not an adiabatic event . the existence of a ` zero - perturbation limit ' is an intriguing feature of the ekpyrotic model and the matching prescription given here , since it suggests a natural explanation for the smallness of the observed density perturbations . recall that a feature of inflation is that it naturally predicts a value of the density perturbation amplitude that is far too large , and fine - tuning of potentials is required to obtain a sufficiently small amplitude . we conclude that , while many issues connected to the microphysics at brane collision remain to be settled by rigorous investigation of string theory in the limit of outer - brane collision , the basic idea of ref . [ kost ] for producing density perturbations during the early stages of the ekpyrotic universe , in a phase which is slowly contracting from the four dimensional point of view , remains viable . * acknowledgements : * we thank v. mukhanov for insightful remarks and contributions . we have greatly benefited from explaining our arguments and criticisms to d. lyth prior to publication and obtaining his comments . we especially thank d. wands for helpful correspondence and important questions . we also thank r. brandenberger for showing us a preliminary version of his preprint . this work was supported in part by the natural sciences and engineering research council of canada ( jk ) , the us department of energy grants de - fg02 - 91er40671 ( jk and pjs ) and de - ac02 - 76 - 03071 ( bao ) , and by pparc - uk ( nt ) . * note added : * as mentioned in the introduction , our conclusions differ from those of lyth , brandenberger and finelli , and hwang@xcite . the disagreement is due to different assumptions about physical conditions near the bounce , and to a different prescription for matching across it . we believe that the assumptions made by the authors of refs . 7 - 9 are inconsistent with what we proposed in refs . 1 and 4 , because @xmath257 they use a scalar field potential which diverges to minus infinity as the bounce is approached . this potential is not compatible with the bounce prescription discussed in ref . 4 , since it is too singular . @xmath257 they choose to match on a surface of constant energy density in the contracting phase , assuming that the scale factor of the universe reverses from contraction to expansion on this surface . this behavior is incompatible with the field equations describing the cosmological background solution . we instead follow the classical field equations all the way to the bounce , and apply a matching prescription consistent with our treatment of the background at this point . as we have explained above the appropriate matching surface in the ekpyrotic setup is defined by the scalar field specifying the inter - brane separation , and not by the energy density . @xmath257 brandenberger and finelli claim our matching prescription based on the energy density perturbation on slices of constant scalar field is inconsistent with results given in the literature for models where the equation of state undergoes a sudden jump@xcite . it is easy to see why matching @xmath110 and its time derivative in this situation is incorrect . the equation of motion ( [ eq : perteq ] ) for @xmath110 involves @xmath136 , related by ( [ eq : relat ] ) to the time derivative of @xmath74 . hence , if @xmath74 jumps , @xmath136 acquires a delta function contribution , which causes a jump in @xmath176 across the matching point , which may be straightforwardly computed . the key point , however , is in the situation we are discussing , @xmath74 _ is continuous _ across the bounce , thus there are no such delta function contributions . therefore this ` counterexample ' is truly a red herring . @xmath257 brandenberger and finelli mistakenly imply that reversal from contraction to expansion would occur at a bulk brane - boundary brane collision . if the four dimensional effective description is valid , then reversal can _ only _ happen at a boundary brane - boundary brane collision , as described in ref . 4 . in the four dimensional effective description the universe continues to contract after a bulk - boundary collision , so @xmath1 remains small and @xmath3 continues to grow . only when the outer boundary brane collision occurs as it must , can the growing perturbations in @xmath3 get converted to long wavelength fluctuations in @xmath1 . 9999 j. khoury , b.a . ovrut , p.j . steinhardt and n. turok , hep - th/0103239 . [ kost ] a. guth and s .- y . pi , _ phys . * 49 * , 1110 ( 1982 ) . l. wang , v. mukhanov and p.j . steinhardt , _ phys . b_414 , 18 ( 1997 ) . j. khoury , b.a . ovrut , n. seiberg , p.j . steinhardt and n. turok , hep - th/0108187 . [ nonsing ] d. lyth , hep - ph/0106153 . [ lyth0 ] d.h . lyth and e.d . stewart , _ phys . lett b_274 , 168 ( 1992 ) . d. lyth , lancaster preprint , in preparation ( 2001 ) . [ lyth2 ] r. brandenberger and f. finelli , hep - th/0109001 . [ branden ] j. hwang , astro - ph/0109045 . m. bucher , a.s . goldhaber and n. turok , _ phys . rev . d_52 , 3314 ( 1995 ) . [ bgt ] j.m . bardeen , physical review * d22 * ( 1980 ) 1882 . v. f. mukhanov , _ jetp lett . _ * 41 * , 493 ( 1985 ) ; sov . phys . jetp * 68 * ( 1988 ) 1297 ; v.f . mukhanov , h.a . feldman and r.h . brandenberger , phys . reports * 215 * ( 1992 ) 203 . bardeen , p.j . steinhardt and m.s . turner , phys . d * 28 * ( 1983 ) 679 . r. brandenberger and r. kahn , _ phys . rev . d_29 , 2172 ( 1984 ) . s. gratton and n. turok , physical review * d60 * ( 1999 ) 123507 . [ gt ] e. witten , hep - th/9602070 . n. deruelle and v.f . mukhanov , gr - qc/9503050 , phys.rev . d52 ( 1995 ) 5549 .	we study the generation of density perturbations in the ekpyrotic scenario for the early universe , including gravitational backreaction . we expose interesting subtleties that apply to both inflationary and ekpyrotic models . our analysis includes a detailed proposal of how the perturbations generated in a contracting phase may be matched across a ` bounce ' to those in an expanding hot big bang phase . for the physical conditions relevant to the ekpyrotic scenario , we re - obtain our earlier result of a nearly scale - invariant spectrum of energy density perturbations . we find that the perturbation amplitude is typically small , as desired to match observation . we recently proposed a novel scenario for the early universe in which the hot big bang is created by the collision between two m - theory branes@xcite . the scenario assumes the universe begins in an almost static , nearly bps initial state consisting of empty , flat , parallel three - branes . in the effective 4d theory , the bps state is homogeneous and has zero spatial curvature . due to non - perturbative effects , however , a tiny force attracts the branes to one another . as the branes come together , quantum fluctuations create ripples in the brane surfaces that result in spatial variations in the time of collision . consequently , some regions heat up and begin to cool before others , producing a spectrum of long wavelength density perturbations which can seed structure formation in the universe . we estimated the perturbation spectrum using a ` time delay ' formalism@xcite , often used in simplified treatments of inflationary models . in that context , spatial variations in the time when inflation ends result in long wavelength density inhomogeneities . we applied the same formalism to variations in the time of collision in the ekpyrotic scenario . the equation for fluctuations in the scalar field @xmath0 describing the inter - brane separation in the ekpyrotic model is almost identical to that describing fluctuations in the inflaton during slow - roll inflation . consequently , a nearly scale - invariant spectrum of fluctuations is found . the result is remarkable because it shows that the harrison - zeldovich spectrum can be obtained without inflation in a space - time which is very nearly static minkowski space . the time delay formalism is a crude approximation , and only quantitatively accurate for a small class of inflationary potentials @xcite . nevertheless , it often gives a good estimate of the spectral index for the power spectrum of perturbations . one of the goals of this paper is to investigate whether the same statement is true for the ekpyrotic model . in the case of the ekpyrotic model , there is the major complication that the perturbations are produced when the effective 4d scale factor is contracting . in order to have a viable scenario , a mechanism must be found to reverse from contraction to expansion . this issue has been addressed in a recent paper we have written with n. seiberg@xcite , where we argue that such a ` bounce ' may be allowed in the context of m - theory , where it corresponds to a collision and rebound of the outer boundary branes . a matching rule linking the homogeneous background variables of the contracting phase to those describing the expanding phase was suggested there . assuming this proposal is valid , what remains is to apply and extend those ideas to describe the evolution of perturbations through the moment of reversal . a stimulus for the present work was a paper by d. lyth@xcite , which calculated the growth of the perturbation variable @xmath1 ( also commonly termed @xmath2 ) representing the curvature perturbation of spatial slices which are comoving with the matter . lyth correctly showed that @xmath1 was not amplified in the contracting phase of the ekpyrotic universe . he claimed this implied that when gravitational back - reaction was included , the spectrum of density perturbations became strongly scale - dependent with negligible power on large scales , making the ekpyrotic scenario incompatible with observations . his analysis employed a certain class of analytically solvable models with exponential potentials , previously used to describe power law inflation@xcite and simply extended to the situation of slow contraction relevant to the ekpyrotic scenario . we repeat his analysis here , but also compute the perturbation in the newtonian potential @xmath3 . we show that in the contracting phase gravitational back - reaction actually enhances rather than suppresses long wavelength fluctuations , but these fluctuations show up purely in @xmath3 and not in @xmath1 . therefore gravitational back - reaction does not spoil the ekpyrotic mechanism , at least in the contracting phase . the remaining issue regards the appropriate matching condition for tracking the perturbations across the bounce and into the expanding hot big bang . consistent with the arguments of ref . 4 , we seek to identify variables which are non - singular at the bounce , both for the background and perturbation variables . we then match the amplitudes of the two linearly independent solutions for the perturbation variables across the bounce . with our prescription , we find that the long wavelength perturbations developed in the contracting phase do indeed survive to the expanding phase , provided there is a change in the equation of state at the bounce , such as occurs if a sub - dominant component of radiation is produced there . our final expression for the density perturbation spectrum agrees well with the more naive time delay estimate . after communicating a preliminary version of this paper to lyth and r. brandenberger , lyth prepared a second manuscript@xcite proposing that contraction be matched to expansion on a time - slice of fixed energy density@xcite . with this procedure , he argues that the curvature perturbation @xmath1 is conserved across the bounce . but since @xmath1 does not acquire a scale invariant spectrum in the contracting phase , he argues that @xmath1 will not have such a spectrum in the expanding phase , where it represents the amplitude of growing mode adiabatic perturbations . lyth s conclusion is that any growing mode density perturbations developed in the contracting phase match perfectly onto pure decaying mode perturbations in the expanding phase . brandenberger and finelli@xcite , and hwang@xcite , have recently produced preprints repeating this argument . in a note added , at the end of this paper , we explain why we do not believe these conclusions are valid for the ekpyrotic scenarios proposed in refs . 1 and 4 . let us outline our approach to the matching problem . we want to evolve background and perturbation variables according to the appropriate field equations , all the way to to zero scale factor in the four dimensional effective theory . we identify a complete set of variables which are non - singular at the bounce , and match those non - singular variables across it . this prescription automatically excludes the variables @xmath3 and @xmath1 , both of which diverge . more generally , geometrical quantities such as the synchronous gauge comoving metric perturbation @xmath4 also diverge . indeed the meaning of the three - geometry is unclear at zero scale factor . instead our approach is essentially algebraic rather than geometrical . we focus on gauge invariant perturbation variables which are consistently small at all times and match these across the bounce at @xmath5 . we argue that this matching would give consistent results for an infinite class of perturbation variables so defined . our prescription can only be fully justified by a satisfactory microscopic description of the relevant degrees of freedom . nevertheless , if string theory shows that the scale factor can truly pass through zero and bounce , then tracking perturbative gauge - invariant degrees of freedom which remain small and finite seems likely to be the right approach to matching fluctuations across the bounce . our analysis shall be performed entirely within the context of four dimensional effective field theory . this does not capture all the low energy degrees of freedom relevant to the five ( or indeed eleven ) dimensional brane - world . as in ref . 1 , we shall assume those other degrees of freedom are frozen , or at least so slowly varying that their inclusion would not substantially alter the result . we shall focus here on single moduli field @xmath0 which determines the outer brane separation in a brane - world universe . in ref . 1 we considered a model in which the perturbations are produced by the collision of a bulk brane with one of the boundary branes . as we pointed out there , if the 4d effective theory is valid , the scale factor in that theory must continue to contract until the outer branes collide and bounce . in ref . 4 we suggested a simplified model in which there is only one collision between the boundary branes and no bulk brane is needed . the perturbations are produced as the outer branes approach one another . for simplicity , here we shall restrict ourselves to this two - brane scenario , in which the same scalar field @xmath0 is responsible both for the development of the perturbations , and for describing the the outer - brane collision and bounce . generalizations to bulk - boundary collisions , as in the original scenario of ref . 1 , are a straightforward extension and will be briefly mentioned when appropriate . the outline of this paper is as follows . in section i we discuss the properties of the inter - brane potential relevant to the ekpyrotic scenario , an issue referred to again in the note added at the end of this paper . we then review the application of the time delay formalism to the ekpyrotic scenario , as described in ref . 1 , showing that a scale invariant spectrum of fluctuations is naturally predicted . in section ii , we show that including gravitational back - reaction has , in newtonian gauge , only negligible effects on the fluctuations acquired by the scalar field . section iii is devoted to a discussion of the role of the curvature perturbation @xmath1 ( or @xmath2 ) conventionally used in the analysis of inflationary models , which is in fact insensitive to the growing mode perturbation in the contracting phase of the ekpyrotic model . this is further elaborated in section iv where we show that @xmath1 is canonically conjugate to the direction of amplification , which is proportional to the variable @xmath3 . as a consequence of liouville s theorem , @xmath1 is ` squeezed ' as the perturbations develop during the collapsing phase of the ekpyrotic scenario . section v addresses the key moment of reversal , which is when , in the four dimensional effective description , the universe ` bounces ' as the scale factor @xmath6 hits zero . since the four dimensional spacetime geometry is singular there , a four dimensional geometrical description using the einstein frame metric is inappropriate . however , our approach is to identify a complete set of dynamical variables which remain finite at the bounce and to match them at the transition from contraction to expansion . we ignore divergent quantities as unphysical , attributing their bad behavior to a singular choice of field and metric variables . we used this approach in ref . [ nonsing ] to describe the homogeneous background evolution during the bounce . here we shall show that there is also a set of gauge invariant linear perturbation variables which are well - behaved at the bounce . when the theory is formulated in these variables , there is a well defined matching prescription , even in the four dimensional effective field theory , which seems to yield physically sensible results . in order for the well - behaved variables to exist , we emphasize that it is important to have two conditions : ( 1 ) there must be a free scalar field ( the modulus @xmath0 in our case ) whose kinetic energy diverges at the bounce and ( 2 ) the scalar potential @xmath7 must be sufficiently non - singular at the bounce . the simplest case is where @xmath7 approaches zero at the bounce , as suggested by the mapping from m - theory to weakly coupled string theory@xcite . if these two conditions are fulfilled , then it is possible to follow the perturbations and match at @xmath8 . our main finding , in sections v and vi , is that as long as radiation is produced or fields are excited at the outer - brane collision so that there is a jump in the first or second time derivative of the equation of state parameter @xmath9 , the scale invariant perturbation spectrum developed during the contracting phase propagates through the bounce and into the final expanding universe . intriguingly , the final density perturbation amplitude derived in section vi is naturally suppressed by a small numerical coefficient , by quantities which are automatically small in horava - witten theory , and by factors involving the efficiency with which brane kinetic energy is converted into radiation . we comment on how this suppression may naturally explain the small perturbation amplitude @xmath10 we see via observations of the cosmic microwave sky in the universe today .
recently ellis and maartens [ 1 ] reconsidered the possibility of a cosmological model in which there is no big - bang singularity , no beginning of time and the universe effectively avoids a quantum regime for space - time by staying large at all times . the universe starts out in the infinite past as an almost static universe and expands slowly , eventually evolving into a hot big - bang era . an interesting example of this scenario is given by ellis , murugan and tsagas [ 2 ] , for a closed universe model with a minimally coupled scalar field @xmath0 , which has a special form of interaction potential @xmath1 . it was pointed out that this potential is similar to what one obtains from a @xmath2 theory after a suitable conformal transformation and identifying @xmath3 with a negative @xmath4 . although the probability of these solutions is not high , the emergent universe scenario nevertheless merits attention as it solves many conceptual and technical problems of the big bang model . in this paper , we point out that the starobinsky model , the original as well as the modified version , permit solutions describing an emergent universe . the solution may be used to model varieties of cosmological scenarios consistent with the observational results , available at present and expected in the near future . thus , it may be possible to build models which avoid the quantum regime for space - time but share the good features of the standard big bang model . in the models considered in @xmath5 [ 1 ] @xmath6 [ 2 ] , a closed universe was considered . however , recent results from boomerang and wmap indicate that the universe is most likely to be spatially flat . if the universe has always been large enough , the field equations become simpler . in fact , in the starobinsky model , the field equations can be written as a second order differential equation for the hubble parameter @xmath7 , vide equation ( 7 ) . in the starobinsky model , one considers the semi - classical einstein equation , @xmath8 where @xmath9 is the vacuum expectation value of the energy momentum tensor . in the case of free , massless , conformally invariant fields the vacuum expectation value @xmath9 can be written , with the robertson walker metric as , @xmath10 where @xmath11 and @xmath12 are numbers and @xmath13 @xmath14 equation ( 2 ) , written for @xmath15 , gives the well - known trace - anomaly , indicating that the conformal invariance is broken by the regularization process . note that the two tensors @xmath16 and @xmath17 have different features . the tensor @xmath16 is identically conserved and can be obtained by varying a local action @xmath18 . however , one of the counter terms to be added to the lagrangian to regularise @xmath15 has the form @xmath19 , where @xmath20 is a logarithmically divergent constant , thus permitting the possibility of an addition of any finite number to @xmath20 . hence , @xmath11 is not determined and can be given any arbitrary value including a negative one . the tensor @xmath17 is conserved only in a conformally flat universe and it can not be obtained by variation of any local action . the constant @xmath12 is fully determined , once the fields are specified , e.g. , @xmath21 where @xmath22 gives the number of quantum fields of spin i. the freedom of choosing @xmath11 arbitrarily has led to the modified starobinsky [ 3,4 ] model , where one adds a counter term @xmath23 , and chooses @xmath24 . the theory then becomes essentially a typical @xmath25 -theory . however , the earlier work on starobinsky models were done in the context of a big bang model . the emergent universe model needs a different scenario and one makes different choices of @xmath11 and @xmath26 , as will be shown below . let us choose @xmath27 and write the evolution equation ( 1 ) of the flat frw universe as @xmath28\ ] ] which can be written in terms of the hubble parameter @xmath29 equation ( 7 ) will determine the evolution of the universe when the initial conditions are provided . we now look for a solution which describes a universe which exists eternally and always remains large so that a classical description is possible at all times . the matter , as indicated earlier , will be described by quantum field theories . we first note the following two exact solutions : @xmath30 @xmath31 is a solution but it describes an eternally static universe . it can be checked that the solution is stable against small linear perturbations , if @xmath11 is positive . however , with @xmath11 negative , the state is unstable . @xmath32 @xmath33 gives a de sitter type solution , but the solution is not stable . to get an emergent universe , we look for a solution of the suggestive form @xmath34 where the constants @xmath4 and @xmath35 will be determined from equation ( 7 ) . the constants @xmath36 and @xmath37 may be determined from initial conditions . since we have @xmath38 the function in ( 8) will be a solution , if @xmath39 and @xmath40 . thus @xmath11 will be chosen negative in this model . the fact that the emergent universe scenario is indeed permitted by equation ( 7 ) comes as a surprise . since @xmath41 and @xmath12 are related , we have a one parameter family of solutions specified by the value of @xmath12 . the general features of the solutions are \1 ) the scale factor @xmath42 has a non - zero value @xmath43 as @xmath44 , which may be chosen much larger than the planck length . the hubble parameter @xmath7 and its time derivatives @xmath45 and @xmath46 all vanish in the limit @xmath47 . thus the solution decribes an emergent universe with natural initial conditions . \2 ) if @xmath48 , the universe remains almost static during the period @xmath49 . thus during the entire infinite past of @xmath50 , the universe expands only by a factor @xmath51 for a large @xmath37 . [ fig.1 : ] for @xmath52 , the universe expands rapidly , eventually reaching an asymptotically de sitter stage . fig 1 , gives the time variation of the hubble parameter and its derivatives , for @xmath53 ( which occurs if the particle number and species correspond to the minimal su(5 ) model ) and @xmath37 is given a value @xmath54 . \3 ) it may be useful to check if the solution is stable under small linear perturbations . we write @xmath55 where @xmath56 is a small perturbation and @xmath57 is the solution obtained above . substituting this in equation ( 7 ) , we obtain the equation for @xmath56 : @xmath58 where @xmath59 @xmath60 note that @xmath61 and @xmath62 for @xmath63 . it is difficult to solve the equation ( 12 ) . however , if we assume that @xmath56 has a solution @xmath64 , we must have , for t negative , @xmath65 and @xmath66 can not be positive , since @xmath4 and @xmath37 are both positive . thus the solution seems to be stable under small perturbations at least in the negative t regime . \4 ) the evolution can be described alternatively in terms of a scalar field by substituting @xmath67 in equation ( 7 ) . this gives @xmath68 where @xmath69 with @xmath70 and @xmath71 . note that @xmath72 has the shape of a double well potential with a maximum at @xmath73 and two minima for @xmath74 , giving @xmath75 . also note that @xmath73 gives an unstable static universe while the evolution of the emergent universe is given by the path from @xmath76 to @xmath77 . however , @xmath7 here is time dependent and is determined self - consistently by the instantaneous value of @xmath0 . the equation ( 15 ) does not make the problem simpler , although it looks very familiar and attractive . \5 ) the prescribed evolution of the universe gets modified when particles are produced due to the expansion of the universe . we have considered above only massless , conformally coupled fields in conformally flat spacetimes , i.e. , the conformally trivial situation where no particles are produced . the perturbative calculations of particle production make use of small deviations from this conformal triviality . one considers either ( i ) a small mass and/or ( ii ) deviations from a conformally flat space time , say by assuming a bianchi i spacetime deviating a little from a frw spacetime or ( iii ) assuming a small non conformal coupling with @xmath78 . for simplicity , we shall consider here the case of non conformal coupling . let us consider a real massive scalar field satisfying the equation @xmath79 with @xmath80 . as the scale factor changes , the changing gravitational field feeds energy into the prturbed scalar field modes . as long as the mode frequency of the field is greater than the hubble expansion rate , a co - moving detector will not respond . modes with lower frequencies will , however , be excited . thus the presence of mass makes the particle production process less efficient . to begin with we consider the case @xmath81 and @xmath82 . zeldovich and starobinsky , starobinsky [ 3 ] , birrell and davies [ 3 ] and vilenkin [ 4 ] have studied this case . the rate of particle production per unit volume per unit conformal time @xmath83 , @xmath84 , is given by @xmath85 where @xmath86 is the scalar curvature . this gives @xmath87\ ] ] in fig . 2 , we have plotted @xmath88 against @xmath89 for @xmath90 and @xmath91 . we note that the rate of particle production peaks around @xmath92 , as expected . increase in the number of species of particles i.e. , a higher value of @xmath12 will indicate a delay in particle production and also a wider peak . the dependence of the rate of particle production on the parameter @xmath37 is shown in fig . 3 . the absolute value of the peaks depends on the value of the scale factor @xmath93 and the constant @xmath36 may be determined from the rate of particle production . although we have considered zero mass particle we expect similar results even when @xmath94 is nonzero but small ( see vilenkin [ 4 ] ) . during the subsequent evolution the particles produced thermalize and the universe enters eventually into a radiation dominated stage of the standard hot bigbang model . the details of the process of particle production and the evolution of the universe will be considered elsewhere . [ fig.2 : ] [ fig.3 : ] we have presented here a one parameter family of solutions of the starobinsky model which describes an emergent universe . the earlier work on starobinsky model was done in the context of big bang singularity and the particle production was achieved by its oscillatory solution . the present solution on the other hand describes the universe which is almost dormant during the infinite past period @xmath95 , after which it undergoes a rapid expansion , see fig . the parameters @xmath36 and @xmath37 are related with initial conditions . they also introduce a new mass scale in the process of particle production . we have not considered here any specific theory of particle interactions which will determine the subsequent evolution of the universe as well as the details of large scale structure formation . whether the present scenario can successfully explain the present observational data needs further study . however , the fact that one encounters solutions describing an emergent universe in different contexts may be a good reason for taking such solutions seriously , although the probabilities for such solutions may not be very high . sm and bcp would like to thank the university of zululand and the university of kwazulu - natal , south africa for hospitality during their visit when a part of the work was done . they would also like to thank iucaa , pune and iucaa reference centre , north bengal university for providing facilities . 000 g. f. r. ellis and r. maartens , _ the emergent universe : inflationary cosmology with no singularity _ , gr - qc/0211082 ( 2003 ) . g. f. r. ellis , j. murugan and c. g. tsagas , _ the emergent universe : an explicit construction _ , gr - qc/0307112 ( 2003 ) . b. zeldovich and a. a. starobinskii , _ jetp lett . _ * 26 * 253 ( 1977 ) ; a. a. starobinsky , _ phys . lett . _ * 91 b * 99 ( 1980 ) ; n. d. birrel and p. c. w. davies , _ j. phys . * 13 * 2109 ( 1980 ) . r. fabri and m. d. pollock , _ phys . lett . _ * 125 b * 445 ( 1983 ) ; l. a. kofman , a. d. linde and a. a. starobinsky , _ phys . lett . _ * 157 b * 361 ( 1985 ) ; a. vilenkin , _ phys . * d 32 * 2511 ( 1985 ) ; m. b. mijic , m. s. morris and w , suen , _ phys . _ * d 34 * 2934 ( 1986 ) ; s. g @xmath96ttlober and v. m@xmath97ller , _ class . quantum grav . _ * 3 * 183 ( 1986 ) ; b. c. paul , d. p. datta and s. mukherjee , _ mod . * a 3 * 843 ( 1988 ) ; g. maganano and l. m. sokolowski , _ phys . rev . _ * d 50 * 5039 ( 1994 ) ; s. w. hawking and j. c. luttrell , _ nucl . * b 247 * 250 ( 1984 ) ; a. a. starobinsky , _ sov . lett . _ * 9 * 302 ( 1983 ) , _ jetp lett . _ * 42 * 152 ( 1985 ) ; t. v. ruzmaikina and a. a. ruzmaikin , _ zh . eksp . fiz . _ * 57 * 680 ( 1969 ) ; v. ts . gurovich and a. a. starobinsky , _ sov . . jetp _ * 50 * 844 ( 1979 ) . i. prigogine , _ thermodynamics of irreversible processes _ ( new york : wiley 1961 ) . i. prigogine , _ gen . * 21 * 767 ( 1989 ) ; k. desikan , _ gen . _ * 29 * 435 ( 1997 ) ; g. montani , _ class . quantum grav . _ * 18 * 193 ( 2001 ) .	we present an emergent universe scenario making use of a new solution of the starobinsky model . the solution belongs to a one parameter family of solutions , where the parameter is determined by the number and the species ( spin - values ) of primordial fields . the general features of the model have also been studied . = 11 pacs numbers : 04.20.jb , 98.80.cq , 98.80.jk
many biological and physical phenomena are due to a complex connectivity among complex elements ; feedback mechanism seems to be a suitable way to understand such phenomena ( see for instance @xcite@xcite@xcite ) . the introduction of feedback through a control parameter leads to models involving global multiplicative coupling among maps . one of the most simple models can be given by using logistic maps . within this context , two - dimensional noninvertible maps with a global multiplicative coupling between one - dimensional logistic maps are of great interest . the understanding of their quite complex behaviour can be a first step in the understanding of dynamics in higher dimensional systems . previous works ( see @xcite@xcite@xcite@xcite ) have already been devoted to such kind of maps . in this paper , our aim is to complete the understanding of dynamics of model ( c ) given in @xcite . the considered model is the noninvertible two - dimensional map @xmath0 defined by : @xmath1 where @xmath2 is a real control parameter , @xmath3 and @xmath4 are real state variables . the map ( [ 1 ] ) includes a time asymmetric feedback . numerical studies show that the system ( [ 1 ] ) is stable when @xmath5 $ ] . section 2 is devoted to recalls about fixed points of the map ( [ 1 ] ) and their bifurcations . section 3 concerns critical manifolds . section 4 analyzes evolution of basins by considering crossing of basin boundary through critical curves . in section 5 we explain evolution of invariant closed curve issued from a fixed point and giving rise to @xmath6 , then to chaotic attractor . as explained in @xcite , @xmath0 possesses at most five different fixed points : @xmath7 @xmath8 @xmath9 @xmath10 for @xmath11 , @xmath12 is an attractive node and @xmath13 and @xmath14 are saddle points on the negative side of the axes , x axis being @xmath13 unstable manifold and y axis being @xmath14 unstable one . when @xmath15 , @xmath16 . for @xmath17 , @xmath12 is a repulsive node and @xmath13 and @xmath14 are saddle points , but located on the positive side of the axes , which are now @xmath13 and @xmath14 stable manifolds ( figures [ fig6][fig7 ] ) . when @xmath18 , @xmath12 undergoes a flip bifurcation and a period doubling cascade occurs when @xmath2 decreases . @xmath19 and @xmath20 exist when @xmath21 . when @xmath22 , @xmath23 ; for @xmath24 , @xmath19 is a saddle point and @xmath20 an attractive node , the whole diagonal segment between @xmath19 and @xmath20 is locus of points belonging to heteroclinic trajectories connecting the two fixed points . when @xmath15 , @xmath20 undergoes a nemark hopf bifurcation and becomes repulsive by giving rise to an attractive invariant closed curve ( @xmath25 ) ( @xmath26 ) . section 5 is devoted to the study of ( @xmath26 ) evolution . an important tool used to study non - invertible maps is that of critical manifold , which has been introduced by mira in 1964 ( see @xcite@xcite for more details ) . a non - invertible map is characterized by the fact that a point in the state space can possess different number of rank-1 preimages , depending where it is located in the state space . a critical curve lc , in the two - dimensional case , is the geometrical locus of points x in the state plane having two coincident preimages , @xmath27(x ) , located on a curve @xmath28 . it is recalled that the set of points @xmath29(x ) constitutes the rank n preimages of a given point x. the curve lc verifies @xmath27(lc)@xmath30@xmath28 or t(@xmath28)=lc and separates the phase plane in two regions where each point has a different number of preimages ( k in one region and k+2 in the other one , k @xmath31 ) . for the map ( [ 1 ] ) , @xmath28 is given by the cancellation of the jacobian of the map . for the map ( [ 1 ] ) , the critical manifold lc@xmath32 is given by four arcs ( figure [ fig1 ] ) : @xmath33 @xmath34 and the critical manifold lc by ( figure [ fig2 ] ) : @xmath35 @xmath36 @xmath37 @xmath38 let us call @xmath39=@xmath40 , @xmath39 is a cusp point of @xmath41 , indeed @xmath42 is constituted of two merged critical half straight lines ( @xmath43-\infty,\ \frac{5\lambda}{8}]$ ] ) , which explains the fact that , by crossing @xmath42 , one goes from @xmath44 area to @xmath45 one , or from @xmath46 to @xmath46 ( cf . figure [ fig2 ] ) . let us remark that @xmath47 . the lc curves separate the state space in different areas @xmath48 ( figure [ fig2 ] ) , where each point possesses @xmath49 distinct rank1 preimages . figure [ fig2a ] shows the state space as a foliated one , each sheet corresponding to the existence of a rank1 preimage . this foliation looks like those given for parameter plane , for instance in @xcite . previous representations of such foliation related to the existence of lc curves can be found in @xcite . the basin of an attractor @xmath50 is defined as the set of all initial conditions that converge towards an attractor at finite distance when the number of iterations of @xmath0 tends towards infinity . the set @xmath50 verifies : @xmath51 and @xmath52 . we call @xmath53 the largest connected part of a basin containing the attracting set . in this section , we study the evolution of basin of attractors when parameter @xmath2 is modified . the situations are shown in figures [ fig3 ] to [ fig17 ] . using the terminology and the results of @xcite@xcite , our aim is to explain how basins become fractal , nonconnected or multiply connected in the case of ( [ 1 ] ) . * figure [ fig4 ] shows the first modification of the basin with the creation of a @xmath54 h , due to the crossing of @xmath55 through the basin boundary . the @xmath54 h is a rank1 preimage of the area @xmath56 , which is located inside @xmath46 , when @xmath2 increases from the value 1 . * the second modification consists in a bifurcation @xmath57 @xmath58 when @xmath2 decreases from the value 1 . islands are created ( cf . figures [ fig6]-[fig7 ] ) after the bifurcation concerning the fixed points @xmath59 . indeed , when @xmath2 decreases from the value 1 , @xmath13 and @xmath14 , located on positive part of axes , go to negative part of axes ; the stable manifolds of these two saddle points is a part of the basin boundary ; when the two points are on the negative part of axes , a tongue is created in the area ( @xmath60 , @xmath61 ) of the state space and crosses through @xmath55 and @xmath42 curves to penetrate inside @xmath46 area , which is located above @xmath42 ; it gives rise to a sequence of infinitely many preimages of any rank of this part of the tongue : these are the islands , the basin becomes non connected . * the third modification of the basin is due to a sequence of @xmath62 ( figures [ fig9]-[fig11 ] ) . the first aggregation occurs when @xmath55 and @xmath42 curves become tangent in points @xmath63 and @xmath64 at the basin boundary ( figure [ fig9a ] ) , rank1 preimages of @xmath63 and @xmath64 , @xmath65 and @xmath66 , correspond to the aggregation of two islands with the immediate basin . the other preimages of @xmath63 and @xmath64 give rise to other aggregations . this first aggregation is followed by infinitely many new aggregations due to tangencies between critical curves and the immediate basin boundary when @xmath2 decreases ; such phenomena are explained more precisely in @xcite . the important point to note is that a crossing between critical curves and basin boundary makes some domains change @xmath67 area , so it makes appear or disappear preimages of the considered domains . the sequence of aggregations is nearly finished in figure [ fig11 ] . the basin size increases when the parameter @xmath2 decreases and tongues corresponding to island ends disappear when crossing critical curves of different rank ( figure [ fig12 ] ) . in figure [ fig13 ] , the basin boundary is smooth . when @xmath2 decreases , being negative , an inverse process gives rise to new appearance of tongues ( figure [ fig14 ] ) . * another modification corresponds to a bifurcation @xmath68 @xmath69 @xmath70 , that means basin with holes inside . the appearance of these holes is also related to crossing of basin boundary through critical curves ; for instance , the hole in the middle of the basin in figure [ fig15 ] is due to the crossing of @xmath55 near its left extremum ; it makes appear an area in @xmath46 region and the preimages of any rank of this area gives rise to infinitely many holes inside the basin ; these holes accumulate along the basin boundary . when @xmath2 decreases , the size of holes increases ( figure [ fig16 ] ) . some holes can open and be changed in @xmath71 ( cf . @xcite@xcite ) , this is the case in figure [ fig17 ] , the basin boundary becomes fractal . this section is devoted to the study of evolution of attractors when @xmath2 increases . this evolution can involve relations with the attractor basin or homoclinic and heteroclinic bifurcations . * when @xmath2= 1 , the attractive focus undergoes a nemark hopf bifurcation and becomes repulsive when giving rise to an attractive @xmath25 @xmath72 ( figure [ fig18 ] ) . oscillations appear on @xmath72 when it is tangent to the cusp point @xmath39 on @xmath42 ( figure [ fig19 ] ) . from @xmath73 , @xmath2 increasing , there is an alternation between existence of @xmath25 with oscillations , frequency lockings , cyclic chaotic behaviors , contact bifurcations with basin boundaries and single chaotic attractor . these results are consistent with the liapunov exponents values obtained in @xcite . we are going to explain what occurs regarding contact with basins of coexisting period3 attractors . * in the same time , two attractive period3 orbits appear simultaneously by a fold bifurcation , coexisting with two period3 orbits of saddle points ( figure [ fig19 ] ) . so , three different attractors coexist ; figure [ fig20 ] shows the basin of the @xmath25 and the basin of the two period3 attractors . when @xmath2 increases , the two period3 attractive cycles undergo a nemark hopf bifurcation and become unstable by giving rise to two period3 attractive @xmath25 around them . figure [ fig21 ] shows the three different basins of the three coexisting attractors : in this case , two period3 @xmath25 and a period29 attractive cycle , which corresponds to a frequency locking for the @xmath25 . * when @xmath2 increases , the @xmath25 appears again ; curles arise , due to crossings of @xmath72 through critical curves , and involve auto intersections of @xmath72 , the @xmath25 is then called a @xmath6 @xcite ( figure [ fig23 ] ) . the existence of curles changes the shape of the @xmath25 and generates contact with its basin boundary ; this implies the explosion of @xmath72 and makes it disappear . then , the basin of the former @xmath25 becomes part of the basin of the other attractors ( figures [ fig26 ] @xmath74 [ fig28 ] ) . * figure [ fig27 ] shows the separate basins of each part of the two period3 attractive @xmath25 , that means basins for @xmath75 are plotted . * another contact bifurcation occurs , which involves the manifolds of the period3 saddle cycles , born by fold bifurcation . this contact creates a new attractor , by mixing the two former period3 @xmath25 ( figure [ fig29 ] ) . this new attractor evolves , undergoes frequency lockings , which give rise to the appearance of cyclic chaotic attractor ( figure [ fig30 ] ) ; curles appear on the attractor and it becomes a @xmath76 ( figures [ fig36][fig38 ] ) , then it becomes a chaotic attractor ; chaos becomes stronger and stronger ( figures [ fig43][fig44 ] ) . * the chaotic attractor disappears when a contact bifurcation occurs between the attractor and its basin boundary ( figure [ fig46 ] ) . complex behaviour of a two - dimensional noninvertible map with a multiplicative global coupling between one - dimensional logistic maps , including a time asymmetric feedback , has been analyzed . the different kinds of basin bifurcations which occur when the parameter is modified and the evolution of attractors issued from nemark hopf bifurcation have been explained . the fundamental point , as previously obtained in two - dimensional noninvertible maps studies , is the position of basins and attractors , regarding critical curves . @xmath77 , @xmath57 @xmath58 , @xmath68 @xmath69 @xmath70 , change of @xmath78 into @xmath79 or @xmath80 are bifurcations in relation with critical curves . we would like to thank prof . mira for helpful discussions , in particular , the figure [ fig2a ] is due to him . r. lopez - ruiz wishes to thank the program europa of cai consi+d ( zaragoza , spain ) for financial support . gardini , l. , abraham r. , fournier - prunaret , d. , record , r. , _ a double logistic map _ , int . chaos , vol . 4 , _ 1 _ ( 1994 ) , 145176 . kaneko , k. , _ chaotic but regular posi - nega switch among coded attractors by cluster size variation _ , lett . , 63,(1989 ) , 219223 . kaneko , k. , _ spatiotemporal chaos in one - dimensional and two - dimensional coupled lattices _ , physica d , 37 , ( 1989 ) , 6084 . kaneko , k. , _ clustering , coding , switching , hierarchical ordering and control in network of chaotic elements _ , physica d , 41 , ( 1990 ) , 137172 . lopez - ruiz , r. , perez - garcia , c. , _ dynamics of maps with a global multiplicative coupling _ , chaos , solitons and fractals , vol.1 , _ 6 _ ( 1991 ) , 511528 . lopez - ruiz , r. , perez - garcia , c. , _ dynamics of two logistic maps with a global multiplicative coupling _ , int . chaos , vol . 2 , _ 2 _ ( 1992 ) , 421425 . lopez - ruiz , r. , fournier - prunaret , d. , _ complex patterns on the plane : different types of basin fractalizations in a 2-d mapping_,int chaos , vol.13 , _ 2 _ ( 2003 ) , 281310 . mira , c.,,_chaotic dynamics . from the one - dimensional endomorphism to the two - dimensional diffeomorphism _ , world scientific publishing , singapore,(1987 ) . mira , c. , fournier - prunaret , d. , gardini , l. , kawakami , h. , cathala , j.c . , _ basin bifurcations of two - dimensional noninvertible maps . fractalization of basins _ , int . chaos , vol . 4 , _ 2 _ ( 1994 ) , 343381 . mira , c.,gardini , l. , barugola , a. , cathala , j.c.,_chaotic dynamics in two - dimensional noninvertible maps _ , world scientific publishing , singapore , series a , vol . 4,(1996 ) . mira , c. , rauzy , c. , _ fractal aggregation of basin islands in two - dimensional quadratic noninvertible maps _ , int . chaos , vol . 5 , _ 4 _ ( 1994 ) , 9911019 . mira , c. , carcasses , j.p . , millerioux , g. , gardini , l. _ plane foliation of two - dimensional noninvertible maps . fractalization of basins _ , int . chaos , vol . 6 , _ 8 _ ( 1996 ) , 14391462 .	previous works have been devoted to the study of two - dimensional noninvertible maps , obtained using a coupling between one - dimensional logistic maps . this paper is devoted to the study of a specific one , in order to complete previous results @xcite @xcite , regarding the evolution of basins and attractors , when considering the tool of critical manifolds .
we acknowledge the scientific contributions and advice from jeremy bredfeldt , selezion hambir , claudio vinegoni , martin gruebele , dana dlott , amy wiedemann , and barbara kitchell from the university of illinois at urbana - champaign . this research was supported in part by the national aeronautics and space administration ( nas2 - 02057 ) , the national institutes of health ( national cancer institute ) , and the beckman institute for advanced science and technlogy .	coherent anti - stokes raman scattering ( cars ) processes are `` coherent , '' but the phase of the anti - stokes radiation is usually lost by most incoherent spectroscopic cars measurements . we propose a novel raman microscopy imaging method called nonlinear interferometric vibrational imaging , which measures raman spectra by obtaining the temporal anti - stokes signal through nonlinear interferometry . with a more complete knowledge of the anti - stokes signal , we show through simulations that a high - resolution raman spectrum can be obtained of a molecule in a single pulse using broadband radiation . this could be useful for identifying the three - dimensional spatial distribution of molecular species in tissue . functional imaging techniques have been developed to provide insight into biological processes . optical functional imaging is frequently limited because dyes or markers must be introduced that alter or damage biological tissues . because it is preferable to use endogenous properties of tissues to identify molecular contents , methods such as infrared and raman spectroscopy are used . in particular , coherent anti - stokes raman scattering ( cars ) processes have been successfully integrated with confocal scanning microscopes @xcite to achieve three - dimensional molecular images of vibrational resonances . however , existing instruments measure only the total power of the received anti - stokes radiation . we propose a novel method , nonlinear interferometric vibrational imaging ( nivi ) , which utilizes nonlinear interferometry to measure the amplitude and phase of the anti - stokes light . an experimental demonstration of the principle of this technique has been demonstrated . @xcite . this additional phase information facilitates the inference of the amplitude and phase of the nonlinear susceptibility of the molecule . by utilizing nivi with properly designed illuminating radiation , a large region of the amplitude and phase of the raman spectrum can be sampled in a single brief pulse . coherent anti - stokes raman scattering processes have only recently been used to probe biological specimens . the appeal of cars is that it can probe the density of molecules with a particular raman resonance frequency while exposing the specimen to relatively low levels of illumination . a typical cars process illuminates the specimen with a pump pulse of frequency @xmath0 , and a stokes pulse of frequency @xmath1 , which are separated by the vibrational frequency of interest @xmath2 . if a molecule with a raman resonance at frequency @xmath3 is present , an anti - stokes pulse at frequency @xmath4 is produced . in cars microscopy , tightly focused pump and stokes beams are scanned through the specimen , and the anti - stokes photon count is measured at each point . this photon count is proportional to the square of the molecular bond density and the magnitude of the raman susceptibility . nonlinear interferometry has been used to characterize the magnitude of the stimulated raman scattering nonlinearity in liquids @xcite , and also of two cars signals @xcite in gases . the model of cars with broadband pulses used here is similar to that described by @xcite . we do not assume that the illuminating radiation is narrowband . however , we stipulate that the molecule does not resonantly interact directly with any of the frequencies inside the illumination bandwidth or the generated anti - stokes bandwidth of the optical signal . the cars process is composed of two stimulated raman scattering processes involving four photons that begins and ends with the molecule in the ground state . to describe cars , we denote the electric field incident on the molecule as @xmath5 . the first process is modeled by eq . [ eq : cars1 ] , and excites the nonlinear dipole polarization of the resonant transition . @xmath6 each pair of frequencies that are separated by a resonance of the molecule at frequency @xmath3 produces a nonlinear polarization in the molecule . another way to look at eq . [ eq : cars1 ] is that , in the time domain , the molecule has a nonlinear polarization that is driven not by the electric field but by the instantaneous intensity envelope of the signal . in this formulation , we are neglecting any changes in @xmath7 that are dependent on the `` carrier '' envelope frequency on which the beats are imposed . therefore any pulse train with intensity beats at the resonance frequency will stimulate the nonlinear polarization . examples of this are pulses that are modulated periodically in the spectral domain with period @xmath3 @xcite , and interfering two relatively delayed chirped pulses to achieve a beat frequency proprortional to time delay @xcite . the second step creates anti - stokes radiation by mixing the incoming radiation field with the polarization in the time domain , and is modeled by eq . [ eq : cars2a ] . the eqs . [ eq : cars1 ] and [ eq : cars2a ] allow one to calculate the emitted cars radiation @xmath8 for a given @xmath5 and @xmath9 . while these relations do not constitute a linear relationship between @xmath10 and @xmath8 , there is a linear dependence of @xmath8 on @xmath11 given an input field @xmath5 . this suggests that with knowledge of the complex @xmath8 , one can do linear estimation to find @xmath11 . the advantage of nivi over incoherent detection is that nonlinear interferometry enables the recovery of the complex @xmath8 . with a properly designed input pulse @xmath5 , the nonlinear susceptibility can be found in a particular frequency range . nivi takes advantage of the coherent nature of the cars process to allow the phase - sensitive measurement of the anti - stokes radiation . conventional linear interferometry involves splitting a source light beam into two parts , each of which scatters linearly in the field , and which are then recombined and detected . nivi differs in that a cars process occurs to one of the split beams . nivi can be implemented with the setup detailed in fig . [ fig : fig1 ] . we start with a broadband , phase - locked source of light pulses , such as those from a mode - locked laser . sources that can produce such light are ultrabroadband mode - locked ti - sapphire oscillators @xcite , and supercontinuum sources @xcite . because the source is phase - locked , there is a deterministic relationship between phases at various frequencies . this deterministic relationship will be preserved by coherent processes such as cars . to utilize this determinacy , the source bandwidth is split into higher and lower frequency bands , as shown in fig . [ fig : fig1 ] with a dichroic beamsplitter . the higher frequency band will be a reference signal @xmath12 , and will correspond to the bandwidth of the anti - stokes frequencies produced by a sample . the lower frequency band is temporally shaped to stimulate cars in the sample , a signal we denote by the frequency spectrum @xmath5 . some of the illumination signal will be converted to anti - stokes radiation by the sample . because cars processes are usually rather weak , we will assume that any new anti - stokes radiation created in the same bandwidth as the illumination will be inseparable from the illumination . therefore we will discard all anti - stokes light inside the illumination bandwidth with a high - pass frequency filter . the remaining anti - stokes light that passes through the filter , which we denote by the spectrum @xmath13 , corresponds to frequencies in the reference signal . we combine the reference signal @xmath14 and the anti - stokes spectrum @xmath13 with a 50:50 beam splitter , and utilize balanced detection to measure the intereference component on two photodetectors . there is a delay of time @xmath15 placed in the reference path to facilitate measuring the temporal cross - correlation between the reference and anti - stokes signals . the difference in the two intensities @xmath16 as a function of delay between the signal and reference will be : @xmath17 if we call @xmath18 the fourier transform of @xmath16 with respect to @xmath15 , we find that @xmath19 . thus the measured data retains its linear relationship with respect to the anti - stokes spectrum @xmath20 and therefore the nonlinear susceptibility @xmath11 . besides the ability to find the complex - valued @xmath11 , interferometry eliminates the need for photon - counting detectors . another advantage is that interference will only occur when the anti - stokes light and the reference light arrive at the beam splitter at the same time . because of this , temporal gating can be used to produce three - dimensional vibrational images in a manner analogous to optical coherence tomography @xcite . coherent detection is also far less sensitive to stray light than photon counting . because of this , nivi may be more adaptable to various scanning configurations and environments outside the laboratory . to show that nivi can measure intervals of the raman spectrum in a single pulse , a pulse must be designed that can stimulate molecules in a broad raman spectrum . the approach we take creates beats that instead of being of a constant frequency @xcite , will be themselves chirped . this can be accomplished by combining two chirped pulses with a relative delay , but with different chirp rates . if we have a transform - limited pulsed source of center frequency @xmath21 and bandwidth @xmath22 , and we wish to sweep the beat frequency from @xmath23 to @xmath24 in time @xmath25 , we can design a pulse @xmath10 such that : @xmath26\mbox { for } \omega_0-\frac{\delta \omega}{2}<\omega<\omega_0 + \frac{\delta \omega}{2 } \\ \tilde{e_i}(\omega ) = 0 \mbox { otherwise } \\ \ \ \ \ \ \ \ \ \ \ \mbox{where } \alpha = \frac{2 \delta \omega - \omega_h - \omega_l}{2 t } , \beta = \frac{\omega_h - \omega_l}{2 t } , \mbox { and } \tau = \frac{t}{2}\left(\frac{\omega_h}{\delta \omega-\omega_h}+\frac{\omega_l}{\delta \omega-\omega_l}\right ) \end{array } \label{eq : nivi2}\ ] ] the variable @xmath27 is the common chirp to both pulses , @xmath28 is the difference chirp , @xmath29 is the time delay between the two pulses , and @xmath30 is the difference in field magnitude between the two pulses . the pulse bandwidth has been apodized with a cosine window because in practice it seems to help the stability of the inversion . note that the bandwidth of the source @xmath31 must exceed @xmath24 so that beats can be formed at all raman frequencies . when creating the pulse , the chirp time @xmath25 will control the resolution with which one will be able to resolve frequencies in the raman spectrum . the largest practical @xmath25 is determined by the dephasing time of the resonances , which in most liquids is on the order of picoseconds . to demonstrate the feasibility of nivi , we simulate the illumination of a target molecule with the broadband pulse of eq . [ eq : nivi2 ] and use the returned signal to estimate the complex susceptibility @xmath9 . we will show two simulations : one that is able to probe a wide bandwidth of raman resonances in a single pulse , and the other which is able to distinguish between two nearby resonances . we take as our hypothetical laser source a mode - locked ti - sapphire laser that can produce a pulse with a uniform bandwidth from 7001000 nm , and the setup of fig . [ fig : fig1 ] . the bandwidth from 8001000 nm will be reserved for stimulating cars , with the remainder used as the reference signal . for the first simulation , the cars excitation bandwidth will be shaped such that @xmath32 , @xmath33 , and @xmath34 . to show that the system can reconstruct several simultaneous resonances over the entire bandwidth , we create a hypothetical @xmath9 with several lorentzian resonances centered at @xmath35 , @xmath36 , @xmath37 , and @xmath38 . these frequencies are in the raman `` fingerprint '' region and would be useful for practical molecular identification . the simulation was implemented by sampling the spectra of @xmath9 , @xmath5 , and @xmath13 with 20,000 points spaced at equal intervals from @xmath39 to @xmath40 in @xmath41 steps . the cross - correlations of eqs . [ eq : cars1 ] and [ eq : cars2a ] were computed using the fast fourier transform . these two equations form a `` forward '' cars linear operator computing @xmath11 from @xmath8 which we call @xmath42 . to find the inverse of this operator , we used the tikhonov - regularized least - squares inversion operator , which is formally denoted by @xmath43 . the tikhonov regularization was included to improve the stability of the inverse and to account for potential noise sources such as thermal and photon noise in a practical experiment . the constant @xmath44 is chosen to account for the magnitude of additive white gaussian noise in a realistic experiment . in practice , @xmath45 was computed using the preconditioned conjugate gradient @xcite method to avoid the very computationally expensive direct matrix inversion . while we do not model a real noise source here , tikhonov regularization adjusts the inverse operator such that features of the estimated spectrum @xmath11 that would be unstable due to insufficient information for reconstruction would tend towards zero . the left column of figure [ fig : fig3 ] shows the temporal and spectral shapes of the input pulse . part ( a ) shows two chirped pulses that partially overlap , producing a beat pattern that stimulates the resonance . part ( b ) shows the power spectrum of the input pulse . part ( c ) shows the anti - stokes radiation spectrum , that is calculated using eqs . [ eq : cars1 ] and [ eq : cars2a ] . because the excitation light is assumed to be much more powerful that the anti - stokes light , we filter out all of the excitation bandwidth and utilize only wavelengths shorter than 800 nm for the inverse . the right column of figure [ fig : fig3 ] shows the original and reconstructed @xmath9 . part ( d ) is the magnitude of the spectrum of the intensity of the original pulse , _ i.e. _ the beat frequencies of the pulse . it shows the possible measurable raman frequencies with this pulse . part ( e ) shows the original @xmath11 raman spectrum magnitude . finally , part ( f ) is the tihkonov- regularized least - squares reconstructed @xmath11 based on only the anti - stokes frequencies from 700800 nm . in simulation , all of the spectral lines can be recovered . the minimum discernible separation in raman frequencies tends to increase as the raman frequency decreases because the anti - stokes radiation created by lower frequency resonances tends to overlap the original spectrum more . as a second demonstration with two closely spaced raman lines , we consider deoxyribonucleic acid ( dna ) , which would be contained in the nucleus of a cell , and ribonucleic acid ( rna ) located throughout the cell . both macromolecules have @xmath46 phosphodiester resonances , but the resonance occurs in dna at @xmath47 and in rna at @xmath48 . to show that a properly designed pulse can recover both resonances distinctly , we design a @xmath11 that has resonances both at @xmath49 and @xmath48 , that could be created by mixing dna and rna . to probe this mixture , we create a pulse using eq . [ eq : nivi2 ] with @xmath50 , @xmath51 , and @xmath52 . the results of this simulation are shown in figure [ fig : fig6 ] . figure [ fig : fig6 ] shows the beat frequency spectrum , and original and reconstructured @xmath9 raman spectra . while the reconstructed lines are broadened , they are still quite distinct and would be useful for discerning the two molecules . nivi is a flexible tool utilizing ultrafast pulses that can measure small or large portions of a raman spectrum of a molecule in a single pulse . it does so by interferometrically measuring the anti - stokes radiation from a molecule , stimulated by beats in intensity of an excitation field . from this anti - stokes field , the complex raman susceptibility can be estimated . it is especially suited to biological imaging because while the pulse energy can be large , the peak power can remain small by chirping the pulse . for these reasons , we believe that nivi can be a general tool for noninvasively probing the molecular content of biological tissues .
surveys in the optical are known to be affected by strong selection effects ( e.g. disney 1976 ; disney & phillipps 1983 ; mcgaugh 1996 ) which bias our understanding of the local galaxy population . corrections for these selection effects can be made , but these are large and controversial and must be applied to small numbers of sources , leading to large uncertainties ( impey & bothun 1997 ; disney 1999 ) . low surface - brightness ( lsb ) galaxies could make a significant contribution to the universe that would not be recognised in an optical survey . they may dominate the luminosity , baryon or mass density of the galaxies in the universe ( e.g. fukugita , hogan , & peebles 1998 ) and could contribute significantly to qso absorption line spectra ( e.g. churchill & le brun 1998 ) . zwaan et al . ( 2003 ) derived the hi mass function for the 1000 galaxies in the hi parkes all sky survey ( hipass ) bright galaxy catalogue ( koribalski et al . 2004 ) and put a value of 15 per cent on the contribution of lsb galaxies to the neutral hydrogen density of the universe . however , this number depends crucially on their assumption that optical catalogues are as ( in)complete for lsb galaxies as for high surface - brightness ( hsb ) galaxies . this is obviously not the case as lsb galaxies are much harder to detect than hsb galaxies . indeed , most of the new galaxies , outside of the zone - of - avoidance , found in the hipass bright galaxy catalogue are lsb galaxies ( ryan - weber et al . thus the value of zwaan et al . should be considered only a lower limit to the contribution of lsb galaxies to the neutral hydrogen density . the idea of searching for galaxies via the 21-cm neutral hydrogen line has long been considered as an alternative to optical surveys ( e.g. disney 1976 ) . however , until recent advances in technology such as multibeam receivers and powerful correlators , such a survey has not been practical . the hideep survey ( minchin et al . 2003 ; paper 1 hereafter ) was motivated by the desire to reach previously inaccessible surface - brightness levels . if optical surface - brightness ( e.g. luminosity per square arcsec ) were to correlate with hydrogen column - density ( e.g. hi flux per square arcsec ) , then to reach low surface - brightnesses it would be necessary to reach low column - densities . with an integration time of 9000 seconds per beam , hideep is significantly more sensitive to low column - density gas than any previous survey and thus potentially more sensitive to low surface - brightness galaxies . hideep followed the same survey strategy and data - reduction path as hipass ( barnes et al . 1998 ) , but with twenty times the integration time . the parkes multibeam system was actively scanned in declination to cover a @xmath10 region , with full sensitivity being reached over the central @xmath11 . scans were interleaved to give nyquist sampling and repeated to reach the required depth . the data were treated in the same way as hipass data : they were reduced in aips++ using the livedata and gridzilla programs and continuum sources were removed by template - fitting with the luther program . in order to minimise the noise , all the observations were carried out at night . the full sensitivity reached was 3.2 mjy beam@xmath12 , which is consistent with a @xmath13 improvement on hipass . to complement the deep hi data , we obtained eight 1-hour exposures on tech pan films at the uk schmidt telescope ( ukst ) . these covered a sky area of @xmath14 , with the same centre as the hi survey data . the films were digitised using the supercosmos machine at the royal observatory , edinburgh ( hambly et al . digital stacking of these films brings a @xmath13 improvement in signal to noise ( bland - hawthorn , shopbell & malin 1993 ) , thus gaining us a little over a magnitude in depth . the use of tech pan films gives a further magnitude improvement over the iii - af _ r_-band plates previously used at the ukst ( parker & malin 1999 ) . for galaxies with ordinary colours , our @xmath15 surface - brightness limit of 26.5 _ r _ mag arcsec@xmath1 is equivalent to @xmath16 _ b _ mag arcsec@xmath1 . the bandpass limit of the multibeam system means that only objects with a heliocentric velocity less than 12,700 kms@xmath12 can be detected ( 173 mpc for h@xmath17 = 75 kms@xmath12 mpc@xmath12 , after correction to cmb rest frame velocities ) , thus the volume of the 36 square degree optical overlap region is 19,000 mpc@xmath18 . within this volume we find a total of 129 sources , of which 107 can be identified with individual galaxies . in paper 1 , we identified sources found in the hi data with optical counterparts on the tech pan films . a comparison of the offsets of the optical positions from the hi positions for the sources with secure identifications ( those with catalogued velocities matching the hi velocity and those confirmed by hi interferometry or optical spectroscopy follow - up observations ) with the offsets for less certain counterparts ( those without catalogued velocities or follow - up observations ) showed no significant difference . of the 107 galaxies in this paper , 87 are secure identifications ; this includes 18 confirmed with our interferometric follow - up and 2 with our spectroscopic follow - up . of the remaining 20 , 18 are previously catalogued galaxies that do not have redshifts in the literature and two are new detections . an hi - selected sample is expected to appear different from an optically - selected one . in particular low surface - brightness galaxies which are seen over much smaller volumes than high surface - brightness galaxies with the same total luminosity in optical surveys ( disney & phillipps 1983 ) are expected to be seen much more easily . however , gas - poor galaxies , such as ellipticals and dwarf spheroidals , which are found by optical surveys will be invisible at 21-cm . we find that this is indeed the case with our sample . as expected , our surface - brightness distribution ( figure [ sbhist ] ) is higher towards lower surface - brightnesses than one finds in optically selected samples such as the eso - lv ( lauberts & valentijn 1989 ) , which is also shown in the figure . a kolmogorov - smirnov test on our observed surface - brightness distribution and the surface - brightness distribution of the eso - lv , which also uses _ r_-band effective surface - brightness , shows that the hypothesis that both are drawn from the same parent population has a significance of less than 1 per cent , due to the larger number of lsb galaxies seen in the hideep sample . this confirms that hi surveys do , as expected , avoid some of the surface - brightness selection effects present in optical surveys . optical photometry of hideep galaxies was carried out using a combination of sextractor ( bertins & arnouts 1996 ) and the stsdas package in iraf , where the ellipse task was used to obtain radial surface - brightness profiles . the measured optical parameters are displayed in table 1 ( a sample is shown here , the full table is available in the online edition ) where column ( 1 ) gives the source name and ( 2 ) the catalogued identification if there is one . column ( 3 ) gives the type of the identification graded into 3 classes : `` i d '' denotes objects where both positions and velocities coincide , `` pcg '' ( for previously catalogued galaxy ) where there is a positional coincidence with a previously catalogued galaxy in the nasa extragalactic database ( ned ) and `` pug '' ( for previously uncatalogued galaxy ) where there is no possible counterpart in ned , and thus no previous positional or redshift data . three further sub - classes , `` id - vla '' , `` id - atca '' , and `` id - spec '' denote where a previously catalogued galaxy without a redshift has been confirmed as the optical counterpart through our follow - up observations ( by interferometry with the vla or the atca or by spectroscopy at the anu 2.3-m telescope at siding spring respectively ) . the tabulated optical parameters are as follows : column ( 4 ) gives the apparent magnitude ( @xmath19 ) as measured by the _ mag___best _ parameter in sextractor . this takes the form of either an adaptive aperture magnitude ( kron 1980 ) , where the size of the aperture is 2.5 times the first - moment radius of the galaxy , or , for crowded fields , an isophotal magnitude ( to the 26 @xmath0 mag arcsec@xmath1 isophote ) with a correction made for the part of the galaxy beyond the isophotal limit ( maddox et al . 1990 ) , the best algorithm being selected automatically by the program . simulations by bertin & arnouts ( 1996 ) of an _ r_-band ccd image with an isophotal limit of 26 _ r _ mag arcsec@xmath1 ( i.e. very similar to our image ) show that , to the magnitude level of the faintest galaxy in our survey ( 18.6 _ r _ mag ) , the true magnitude is recovered to better than 0.1 mag . column ( 5 ) gives the effective radius , @xmath20 the radius enclosing half the _ r_-band light . column ( 6 ) gives the effective surface - brightness @xmath21 the _ r_-band surface - brightness at the effective radius . this gives a model - independent measure of the surface brightness which can be applied to all types of galaxies . column ( 7 ) gives the sextractor ellipticity ( @xmath22 ) and ( 8) the estimated _ r_-band absorption at the position of the object from schlegel et al . ( 1998 ) supplied by ned . column ( 9 ) gives the absolute magnitude @xmath23 , including corrections for galactic absorption , cosmological dimming and internal absorption ( estimated using @xmath24 , where @xmath25 is calculated from @xmath26 ) . the k - correction out to the bandpass limit of 12,700 kms@xmath12 is negligible . we use the distances from paper 1 except for galaxies in the centaurus a group ( taken to be those within 1000 kms@xmath12 ) , where we use the distance to the m83 subgroup of 4.5 mpc from thim et al . column ( 10 ) gives the physical effective radius @xmath27 in kpc . finally column ( 13 ) gives the surface brightness @xmath28 corrected for @xmath29 cosmological dimming , inclination ( estimated using @xmath30 from graham & de blok ( 2001 ) ) and galactic absorption . table [ optpos ] gives the positions of the optical counterparts of the pugs . where these have been confirmed by interferometric follow up , this is indicated by a code in column ( 4 ) giving the run in which they were detected . the pugs that were not detected in earlier runs were followed up again in later runs , with the result that there are no galaxies in the table that have been followed - up but not detected , although there are two sources which have not been followed - up interferometrically . .positions of newly - discovered galaxies in the hideep optical area . for those detected interferometrically , column ( 4 ) gives the code for the observing run : a1 - atca , 11/1999 ; a2 - atca , 01/2000 ; v - vla , 01 - 03/2003 ; a3 - atca , 04/2003 . those without a code have not been observed with an interferometer . [ cols= " < , < , < , < " , ] table [ importantlsb2 ] summarises the findings of this section . it can be seen that lsb galaxies make up over half of all gas - rich galaxies , yet have less than 10 per cent of the luminosity density . the relationship between dynamical @xmath31 and surface - brightness means that luminosity is a biased indicator of the cosmological significance of lsb galaxies . similarly , the higher hi mass - to - light ratios of lsb galaxies means that they have more gas than would be indicated by their light on a straight extrapolation of @xmath32 from freeman - law galaxies . the relatively larger sizes of lsb galaxies means that their contribution to cross - sectional area , around 40 per cent , is much higher than would be expected from their luminosity or even from their hi mass . the totals given here are only for galaxies with @xmath33 . these are almost entirely spiral galaxies . dwarf galaxies , even gas - rich ones , tend to have lower gas masses than this , while elliptical galaxies are too gas - poor to be detected . most dwarf galaxies have low surface - brightnesses , therefore if these were included it is likely that the total contributions from lsb galaxies would rise . these numbers should therefore be seen as lower estimates for the total contribution of lsb galaxies . this survey does not find any lsb ` crouching giant ' galaxies , such as malin 1 . from this , we can rule out lsb galaxies making up more that 6 per cent of the population of luminous ( @xmath34 ) galaxies . indeed , there does seem to be a minimum surface - brightness at every magnitude level given approximately by @xmath35 ( figure [ bbd ] ) . to higher surface - brightnesses , galaxies seem to populate fairly evenly a band approximately 4 magnitudes wide in luminosity and 2 in surface - brightness . beyond this , high surface - brightness , low - luminosity galaxies also appear to be rare in the gas - rich population . the populated band broadens slightly towards lower luminosities , giving it an approximate slope of @xmath36 ( where @xmath37 is surface brightness in luminosity per unit area and @xmath38 is the total luminosity ) . from this , it can be calculated that the radius , @xmath0 , is related to the surface - brightness as @xmath39 , i.e. the radius only changes very slightly with surface - brightness . once volumetric corrections are made , the number of galaxies per unit volume is flat as we go to lower surface - brightnesses . even within the surface - brightness limit reached by this survey , lsb galaxies contribute over half of the number density of galaxies . furthermore , we find that lsb galaxies contribute approximately 30 per cent of the neutral - hydrogen density , twice the lower limit set by zwaan et al . lsb galaxies may also contribute around 20 per cent of the total mass density ( again with large errors ) , but only 7 per cent of the luminosity - density . overall , hi - rich lsb galaxies do not appear to contribute much to the universe . however , the cross - section to qsoals made by lsb galaxies , which is around 40 per cent of the total cross - section , is disproportionate to their luminosity or baryonic content . this implies that lsb galaxies could contribute substantially to those qsoals where galaxies are the absorbers . future work in progress will see an order of magnitude larger sample assembled from the overlap between hipass and the sdss . this will give not only an increase in the significance of the results but will also give 5-colour optical data , thus greatly extending our knowledge of hi - selected galaxies . a larger sample of hi - selected samples from the sdss may be assembled in the medium term by the proposed alfalfa drift - scan survey with the arecibo l - band feed array or in the longer term by surveys carried out with the square kilometre array . however , as neither hipass nor alfalfa reach column - density levels as low as hideep it is unlikely that they will turn up any significant population of extremely lsb galaxies that we have missed . this is not to say that extremely lsb galaxies do not exist , just that if they do make up a significant population then the amount of hi they contain must be small . in order to search for these galaxies , techniques other than hi surveys will be necessary . the optical channel available to us on the ground is fundamentally unsuited to looking for lsb galaxies , but a quite modest switch in wavelengths to either side of the solar peak , which is possible from space , could yield dramatic results . oconnell ( 1987 ) has discussed the idea of using uv wavelengths , while in the @xmath40-band at 1.6 @xmath41 m the contrast between the sun s scattered zodiacal spectrum and the light from red stars in lsb galaxies could be two orders of magnitude higher than the contrast from the ground . wide - field near - ir cameras , such as the wfc-3 instrument currently awaiting shipment to the hst , will be needed to exploit this window which is , in principle , capable of revealing galaxies 7 magnitudes lower in surface - brightness than we can detect from the ground . the authors would like to thank lister staveley - smith , tony fairall , david barnes , jon davies and suzanne linder for useful discussions . rfm , wjgdeb and pjb acknowledge the support of pparc . bgk acknowledges the support of the australian research council . the authors would also like to thank the staff of the csiro parkes and narrabri observatories , of the nrao array operations center and of the anu siding spring observatory for their help with observations . we acknowledge pparc grant gr / k/28237 to mjd towards the construction of the multibeam system and pparc grants ppa / g / s/1998/00543 , ppa / g / s/1998/00620 , ppa / g / s/2002/00053 and ppa / g / s/2003/00085 to mjd towards its operation . we would particularly like to thank the anonymous referee for an extremely helpful report . this research has made use of the nasa / ipac extragalactic database ( ned ) which is operated by the jet propulsion laboratory , caltech , under agreement with the national aeronautics and space administration . this research has also made use of the digitised sky survey , produced at the space telescope science institute under us government grant nag w-2166 and of nasa s astrophysics data system bibliographic services bertins e. , arnouts s. , 1996 , a&as , 117 , 393 bland - hawthorn j. , shopbell p. l. , malin d. f. , 1993 , aj , 106 , 2154 blanton m. r. et al . 2001 , aj , 121 , 2358 bowen d. v. , tripp t. m. , jenkins e. b. , 2001 , aj , 121 , 1456 boyce p. j. , phillipps s. , 1995,a&a , 296 , 26 briggs f. h. , rao s. , 1993 , apj , 417 , 494 chen h .- w . , lanzetta k. m. , webb j. k. , barcons x. , 1998 , apj , 498 , 77 churchill c. w. , le brun v. , 1998 , apj , 499 , 677 davies j. i. , 1990 , mnras , 244 , 8 de blok w. j. g. , mcgaugh s. s. , van der hulst j. m. , 1996 , mnras , 283 , 18 de jong r. s. , 1996a , a&a , 313 , 45 de jong r. s. , 1996b , a&a , 313 , 377 disney m. j. , 1976 , nature , 263 , 573 disney m. j. , 1999 , in davies j. i. , impey c. , phillipps s. , eds , asp conf . 170 , the low surface brightness universe , asp , san francisco , p. 9 disney m. j. , phillipps , s. , 1983 , mnras , 205 , 1253 driver s. p. , 1999 , apj , 576 , l69 freeman k. c. , 1970 , aj , 160 , 811 fukugita m. , hogan c. j. , peebles p. j. e. , 1998 , apj , 503 , 518 graham a. w. , de blok w. j. g. , 2001 , apj , 556 , 177 hambly n. c. , macgillivray h. t. , read m. a. , tritton s. b. , thomson e. b. , kelly b. d. , morgan d. h. , smith r. e. , driver s. p. , williamson j. , parker q. a. , hawkins m. r. s. , williams p. m. , lawrence a. , 2001 , mnras , 326 , 1279 henning p. a. , staveley - smith l. , ekers r. d. , green a. j. , haynes r. f. , juraszek s. , kesteven m. j. , koribalski b. , kraan - korteweg r. c. , price r. m. , sadler e. m. , schrder a. , 2000 , aj , 119 , 2686 impey c. d. , bothun g. d. , 1997 , ara&a , 35 , 367 kilborn v. a. , 2001 , phd thesis , university of melbourne koribalski , b. s. et al . , 2004 , aj , 128 , 16 kron r. g. , 1980 , apjs , 43 , 305 lauberts a. , valentijn e.a . , 1989 , the surface photometry catalogue of the eso - uppsala galaxies , eso , garching linder s. m. , 1998 , apj , 495 , 637 linder s. m. , 2000 , apj , 529 , 644 mcgaugh s. s. , 1996 , mnras , 280 , 337 mcgaugh s. s. , schombert j. m. , bothun g. d. , de blok w. j. g. , 2000 , apj , 533 , 99 maddox s. j. , efstathiou g. , sutherland w. j. , 1990 , mnras , 246 , 433 masters k .l . , haynes m. p. , giovanelli , r. , 2004 , apjl , accepted minchin r. f. , 1999 , pasa , 16 , 12 minchin r. f. et al . 2003 , mnras , 346 , 787 ( paper 1 ) oconnell r. , 1987 , aj , 94 , 867 parker q. a. , malin d. f. , 1999 , pasa , 16 , 288 phillipps s. , disney m. j. , davies j. i. 1993 , mnras , 260 , 453 rosenberg j. l. , schneider s. e. , 2002 , apj , 567 , 247 schlegel d. j. , finkbeiner d. p. , davis m. , 1998 , apj , 500 , 525 schneider s. e. , spitzak j. g. , rosenberg j. l. , 1998 , apj , 507 , 9 thim f. , tammann g. a. , saha a. , dolphin a. , sandage a. , tolstoy e. , labhardt l. , 2003 , apj , 590 , 256 turnshek d. a. , rao s. , nestor d. , lane w. , monier e. , bergeron j. , smette a. , 2001 , apj , 553 , 288 zwaan m. a. , van der hulst j. m. , de blok w. j. g. , mcgaugh s. s. , 1995 , mnras , 273 , l35 zwaan m. a. , briggs f. h. , sprayberry d. , sorar e. , 1997 , apj , 490 , 173 zwaan m. a. et al . 2003 , mnras , 125 , 2842	we have placed limits on the cosmological significance of gas - rich low surface - brightness ( lsb ) galaxies as a proportion of the total population of gas - rich galaxies by carrying out a very deep survey ( hideep ; minchin et al . 2003 ) for neutral hydrogen ( hi ) with the parkes multibeam system . such a survey avoids the surface - brightness selection effects that limit the usefulness of optical surveys for finding lsb galaxies . to complement the hideep survey we have digitally stacked eight 1-hour @xmath0-band tech pan films from the uk schmidt telescope covering 36 square degrees of the survey area to reach a very deep isophotal limit of 26.5 @xmath0magarcsec@xmath1 . at this level , we find that all of the 129 hi sources within this area have optical counterparts and that 107 of them can be identified with individual galaxies . we have used the properties of the galaxies identified as the optical counterparts of the hi sources to estimate the significance of lsb galaxies ( defined to be those at least 1.5 magnitudes dimmer in effective surface - brightness than the peak in the observed distribution seen in optical surveys ) . two different methods of correcting for ease - of - detection do not yield significantly different results : lsb galaxies make up @xmath2 per cent of gas - rich galaxies by number according to our first method ( weighting by hi mass function ) , which includes a correction for large scale structure , or @xmath3 per cent when calculated by our second method ( @xmath4 correction ) . we also find that lsb galaxies provide @xmath5 per cent of the contribution of gas - rich galaxies to the neutral hydrogen density of the universe , @xmath6 per cent of their contribution to the luminosity density of the universe ; @xmath7 of their contribution to the baryonic mass density of the universe , @xmath8 per cent of their contribution to the dynamical mass density of the universe and @xmath9 per cent of their cross - sectional area . we do not find any ` crouching giant ' lsb galaxies such as malin 1 , nor do we find a population of extremely low surface - brightness galaxies not previously found by optical surveys . such objects must either be rare , gas - poor or outside the survey detection limits . surveys galaxies : luminosity function radio lines : galaxies galaxies : fundamental parameters
by observing the temperature anisotropy of cosmic microwave background ( cmb ) , the wmap collaboration reported that the density parameter of dark matter ( dm ) or cold dark matter ( cdm ) @xmath0 is @xmath1 at 68 % c.l . @xcite where @xmath2 is the normalized hubble constant . one of the most popular candidates for dm is the lightest supersymmetric particle ( lsp ) in supersymmetric theories with r - parity . recent studies showed that big bang nucleosynthesis ( bbn ) puts interesting limitations on the candidates of lsp dm . a long - lived next lsp ( nlsp ) whose lifetime is longer than @xmath3 sec is dangerous for the successful bbn because the decaying nlsp produces a lot of ( high - energy ) daughter particles such as photons and hadrons during / after bbn , which can destroy @xmath4he ( and d ) and non - thermally produce another light elements copiously @xcite . in particular , a non - thermal hadron emission due to the decaying nlsp is severely constrained by the observational light - element abundances of d , @xmath4he , @xmath5li , and @xmath6li @xcite . in the popular case of the neutralino ( gaugino or higgsino ) lsp , as is well - known , the unstable gravitino decay puts stringent upper bound on the reheat temperature after primordial inflation , which could be problematic for various cosmological considerations , e.g. , of the inflation models . this problem can be avoided if the gravitino is the lsp . ( see also @xcite for constraints on the reheat temperature in the gravitino lsp scenario ) . however , the decays of ordinary supersymmetric particles to the gravitino turn out to be dangerous as well . the case of neutralino nlsp , because the branching ratio into hadrons is close to the order of @xmath7 , is almost excluded @xcite by considering bbn constraints @xcite . slepton nlsp scenarios appear much more favorable than neutralino nlsp scenarios in cosmology . although the sneutrino nlsp scenario should be possible because of the smaller hadronic and radiative decaying modes , the primordial abundance of the sneutrino is too small to satisfy the dark matter density for its daughter lsp particle @xcite . therefore , we would need another production mechanism of the lsp in cosmology , which usually requires a fine tuning . recently , the stau nlsp scenario has been discussed a lot as there are specifically severer / more interesting problems in bbn related to the long - lived negatively charged particle @xcite . see also long - term problems discussed in literature @xcite and topics in doubly - charged particles @xcite . if the lifetime of the negatively charged particle @xmath8 such as @xmath9 is longer than 10@xmath10 sec , bound states with ambient light elements can be produced . in particular , the bound state with @xmath4he denoted by ( @xmath4he,@xmath8 ) , which is produced after 10@xmath11 sec , significantly enhances the reaction rate d+(@xmath4he,@xmath8 ) @xmath12 @xmath13li + @xmath8 + @xmath14 @xcite by a factor of @xmath15 @xcite . then @xmath13li is over - produced by the catalyzed bbn , and this scenario is severely constrained by observations @xcite . various attempts @xcite have been made to tackle this problem and realize a successful dark matter production mechanism . in this paper , we suggest the `` flaxino '' dm as a viable option to avoid the above - mentioned difficulties for the neutralino or slepton nlsp . the flaxino is an axino - like fermion appearing in flat - direction axion models where the spontaneous breaking of peccei - quinn symmetry comes from soft supersymmetry breaking and thus all of the fermionic partners of the symmetry - breaking fields have mass of order the gravitino mass . the flaxino has a coupling to the ordinary superparticles suppressed by the axion scale @xmath16 and thus attractively leads to a fast decay of nlsp , which should be compared with the gravitino coupling suppressed by the planck scale . the strong cp problem can be naturally resolved by assuming the peccei - quinn @xmath17 symmetry spontaneously broken at the usual high scale , @xmath18 gev , and thus leads to an axion of the ksvz @xcite or dfsz type @xcite . in the supersymmetric standard model , the realization of the dfsz axion model nicely leads to the origin of the @xmath19 term @xcite . furthermore , the axion scale @xmath16 appears as the geometric mean of the planck scale and soft supersymmetry breaking scale by invoking an almost flat potential for the peccei - quinn symmetry breaking fields @xcite . for instance , let us introduce two singlet fields @xmath20 with appropriate peccei - quinn charges to allow @xcite @xmath21 where @xmath22 is the reduced planck scale . then , the @xmath19 term and axion scale can be related by @xmath23 and @xmath24 where @xmath25 gev is the typical soft mass scale ( correctly @xmath26 in this model ) . this is below the scale @xmath27 gev required to make the axion the cdm . this model contains two scalar flatons @xmath28 , one pseudoscalar flaton @xmath29 , and two flatinos @xmath30 . this kind of flat dfsz axion models lead to thermal inflation @xcite and the model parameters are rather severely constrained in order to avoid the over - production of unwanted relics like axions or the lsp after thermal inflation @xcite . an interesting collider implication of such models has been pointed out in ref . the lsp can be some linear combination of the fermionic parts of @xmath31 and @xmath32 , which we now call `` flaxino '' ( denoted by @xmath33 ) , and then the usual neutralino or stau nlsp decays to the flaxino mainly through its mixing with higgsinos driven by the @xmath19 term . although the corresponding decay lengths can be larger than the size of detectors , the copious production of the nlsps in future colliders may enable us to observe their decays for the low axion scale @xmath34 gev @xcite . the general idea of the axino dm has been put forward in ref . @xcite ( see also a recent related paper @xcite ) . some of the current authors have investigated the scenario of the `` flaxino '' lsp and the neutralino nlsp in the flat dfsz axion models using gravity - mediated susy breaking @xcite . the axino lsp in ksvz type models has been considered in ref . the ksvz type models ( hadronic axion models ) do not agree with the framework of the flat - direction models because of the overproduction of the flaxino abundance , which is excluded by bbn @xcite . extending the analysis of ref . @xcite where the neutralino nlsp is assumed , we will consider the possibility of the stau nlsp which is also allowed in a large parameter space of the mssm with the minimal supergravity scheme . let us first note that the dfsz models predict a shorter lifetime of the stau nlsp compared to the ksvz models since the stau axino / flaxino coupling arises at tree level in the former case and at one - loop level in the latter . as we will see later , even in the case of @xmath35 gev , the stau lifetime does not exceed 10@xmath11 sec for typical mass scales of susy particles in the dfsz axion models and does not induce the @xmath13li over - production problem . this is a distinct feature of the dfsz axion models . recall that the usual upper bound , @xmath36 gev , can be relaxed in late - time entropy production scenarios such as thermal inflation models which we are considering at present . from the viewpoint of detectabilities of a superparticle at large hadron collider ( lhc ) , such a long - lived charged particle is promising as it can be easily traced within the detector . moreover , quite recently refs . @xcite have discussed aggressive future plans to place new stopper , e.g. , 5 - 10 meter wall made of iron , water tanks , rocks and so on , outside regular detectors in lhc such as atlas or cms to electromagnetically stop the produced staus . if these plans are realized , we will be able to measure the stau s lifetime even if its decay length is longer than the size of the detectors ( @xmath37 m ) , and reconstruct the track of the neutral lsp by catching other daughter particles . then , according to a similar discussion in brandenburg et al s analysis @xcite , observing three - body decay of the staus will enable us to distinguish between signals of the axino / flaxino lsp and the gravitino lsp . we refer to @xcite for another topics of axino dm and gauge - mediation supersymmetry breaking models , and ref . @xcite for general cosmological constraints on a decaying saxion or flaton . see also @xcite for an inverted mass case such as the axino nlsp and the neutralino lsp , and @xcite for the gravitino nlsp and the axino lsp in gauge - mediated supersymmetry breaking models . following ref . @xcite , we calculate the lifetime of the stau nlsp which decays into the flaxino lsp @xmath33 in the dfsz axion models . for the sake of simplicity we show only mixing between @xmath38 and the usual four neutralino components ; bino , wino , and two higgsinos . this case is sufficient if we are not interested in the heavier flaxino @xmath39 whose mass is larger than @xmath40 . then , we get @xmath41\end{aligned}\ ] ] where @xmath42 is the lightest stau mass ( @xmath43 in this paper ) , @xmath44 , @xmath45 with @xmath46 and @xmath47 - c_{\tilde{f}}^ { * } g n_{03}^ { * } m_{\tau}/\sqrt{2 } c_{\beta}m_{w},\end{aligned}\ ] ] @xmath48 here @xmath49 and @xmath50 are the masses of the tau lepton and the weak boson , @xmath51 and @xmath52 are the weak and electromagnetic coupling constants , respectively , and the z - boson couplings to quarks and leptons for the tau lepton are give by @xmath53 . the matrix elements @xmath54 ( @xmath55=1,2.3,4 ) are flaxino neutralino components of a unitary matrix @xmath56 which is the 5@xmath575 mass matrix of four neutralinos plus one flaxino , @xmath58 ( @xmath59=0,1,2.3,4 ) in the @xmath60 basis . it can be diagonalized as @xmath61 where @xmath62 here @xmath63 , @xmath64 , @xmath65 , and @xmath66 with the weinberg angle @xmath67 , and @xmath68 with the electroweak scale @xmath69 = 264 gev , @xmath70 , the mixing angle @xmath71 between @xmath33 and @xmath72 which is expressed by using @xmath73 to be @xmath74 and @xmath75 @xcite . the mass eigenstate labels follow the convention : @xmath76 ( note that @xmath77 can be larger than @xmath78 although we did not write it down explicitly in ( [ eq : m5 ] ) ) . in the stau sector , the diagonalization is taken as follows ; @xmath79 because the mixing angle in @xmath80 and @xmath81 is an unknown parameter , we will study the @xmath80 dependence by changing it from zero to unity . note that the stau decay rate is approximately represented by @xmath82 for @xmath83 gev and @xmath84 gev with a suppression factor @xmath85 caused by electroweak couplings , mixing angles , kinematic suppressions and so on . thus , we can expect the decay signals of staus within the meter - size detectors in lhc . the flaxino lsp abundance is determined by the thermal generation and the non - thermal generation from the stau nlsp decay : @xmath86 first , the freeze - out density of the relic stau can be calculated by taking the leading @xmath87wave stau annihilation to photons whose thermal - averaged cross section is given by @xmath88 with the fine structure constant @xmath89 . then , following the standard procedure presented in @xcite , the non - thermal density of @xmath38 can be estimated as@xmath90 by using the analytical calculation in ref . @xcite , the density parameter of the thermal component of @xmath38 is estimated to be @xmath91}.\end{aligned}\ ] ] here @xmath92 , with @xmath93 the squark mass and @xmath94 the reheat temperature after the lightest flaton @xmath95 decay , @xmath96 with @xmath97 the branching ratio of @xmath95 decaying into axions . @xmath94 should be sufficiently high for daughter axions produced by the decay of @xmath95 not to induce @xmath4he overproduction ( @xmath98 @xcite ) and must be also larger than @xmath99 to thermally produce the nlsp staus . the quantity @xmath100 denotes the squark - quark - flaxino mixing which becomes the largest for the stop having an order - one top yukawa coupling @xmath101 : @xmath102 with @xmath103 . by assuming the superpotential given in ( [ eq : superpot ] ) , we have a scalar potential including soft supersymmetry breaking terms , @xmath104 then the flaton masses are given by @xmath105 where @xmath106 and @xmath107 with @xmath108 then , the flatino masses are represented by @xmath109 allowed parameter regions for @xmath73 and @xmath110 of our interest ( i.e , parameters where the flatons can not decay into flatinos , @xmath111 ) are given in ( 28 ) in ref . @xcite . we present our results on the dm abundance and the stau lifetime by taking some typical parameter sets of the flat dfsz model . following the notations and the parameters presented in the previous sections and ref . @xcite , we take a set of @xmath112 , @xmath113 , @xmath114 , @xmath115 , @xmath116 , @xmath117 , and @xmath118 . in this paper we show two cases of the mass parameters 1 ) @xmath19 = 287 gev ( @xmath119 ) and @xmath120 gev , and 2 ) @xmath19 = 813 gev ( @xmath121 ) and @xmath122 gev . we can calculate the mass spectrum in gev unit of the flatons @xmath28 , the pseudoscalar flaton @xmath29 , and the flatinos @xmath30 which is given as @xmath123 = \left\ { \begin{array}{l } \left [ 216 , 175 , 162 , 184 , 112 \right ] \quad \left(\tilde{\mu}= 12 \right ) , \\ \left [ 616 , 500 , 462 , 525 , 320 \right ] \quad \left(\tilde{\mu}= 34 \right ) . \end{array } \right.\end{aligned}\ ] ] in each case , by using eq.([eq : trh ] ) the reheat temperature after the thermal inflation ( from the decay of the lightest flaton @xmath95 ) is given by @xmath124 ( @xmath125 gev ) for @xmath126 ( @xmath127 ) . in addition , we can diagonalize the 6@xmath576 mass matrix by using the approximated 5@xmath575 method discussed in sec . [ sec : lifetime ] and get the mass spectrum of four neutralinos plus two flatinos . then we can calculate the mass of the lightest neutralino @xmath128 ( or @xmath129 ) and get 1 ) @xmath130 120 gev and 2 ) 525 gev , respectively . in figs . [ fig : omega_mu12 ] and [ fig : omega_mu34 ] we plot the observationally - allowed region which is enclosed by two solid lines in ( @xmath131,@xmath93 ) plane for @xmath132=12 and 34 , respectively . here we have @xmath133 gev . in both figures , the vertical area enclosed by solid lines comes from the non - thermal contribution and the narrow horizontal area from the thermal contribution which is very sensitive function of @xmath134 . the vertical band between two dot - dashed lines denotes a region where the lightest flatino ( flaxino ) lsp and the stau nlsp are realized . in fig . [ fig : omega_mu34 ] , we see that the flaxino dark matter naturally agrees with observations in broad parameter spaces of @xmath93 at the weak scale . in figs . [ fig : lifetime ] and [ fig : lifetime2 ] we plot the lifetime of stau as a function of stau mass in case of @xmath135 , and 34 , respectively . the thick ( thin ) solid line denotes the case of the mixing angle @xmath136 . here we have taken @xmath133 gev to minimize the decay length . in case of @xmath135 , the decay length is shown to be around 1 m and thus the stau nlsp decay to the flaxino lsp will be possibly searched for at lhc . note that the lifetime scales @xmath137 . from these figures we see that the lifetime is sufficiently shorter than @xmath138 sec excluding the special case of degenerate masses @xmath139 ( @xmath140 gev ) . hence this scenario is not constrained by the @xmath13li overproduction through bbn if we adopt the standard range of @xmath133 gev @xmath141 gev . in other words , even lager value @xmath142 gev can be allowed from the viewpoint of bbn . such a high value might not lead to the overclosure of the universe by the axions because of a sufficiently low reheating temperature and large entropy productions after the thermal inflation @xcite . see also ref.@xcite for recent detailed analyses of the thermally - produced axions for a relatively low reheating temperature . = 12 , @xmath73=4 , @xmath110=13 , @xmath143=3 , and @xmath133 gev . the region between the two solid lines satisfies the wmap constraints on @xmath144 given in eq . ( [ eq : omegawmap3 ] ) . the lightest flatino lsp and the stau nlsp can be realized in the region between the two dot - dashed lines . the intersection of these regions is allowed . , width=377 ] but for @xmath132=34 . , width=377 ] ( @xmath145 ) . here we have adopted @xmath133gev and the electroweak scale @xmath69 = 264 gev . note that @xmath146 = 112 gev , and the mass of the lightest neutralino @xmath129 ( @xmath147 gev ) is smaller than that of @xmath72 ( @xmath77 = 184 gev ) in this case.,width=377 ] but for @xmath132=34 ( @xmath148 gev ) and @xmath122 gev . then , @xmath149 320 gev , and @xmath150 525 gev.,width=377 ] the lightest supersymmetric particle ( lsp ) is a dark matter candidate . it is usually said that the lsp will be either the lightest superpartner of a standard model particle ( so called losp ) , the gravitino or the axino . for cosmological constraints and collider signals of the lsp , it matters what is the nlsp . in this paper we revisited the axino - like lsp arising in the flat dfsz axion models for the case of the stau nlsp . following @xcite , we assumed that the spontaneous breaking of peccei - quinn symmetry comes only from supersymmetry breaking . then all of the pq particles except the axion will have masses of order the gravitino mass . the scalar particles are called flatons and their superpartners are called flatinos . in this scenario the saxion is a linear combination of flatons and the axino a linear combination of flatinos , but they are not mass eigenstates and have no special status . so the lsp candidate from the pq sector is not the axino , but the lightest flatino which we have dubbed the flaxino . we have explored the simplest version of this scenario , which has two flatinos and assumes gravity mediated supersymmetry breaking . we have found that the stau nlsp decay to the flaxino lsp is fast enough to maintain the standard predictions of big bang nucleosynthesis . the big difference from the gravitino dark matter scenario is that the flaxino / axino coupling ( @xmath151 ) is much larger than the gravitino coupling ( @xmath152 ) , which makes the stau nlsp decay much faster . in the dfsz axion model , the stau - flaxino / axino coupling appears at tree level as a consequence of the @xmath19 term mixing between the higgsino and flaxino / axino . this makes the nlsp decay more efficient compared with the ksvz axion models . in a specific dfsz model whose parameters are constrained by various cosmological considerations , we identified the region of the stop and stau masses which is compatible with the required dark matter density . within the conventional bound on the axion scale @xmath153 gev , the stau decay to the flaxino lsp never becomes dangerous for bbn . furthermore , with the chosen parameter sets in the first case ( @xmath135 ) , the stau decay length turns out to be @xmath154(1 ) m so that a fair number of stau decays can be captured in colliders when the axion scale is close to its lower bound , @xmath155 gev . in the second case ( @xmath121 ) , the flaxino lsp is naturally realized in the broad parameter space of the squark mass . although the decay length is longer than @xmath154(10 ) m , the additional detectors proposed by @xcite will catch the staus even in this case . let us finally remark that , in our scheme , the unstable gravitino decaying to ordinary superparticle or flaxino does not contradict with bbn as the primordial gravitino abundance will be diluted away by the entropy production after thermal inflation caused by the flat direction in the model . a gravitino even lighter than the flaxino is also allowed since the flaxino decay to the gravitino and the axion causes no trouble with bbn . this work was supported by the science research center program of the korea science and engineering foundation through the center for quantum spacetime(cquest ) of sogang university with grant number r112005021 ( h.b.k . ) . the research at lancaster is supported by pparc grant pp / d000394/1 and by eu grants mrtn - ct-2004 - 503369 and mrtn - ct-2006 - 035863 , the european union through the marie curie research and training network `` universenet '' . s. weinberg , phys . lett . * 48 * , 1303 ( 1982 ) ; j. silk and a. stebbins , astrophys . j. * 269 * ( 1983 ) 1 ; j. r. ellis , j. e. kim and d. v. nanopoulos , phys . b * 145 * , 181 ( 1984 ) ; d. lindley , astrophys . j. * 294 * , 1 ( 1985 ) ; m. y. khlopov and a. d. linde , phys . lett . b * 138 * ( 1984 ) 265 ; j. r. ellis , d. v. nanopoulos and s. sarkar , nucl . b * 259 * , 175 ( 1985 ) ; r. juszkiewicz , j. silk and a. stebbins , phys . b * 158 * ( 1985 ) 463 ; j. r. ellis , g. b. gelmini , j. l. lopez , d. v. nanopoulos and s. sarkar , nucl . b * 373 * ( 1992 ) 399 . m. kawasaki and t. moroi , prog . theor . * 93 * , 879 ( 1995 ) ; e. holtmann , m. kawasaki , k. kohri and t. moroi , phys . rev . d * 60 * , 023506 ( 1999 ) ; k. jedamzik , phys . lett . * 84 * , 3248 ( 2000 ) ; m. kawasaki , k. kohri and t. moroi , phys . d * 63 * , 103502 ( 2001 ) ; r. h. cyburt , j. r. ellis , b. d. fields and k. a. olive , phys . d * 67 * , 103521 ( 2003 ) ; j. r. ellis , k. a. olive and e. vangioni , phys . b * 619 * , 30 ( 2005 ) . m. kawasaki , k. kohri and t. moroi , phys . b * 625 * , 7 ( 2005 ) ; phys . d * 71 * , 083502 ( 2005 ) ; k. kohri , t. moroi and a. yotsuyanagi , phys . d * 73 * , 123511 ( 2006 ) ; k. jedamzik , phys . d * 74 * , 103509 ( 2006 ) . k. kohri and f. takayama , phys . d * 76 * , 063507 ( 2007 ) . m. kaplinghat and a. rajaraman , phys . d * 74 * , 103004 ( 2006 ) . r. h. cyburt , j. r. ellis , b. d. fields , k. a. olive and v. c. spanos , jcap * 0611 * , 014 ( 2006 ) . r. n. cahn and s. l. glashow , science * 213 * , 607 ( 1981 ) ; r.n . boyd , r.j . perry , t.a . miler and k. takahashi , science * 244 * , 1450 ( 1989 ) ; a. de rujula , s. l. glashow and u. sarid , nucl . b * 333 * , 173 ( 1990 ) ; s. dimopoulos , d. eichler , r. esmailzadeh and g. d. starkman , phys . rev . d * 41 * , 2388 ( 1990 ) . j. pradler and f. d. steffen , phys . b * 648 * , 224 ( 2007 ) ; t. jittoh et al . d * 76 * , 125023 ( 2007 ) ; s. kasuya and f. takahashi , jcap * 0711 * , 019 ( 2007 ) ; f. takayama , arxiv:0704.2785 [ hep - ph ] ; j. pradler and f. d. steffen , arxiv:0710.2213 [ hep - ph ] ; n. okada and o. seto , arxiv:0710.0449 [ hep - ph ] ; j. kersten and k. schmidt - hoberg , arxiv:0710.4528 [ hep - ph ] . h. murayama , h. suzuki and t. yanagida , phys . b * 291 * , 418 ( 1992 ) . k. choi , e. j. chun and j. e. kim , phys . b * 403 * , 209 ( 1997 ) . e. j. chun , d. comelli and d. h. lyth , phys . rev . d * 62 * , 095013 ( 2000 ) . e. j. chun , h. b. kim and d. h. lyth , phys . d * 62 * , 125001 ( 2000 ) . s. p. martin , phys . d * 62 * , 095008 ( 2000 ) . l. covi , j. e. kim and l. roszkowski , phys . lett . * 82 * , 4180 ( 1999 ) ; l. covi , h. b. kim , j. e. kim and l. roszkowski , jhep * 0105 * , 033 ( 2001 ) . a. brandenburg and f. d. steffen , jcap * 0408 * , 008 ( 2004 ) . l. covi , l. roszkowski , r. ruiz de austri and m. small , jhep * 0406 * , 003 ( 2004 ) ; a. brandenburg , l. covi , k. hamaguchi , l. roszkowski and f. d. steffen , phys . b * 617 * , 99 ( 2005 ) . k. hamaguchi , y. kuno , t. nakaya and m. m. nojiri , phys . d * 70 * , 115007 ( 2004 ) . j. l. feng and b. t. smith , phys . d * 71 * , 015004 ( 2005 ) [ erratum - ibid . d * 71 * , 0109904 ( 2005 ) ] . a. de roeck , j. r. ellis , f. gianotti , f. moortgat , k. a. olive and l. pape , eur . j. c * 49 * , 1041 ( 2007 ) .	if the spontaneous breaking of peccei - quinn symmetry comes from soft supersymmetry breaking , the fermionic partners of the symmetry - breaking fields have mass of order the gravitino mass , and are called flatinos . the lightest flatino , called here the flaxino , is a cdm candidate if it is the lightest supersymmetric particle . we here explore flaxino dark matter assuming that the lightest ordinary supersymmetric particle is the stau , with gravity - mediated supersymmetry breaking . the decay of the stau to the flaxino is fast enough not to spoil the standard predictions of big bang nucleosynthesis , and its track and decay can be seen in future colliders .
the alternating voltage produces the stationary current by pumping electrons between leads @xmath65 and @xmath66 . the effect is sensitive to the phase coherence of alternating signals and can exist even in symmetric systems . the stationary current is possible also in the case of different amplitudes of alternating fields . we have studied the system both analytically and numerically . the analytical approach is based on the perturbational ( with respect to amplitudes of a.c . signal ) consideration . the current contains independent contributions caused by @xmath3 and an interference term . the elastic , absorption and emission channels participate in the process . the case of strong alternating signal was studied numerically . the motivation to consider the symmetric case is that in this case the stationary current caused by a single frequency is suppressed . only two intercoherent frequencies can cause the stationary current . hence , the system with specially symmetrized barriers can be used for heterodyning of signals with @xmath1 and @xmath2 frequencies and for frequency binding of similar laser sources . the controllability of the system permits to realize the symmetric conditions intentionally with good accuracy . the work was supported by grants of the program for support of scientific schools of the russian federation no . 4500.2006.2 , the grant of the president of the russian federation no . mk-8112.2006.2 and the dynasty foundation . we are grateful to l. s. braginsky for useful discussions . more recent study of photovoltaic effect in artificial 2d lattices of antidots , see for example : a. d. chepelianskii , and d. l. shepelyansky , phys . b * 71 * , 052508 ( 2005 ) ; g. cristadoro , and d. l. shepelyansky , phys . rev . e * 71 * , 036111 ( 2005 ) ; a. d. chepelianskii , cond - mat/0509679 ; m.v . entin , l.i . magarill , phys . b * 73 * , 205206 ( 2006 ) .	a one - dimensional system with two @xmath0-like barriers or wells bi - chromaticaly oscillating at frequencies @xmath1 and @xmath2 is considered . the alternating signal leads to the direct current across the structure ( even in a symmetric system ) . the properties of this quantum pump are studied in a wide range of the system parameters . the quantum pump is a device that generates the stationary current under action of alternating voltage ; it is a subject of numerous recent publications ( for example , @xcite . the quantum pump is essentially analogous to various versions of the photovoltaic effect studied in details from the beginning of the eighties @xcite . the difference is that the photovoltaic effect is related to the emergence of a direct current in a homogeneous macroscopic medium ( the only exception is the mesoscopic photovoltaic effect ) , while the pump is a microscopic object . from the phenomenological point of view , the emergence of a direct current in the pump is not surprising , since any asymmetric microcontact can rectify ac voltage . however , analysis of adiabatic transport in the quantum - mechanical object leads to new phenomena , such as quantization of charge transport @xcite . just this , analytically solvable , adiabatic approach was utilized in the most of studies of quantum pumps @xcite . in the recent papers we have carried out the extensive study of the simplest model of the one - dimensional quantum pump , containing two delta - like harmonically oscillating barriers / wells . this model demonstrates rich behavior which is ruled by a variety of system parameters . the present paper deals with similar system to which alternating bi - chromatic voltages are applied . the system can be exemplified by a quantum wire with two narrow gates . the stationary bias between the source and the drain is supposedly absent . the system has a variety of regimes of the pump operation , depending on the system parameters , e.g. frequency and amplitudes . the effect is sensitive to the phase coherence of alternating signals and can exist even in symmetric systems . the stationary current is possible also in the case of different amplitudes of alternating fields . we have studied the system both analytically and numerically . the analytical approach is based on the perturbational ( with respect to amplitudes ( @xmath3 ) of a.c . signal ) consideration . the current contains independent contributions caused by @xmath3 and an interference term . the elastic , absorption and emission channels participate in the process . the case of strong alternating signal was studied numerically . we mostly concentrate on the case of symmetric system as more interesting by its phase sensitivity . the considered model is described by the one - dimensional time - dependent potential : @xmath4 where @xmath5 is the time , @xmath6 is the distance between @xmath0-barriers ( wells ) ; quantities @xmath3 are measured in units of @xmath7 ( @xmath8 is the electron mass ) ; @xmath9 , @xmath10 , and @xmath1 are the momentum , energy , and frequency measured in units of @xmath11 , @xmath12 , and @xmath13 , respectively . the solution to the schrdinger equation with the potential ( [ 1 ] ) is searched in the form @xmath14\left\ { \begin{array}{rl}\delta_{n,0}\exp\left(\frac{ip_nx}{d}\right)+r_n\exp\left(-\frac{ip_nx}{d}\right ) , & x<-d,\\a_n\exp\left(\frac{ip_nx}{d}\right)+b_n\exp\left(-\frac{ip_nx}{d}\right ) , & -d < x < d , \\ t_n\exp\left(\frac{ip_nx}{d}\right ) , & x > d . \end{array}\right.\end{aligned}\ ] ] here , @xmath15 and @xmath16 . the wave function ( [ wf ] ) corresponds to the wave incident on the barrier from the left . ( in the final formulas , we mark the directions of incident waves by the indices @xmath17 and @xmath18 ) . the form of solution ( [ wf ] ) corresponds to absorption ( for @xmath19 ) or emission ( @xmath20 ) of @xmath21 field quanta by an electron after interaction with the vibrating barriers ; @xmath22 relates to the elastic process . quantities @xmath23 and @xmath24 give the corresponding amplitudes of transmission ( reflection ) . if the value of @xmath25 becomes imaginary , the waves moving away from the barriers should be treated as damped waves , so that @xmath26 . the transmission amplitudes obey the equations : @xmath27 , @xmath28 and @xmath29 here , @xmath30 . provided that electrons from the right and left of the pump are in equilibrium , and that they have identical chemical potentials @xmath31 , the stationary current is @xmath32 where @xmath33 is the fermi distribution function , and @xmath34 is the heaviside step function . the current is determined by the transmission coefficients with real @xmath25 only . at a low temperature , it is convenient to differentiate the current with respect to the chemical potential : @xmath35 here @xmath36 is the conductance quantum and @xmath37 is the fermi momentum . the resultant quantity @xmath38 can be treated as a two - terminal photoconductance ( the conductance for simultaneous change of chemical potentials of source and drain ) . in low - amplitude limit the stationary current ( its derivative @xmath38 ) is proportional to @xmath39 let s consider the case of symmetric system with @xmath40 . the systems symmetry leads to the dependence @xmath41 . this contribution arises from corresponding terms in the transmission coefficients @xmath42 : @xmath43 the quantities @xmath44 , @xmath45 , @xmath46 , @xmath47 , @xmath48 depend on @xmath49 , @xmath50 which contains one- and two - photon singularities . the calculations support this dependence . the figures [ fig1]-[fig7 ] show typical plots of the derivative of the stationary current with respect to the fermi energy @xmath51 as a function of the fermi momentum ( figs . [ fig1]-[fig5 ] ) , the amplitude ( fig . [ fig6 ] ) and the frequency ( fig . [ fig7 ] ) . the figures [ fig1]-[fig5 ] present the quantity @xmath52 as a function of the fermi momentum for @xmath53 and @xmath54 ( fig . [ fig1 ] ) , 2 ( fig . [ fig2 ] ) . the peaks of the curves correspond to the multi - photon threshold resonances with zero energy state . the increasing of @xmath55 leads to the strengthening of multi - photon singularities . the curve in the fig . [ fig1 ] corresponds to antipodal signals with same amplitudes @xmath56 , that demonstrate the importance of the phase coherency of the signals ( in such conditions the stationary current in the same system under monochromatic voltage is vanishing @xcite ) . the current oscillates with the fermi momentum due to the interference of the electron waves in the structure . besides , the current possesses singularities caused by the resonances with the zero energy state and their photon repetitions . the figure [ fig5 ] shows the behavior of @xmath52 for large enough frequency where the only threshold singularity exists in the concerned range of fermi momentum values . the figure [ fig6 ] depicts @xmath52 _ versus _ @xmath57 for the symmetric system ( @xmath58 ) . these values are chosen so that there should be no current caused by harmonic signals only ( if @xmath59 or @xmath60 ) . besides , the current vanishes in the adiabatic approximation which is commonly used for consideration of quantum pumps . in the adiabatic regime the charge transfer per a cycle is proportional to the area covered by two parameters in the phase space . in our case the parameters are @xmath61 . if @xmath58 the trajectory in the phase space @xmath61 is a straight line and covers zero area ( see insert ) . hence in the adiabatic ( @xmath62 ) approximation @xmath63 . at low @xmath3 @xmath64 . this behavior corresponds to coherent photovoltaic effect @xcite . at large @xmath3 the amplitude dependence exhibits distinctive oscillations which are determined by the commensurability of the characteristic wavelength of excited electrons with the distance between delta - functions . the curves for different frequencies demonstrate splitting and beating of oscillations . their amplitude decays with @xmath3 . the oscillations period changes with the frequency . the figure [ fig7 ] demonstrates the dependence of @xmath52 on the frequency . the complicated structure of @xmath52 in the low frequency region is explained by multi - photon resonances reducing for larger frequencies
a variety of metal surfaces , such as the ( 111 ) surfaces of the noble metals , are known to support a partially occupied band of shockley surface states with energies near the fermi level,@xcite whose dynamics have been the subject of long - standing interest . @xcite in particular , the lifetimes of excited holes at the band edge ( @xmath1 ) of these surface states have been investigated with high resolution angle - resolved photoemission ( arp)@xcite and with the use of the scanning tunneling microscope ( stm).@xcite stm techniques have also allowed the determination of the lifetimes of excited shockley and image electrons over a range of energies above the fermi level . @xcite many - body calculations of the electron - electron ( e - e ) inelastic lifetimes of excited holes at the surface - state band edge of the ( 111 ) surfaces of the noble metals cu , ag , and au , which were based upon the @xmath6 one - dimensional scheme that had been introduced by chulkov _ et al._@xcite to describe the lifetimes of image - potential states,@xcite showed considerable agreement with experiment.@xcite these calculations were then extended to treat the case of excited surface - state and surface - resonance electrons above the fermi level.@xcite in order to account approximately for the potential variation in the plane of the surface , the original one - dimensional model potential , which had been introduced to describe surface states at the @xmath0 point , was modified along with the introduction of a realistic effective mass for the dispersion curve of both bulk and surface states . within this model , however , all shockley states have the same effective mass , so the projected band structure is not correct , especially at energies above the fermi level , as shown in fig . [ fig1 ] . in this paper , we present a new approach that although at the @xmath0 point is less sophisticated than the model used in refs . ( i ) basically reproduces the surface - state probability density of refs . at the band edge of the surface - state band in cu(111 ) and ( ii ) has the merit that it reproduces , through the introduction of a @xmath7-dependent one - dimensional potential , the actual bulk energy bands and surface - state energy dispersion of fig . 1 , thereby allowing for a realistic description of the electronic orbitals beyond the @xmath0 point . adding the contribution from electron - phonon coupling,@xcite which is particularly important at the smallest excitation energies , our calculations of the lifetime broadening of excited shockley electrons and holes in cu(111 ) indicate that ( i ) there is good agreement with experiment at the surface - state band edge and ( ii ) at energies above the fermi level the lifetime broadening is closer to experiment and very sensitive to the actual shape of the surface - state single - particle orbitals beyond the @xmath0 point . let us consider a semi - infinite many - electron system that is translationally invariant in the plane of the surface ( normal to the @xmath8 axis ) . the decay rate or inverse lifetime of a quasiparticle ( electron or hole ) that has been added to the system in the single - particle state @xmath9 of energy @xmath10 is obtained as follows ( we use atomic units , i.e. , @xmath11)@xcite @xmath12 where the @xmath13 sign in front of the integral should be taken to be minus or plus depending on whether the quasiparticle is an electron ( @xmath14 ) or a hole ( @xmath15 ) , respectively , @xmath16 is the fermi energy , @xmath17 and @xmath18 represent the position and wave vectors in the plane of the surface , and @xmath19 is the nonlocal self - energy operator . to lowest order in a series - expansion of the self - energy @xmath19 in terms of the energy - dependent screened interaction @xmath20,@xcite and replacing the interacting green function @xmath21 by its noninteracting counterpart , one finds the following expression for the imaginary part of the so - called @xmath6 self - energy : @xmath22 here , @xmath23 , @xmath24 , and @xmath25 is the noninteracting green function @xmath26(z)},\ ] ] where @xmath27 ( @xmath28 ) is the lesser ( greater ) of @xmath8 and @xmath29 , and @xmath30(z)$ ] is the wronskian @xmath31(z)=f(z)g'(z)-f'(z)g(z).\ ] ] the functions @xmath32 are solutions of the single - particle schrdinger equation @xmath33 regular at @xmath34 , with @xmath35 being a momentum - dependent one - dimensional effective potential that we fit to the projected surface band structure . we use @xmath36 where @xmath37 and @xmath38 are fitted to the bulk energy bands ( which we have obtained from three - dimensional _ ab initio _ calculations ) , @xmath39 represents the interlayer spacing , @xmath40 is the experimentally determined work function , and the matching plane @xmath41 is chosen to give the correct surface - state dispersion represented in fig . 1 by a thick solid line.@xcite in the random - phase approximation ( rpa),@xcite the screened interaction @xmath42 is obtained from the knowledge of the noninteracting density - response function @xmath43 by solving the following integral equation @xmath44 where @xmath45 represents the two - dimensional fourier transform of the bare coulomb interaction . the results presented below have been obtained by using in @xmath43 the eigenfunctions and eigenvalues of the one - dimensional hamiltonian of ref . with all effective masses set equal to the free - electron mass . we have also used the eigenfunctions and eigenvalues of a single - particle jellium - surface kohn - sham hamiltonian ( in the local - density approximation ) with @xmath46,@xcite and we have found that the surface - state lifetimes are rather insensitive to whether one or the other choice is employed . of the shockley surface state at the center of the surface brillouin zone of cu(111 ) , as obtained within two different models : the model presented here ( dashed line ) and the model reported in ref . ( solid line ) . the dotted line represents the probability - density that we obtain at @xmath47 . full circles represent the atomic positions of cu in the ( 111 ) direction . the geometrical ( jellium ) electronic edge ( @xmath48 ) has been chosen to be located half an interlayer spacing beyond the last atomic layer . ] the abrupt step model potential of eq . ( [ model ] ) , which does not account for the image tail outside the surface , could not possibly be used to describe image states . however , shockley surface states are known to be rather insensitive to the actual shape of the potential outside the surface ; indeed , the model potential of eq . ( [ model ] ) is found to yield a surface - state probability - density @xmath49 at the band edge of the shockley surface - state band of cu(111 ) ( @xmath50 ) that is in reasonably good agreement with the more realistic surface - state probability density used in refs . , as shown in fig . 2 . both probability - densities coincide within the bulk , although our approximate probability - density appears to be slightly more localized near the surface , as expected.@xcite nevertheless , we find that decay rates of an excited hole at @xmath0 based on the use of these two models to describe the wave function @xmath51 entering eq . ( [ eq1 ] ) agree within less than @xmath52 . differences between our new calculations , which are based on the use of the @xmath53-dependent model potential of eq . ( [ model ] ) , and those reported previously,@xcite are due to our more realistic description of the band structure and surface - state wave functions beyond the @xmath0 point . .@xmath6 decay rates , in linewidth units ( mev ) , of an excited hole at the band edge of the shockley surface - state band of cu(111 ) ( @xmath54 and @xmath50 ) . @xmath55 and @xmath56 represent interband and intraband contributions to the e - e decay rate @xmath57 . the total decay rate @xmath58 includes the e - ph decay rate of 7 mev reported in ref . . the full calculation represents the result we have obtained by using in eq . ( [ eq3 ] ) the actual @xmath53-dependent model single particle wave function @xmath59 . the approximate calculation represents the result we have obtained by replacing all surface - state wave functions with @xmath60 by that at the @xmath0 point . the third row represents the calculations reported in refs . . the experimental linewidth has been taken from the stm measurements reported in ref . . [ cols="<,^,^,^,^,^",options="header " , ] first of all , we consider an excited hole at the band edge of the shockley surface - state band of cu(111 ) , i.e. , with @xmath61 and @xmath50 ( see fig . 1 ) . the decay of this excited quasiparticle may proceed either through the coupling with bulk states ( interband contribution ) or through the coupling , within the surface - state band itself , with surface states of different wave vector @xmath62 parallel to the surface ( intraband contribution ) . in order to investigate the impact of the actual shape of the surface - state wave functions with @xmath60 on the decay of the surface - state hole at @xmath0 , we have compared in table i the decay rates that we have calculated either by using in eq . ( [ eq3 ] ) the actual @xmath53-dependent model wave function @xmath59 ( full calculation ) or by replacing all surface - state wave functions @xmath59 with @xmath60 by that at the @xmath0 point ( approximate calculation ) . this comparison shows that the penetration of the actual @xmath53-dependent surface - state wave functions @xmath59 being larger than at @xmath0 ( compare the dashed and dotted lines of fig . [ fig1 ] ) yields a reduction in the decay rate from @xmath63 to @xmath64 , which is due to the fact that the coupling of the surface - state hole at @xmath0 with actual surface states of different wave vector @xmath62 ( intraband contribution ) is smaller than the coupling that would take place with surface - state orbitals that do not change with @xmath62 . the difference between our predicted surface - state lifetime broadening of @xmath64 and that reported before ( @xmath65)@xcite is entirely due to our more accurate description of ( i ) the projected band structure and ( ii ) the wave - vector dependence of the surface - state wave functions @xmath66 entering the evaluation of the green function of eq . ( [ eq3 ] ) . we have also carried out a full calculation of the decay of an excited hole at @xmath0 but replacing the actual surface - state wave vector @xmath67 entering eq . ( [ eq2 ] ) by the wave vector that would correspond to a parabolic surface - state dispersion of the form dictated by the thin solid line of fig . [ fig1 ] , and we have found that the linewidth is _ reduced _ ( as expected , since the parabolic dispersion results in fewer final states ) by less than @xmath68 . however , if one further replaces our wave - vector dependent surface - state orbitals entering eq . ( [ eq2 ] ) by their less accurate counterparts used previously,@xcite the lifetime broadening is increased considerably ( from @xmath69 to @xmath70 ) , showing the important role that the actual coupling between initial and final states plays in the surface - state decay mechanism . our model , which correctly reproduces the behaviour of @xmath71-@xmath72 valence states , does not account for the presence of @xmath73 electrons with energies a few electronvolts below the fermi level . the screening of @xmath73 electrons is known to play a crucial role in the decay mechanism of bulk states.@xcite however , in the case of shockley holes , whose decay is dominated by intraband transitions that are associated with very small values of the momentum transfer , the screening of @xmath73 electrons is expected to reduce the lifetime broadening only very slightly,@xcite and will not be included in the present work . adding to our estimated e - e linewidth of @xmath64 the electron - phonon ( e - ph ) linewidth of @xmath74 reported in ref . , we find @xmath75 in close agreement with the experimentally measured linewidth of @xmath76 , as shown in table i. of shockley surface - state electrons and holes in cu(111 ) ( including the e - ph contribution @xmath77 ) versus the surface - state energy @xmath10 , as obtained from eqs . ( [ eq1])-([eqlast ] ) ( thick solid line ) , from eqs . ( [ eq1])-([eqlast ] ) but replacing all surface - state wave functions with @xmath60 by those at the @xmath0 point ( thin solid line ) , and from ref . ( thin dashed line ) . the dotted line represents the energy - independent e - ph contribution @xmath77 . the dotted line with squares represents the stm measurements reported in ref . , multiplied by a factor of two to correct for an error in the phase coherence length used in that work . the solid circle represents the experimentally determined linewidth of an excited hole at the @xmath0 point , as reported in ref . . separate unscaled interband ( @xmath55 ) and intraband ( @xmath56 ) contributions to the total linewidth ( @xmath78 ) are represented in the inset by dashed and dottel lines , respectively . ] in fig . 3 , we show our full calculation ( thick solid line ) of the inelastic linewidth ( @xmath79 ) of excited shockley holes and electrons in cu(111 ) with energies @xmath10 below and above the fermi level . also shown are separate interband and intraband contributions to the linewidth ( inset ) , the approximate calculation that we have carried out by replacing all surface - state wave functions with @xmath60 by the surface - state wave function at @xmath0 ( thin solid line ) , and the calculations reported in ref . ( thin dashed line ) . the lifetime broadening of excited shockley electrons in cu(111 ) was studied with the stm by brgi _ et al._. @xcite however , it has been argued recently @xcite that due to an error in the identification of the phase coherence length in the measured quantum interference patterns the linewidths reported in ref . should be doubled . these doubled values have been represented in fig . 3 by solid squares . the experimentally determined inelastic linewidth of an excited shockley hole at the @xmath0 point@xcite is represented by a solid circle . 3 shows that a correct description of the wave - vector dependent surface - state wave functions reduces the coupling of holes and electrons within the shockley band , thereby bringing the lifetime broadening into closer agreement with experiment . a comparison between our full calculations and experiment shows that there is close agreement at the surface - state band edge ( at @xmath80 ) and there is also reasonable agreement at energies above the fermi level . at energies where the surface - state band merges into the continuum of bulk states , however , our calculated linewidths are still too low , which might be a signature of the need of a fully three - dimensional description of the surface band structure . we also note that differences between the calculations reported here and those reported previously@xcite indicate that inelastic lifetimes are very sensitive to the actual shape of the surface - state single - particle orbitals beyond the @xmath0 point . the linewidths reported here are smaller ( larger ) for excited holes ( electrons ) below ( above ) the fermi level , thus bringing the theretical predictions closer to experiment . in summary , we have presented a new @xmath6 one - dimensional scheme to calculate the inelastic lifetime broadening of excited shockley electrons and holes in cu(111 ) , which is based on a realistic description of the projected bulk energy bands and the surface - state orbitals beyond the @xmath0 point . adding the contribution from electron - phonon coupling,@xcite which is particularly important at the smallest excitation energies , our calculations indicate that there is reasonable agreement with experiment , especially at low excitation energies . the screening of @xmath73 electrons , not included in this work , is expected to reduce the lifetime broadening only very slightly , at least at the hot - electron energies nearest to the fermi level . the authors thank e. v. chulkov , p. m. echenique , and j. e. inglesfield for enjoyable discussions and a careful reading of the manuscript . m.g.v . and j.m.p . acknowledge partial support by the university of the basque country , the basque unibertsitate eta ikerketa saila , the mcyt , and the ec 6th framework network of excellence nanoquanta ( nmp4-ct-2004 - 500198 ) . j. e. inglesfield , rep . . phys . * 45 * , 223 ( 1982 ) . r. matzdorf , surf . sci . rep . * 30 * , 153 ( 1998 ) . m. echenique _ et al . _ , phys . * 251 * , 1 ( 2000 ) . m. echenique _ et al . _ , rep . * 52 * , 219 ( 2004 ) . s. d. kevan , phys . lett . * 50 * , 526 ( 1983 ) . s. d. kevan and r. h. gaylord , phys . b * 36 * , 5809 ( 1987 ) . b. a. mcdougall _ et al . b , * 51 * , 13891 ( 1995 ) . w. schattke and m. a. van hove , solid - state photoemission and related methods , wiley / vch , weinheim ( 2003 ) . et al . _ , lett . * 81 * , 4464 ( 1998 ) . et al . _ , lett . * 82 * , 4516 ( 1999 ) . p. wahl _ et al . _ , phys . lett . * 91 * , 106802 ( 2003 ) . e. v. chulkov _ et al . _ , sci . * 391 * , l1217 ( 1997 ) . e. v. chulkov _ at al . lett . * 80 * , 4947 ( 1998 ) . j. kliewer _ et al . _ , science * 288 * , 1399 ( 2000 ) . m. echenique _ et al . _ , a * 71 * , 503 ( 2000 ) . p. m. echenique _ et al . _ , sci . * 67 * , 271 ( 2001 ) . l. vitali _ et al . _ , sci . * 523 * , l47 ( 2003 ) . a. eiguren _ et al . _ , 88 * , 066805 ( 2002 ) . m. nekovee and j. m. pitarke , comput . . commun . * 137 * , 123 ( 2001 ) . l. hedin and s. lundqvist , solid state phys . * 23 * , 1 ( 1969 ) . f. aryasetiawan and o. gunnarsson , rep . . phys . * 61 * , 237 ( 1998 ) . at the @xmath0 point ( @xmath81 ) , one finds @xmath82 ( with respect to the fermi level ) , @xmath83 , and @xmath84 ( with respect to the jellium edge ) . at @xmath85 , we find @xmath86 , @xmath87 , and @xmath88 . a. l. fetter and j. d. wallecka , _ quantum theory of many - particle systems _ ( mcgraw - hill , new york , 1964 ) . the valence electron density @xmath89 of the solid is usually characterized by the density parameter @xmath90 , @xmath91 being the bohr radius @xmath92 . for the overlap integral we have found @xmath93 , @xmath94 and @xmath95 being the shockley probability - amplitudes leading to the probability - densities represented in fig . [ fig2 ] by solid and dashed lines . i. campillo _ et al . _ , lett . * 83 * , 2230 ( 1999 ) . a. garca - lekue _ et al . b * 68 * , 045103 ( 2003 ) . s. crampin _ et al . _ , 95 * , 029701 ( 2005 ) .	a theoretical many - body analysis is presented of the electron - electron inelastic lifetimes of shockley electrons and holes at the ( 111 ) surface of cu . for a description of the decay of shockley states both below and above the fermi level , single - particle wave functions have been obtained by solving the schrdinger equation with the use of an approximate one - dimensional pseudopotential fitted to reproduce the correct bulk energy bands and surface - state dispersion . a comparison with previous calculations and experiment indicates that inelastic lifetimes are very sensitive to the actual shape of the surface - state single - particle orbitals beyond the @xmath0 ( @xmath1 ) point , which controls the coupling between the shockley electrons and holes .
chrestenson @xcite has shown that any ( planar ) regular polygon whose vertices are contained in @xmath0 for some @xmath1 must have @xmath2 or @xmath3 vertices . more generally , gardner and gritzmann @xcite have characterized the numbers of vertices of affinely regular _ lattice polygons _ , i.e. , images of non - degenerate regular polygons under a non - singular affine transformation of the plane whose vertices are contained in the square lattice @xmath4 or , equivalently , in some arbitrary planar lattice @xmath5 . it turned out that the affinely regular lattice polygons are precisely the affinely regular triangles , parallelograms and hexagons . as a first step beyond the case of planar lattices , this short text provides a generalization of this result to planar sets @xmath6 that are non - degenerate in some sense and satisfy a certain affinity condition on finite scales ( theorem [ charac ] ) . the obtained characterization can be expressed in terms of a simple inclusion of real field extensions of @xmath7 and particularly applies to _ algebraic delone sets _ , thus including _ cyclotomic model sets_. these cyclotomic model sets range from periodic examples , given by the vertex sets of the square tiling and the triangular tiling , to aperiodic examples like the vertices of the ammann - beenker tiling , of the tbingen triangle tiling and of the shield tiling , respectively . i turns out that , for cyclotomic model sets @xmath6 , the numbers of vertices of affinely regular polygons in @xmath6 can be characterized by a simple divisibility condition ( corollary [ th1mod ] ) . in particular , the result on affinely regular lattice polygons is contained as a special case ( corollary [ cor2](a ) ) . additionally , it is shown that the obtained divisibility condition implies a weak estimate in the _ discrete tomography _ of cyclotomic model sets ( corollary [ corofin ] ) . natural numbers are always assumed to be positive , i.e. , @xmath8 and we denote by @xmath9 the set of rational primes . if @xmath10 , then @xmath11 and @xmath12 denote their greatest common divisor and least common multiple , respectively . the group of units of a given ring @xmath13 is denoted by @xmath14 . as usual , for a complex number @xmath15 , @xmath16 denotes the complex absolute value , i.e. , @xmath17 , where @xmath18 denotes the complex conjugation . the unit circle in @xmath19 is denoted by @xmath20 , i.e. , @xmath21 . moreover , the elements of @xmath20 are also called _ directions_. for @xmath22 and @xmath23 , @xmath24 denotes the open ball of radius @xmath25 about @xmath26 . a subset @xmath6 of the plane is called _ uniformly discrete _ if there is a radius @xmath22 such that every ball @xmath24 with @xmath27 contains at most one point of @xmath6 . further , @xmath6 is called _ relatively dense _ if there is a radius @xmath28 such that every ball @xmath29 with @xmath27 contains at least one point of @xmath6 . @xmath6 is called a _ delone set _ ( or _ delaunay set _ ) if it is both uniformly discrete and relatively dense . for a subset @xmath30 of the plane , we denote by @xmath31 , @xmath32 , @xmath33 and @xmath34 the cardinality , set of finite subsets , convex hull and characteristic function of @xmath30 , respectively . a direction @xmath35 is called an @xmath30_-direction _ if it is parallel to a non - zero element of the difference set @xmath36 of @xmath30 . further , a finite subset @xmath37 of @xmath30 is called a _ convex subset of _ @xmath30 if its convex hull contains no new points of @xmath30 , i.e. , if @xmath38 holds . moreover , the set of all convex subsets of @xmath30 is denoted by @xmath39 . recall that a _ linear transformation _ ( resp . , _ affine transformation _ ) @xmath40 of the euclidean plane is given by @xmath41 ( resp . , @xmath42 ) , where @xmath43 is a real @xmath44 matrix and @xmath45 . in both cases , @xmath46 is called _ singular _ when @xmath47 ; otherwise , it is non - singular . a _ homothety _ @xmath48 is given by @xmath49 , where @xmath50 is positive and @xmath45 . convex polygon _ is the convex hull of a finite set of points in @xmath51 . for a subset @xmath52 , a _ polygon in _ @xmath30 is a convex polygon with all vertices in @xmath30 . regular polygon _ is always assumed to be planar , non - degenerate and convex . an _ affinely regular polygon _ is a non - singular affine image of a regular polygon . in particular , it must have at least @xmath53 vertices . let @xmath54 be a finite set of directions . a non - degenerate convex polygon @xmath55 is called a @xmath56_-polygon _ if it has the property that whenever @xmath57 is a vertex of @xmath55 and @xmath58 , the line @xmath59 in the plane in direction @xmath60 which passes through @xmath57 also meets another vertex @xmath61 of @xmath55 . for a subset @xmath62 , we denote by @xmath63 the intermediate field of @xmath64 that is given by @xmath65 where @xmath66 denotes the difference set of @xmath6 . further , we set @xmath67 , the maximal real subfield of @xmath63 . note that @xmath56-polygons have an even number of vertices . moreover , an affinely regular polygon with an even number of vertices is a @xmath56-polygon if and only if each direction of @xmath56 is parallel to one of its edges . for @xmath68 , we always let @xmath69 , as a specific choice for a primitive @xmath70th root of unity in @xmath71 . let @xmath72 be the corresponding cyclotomic field . it is well known that @xmath73 is the maximal real subfield of @xmath72 ; see @xcite . throughout this text , we shall use the notation @xmath74},\ ; { \thinspace\scriptstyle{\mathcal{o}}\displaystyle}_{n } = { \mathbbm{z}}[{\zeta_{n}}+\bar{\zeta}_{n}]\,.\ ] ] except for the one - dimensional cases @xmath75 , @xmath76 is an imaginary extension of @xmath7 . further , @xmath77 will always denote euler s phi - function , i.e. , @xmath78 occasionally , we identify @xmath71 with @xmath79 . primes @xmath80 for which the number @xmath81 is prime as well are called sophie germain prime numbers . we denote by @xmath82 the set of sophie germain prime numbers . they are the primes @xmath83 such that the equation @xmath84 has solutions . it is not known whether there are infinitely many sophie germain primes . the first few are @xmath85 see entry a005384 of @xcite for further details . we need the following facts from the theory of cyclotomic fields . * theorem 2.5)[gau ] @xmath86 = \phi(n)$ ] . the field extension @xmath87 is a galois extension with abelian galois group @xmath88 , where @xmath89 corresponds to the automorphism given by @xmath90 . since @xmath91 is the maximal real subfield of the @xmath70th cyclotomic field @xmath76 , fact [ gau ] immediately gives the following result . [ cr5 ] if @xmath92 , one has @xmath93= 2 $ ] . thus , a @xmath94-basis of @xmath76 is given by @xmath95 . the field extension @xmath96 is a galois extension with abelian galois group @xmath97 of order @xmath98 = \phi(n)/2 $ ] . consider an algebraic number field @xmath99 , i.e. , a finite extension of @xmath7 . a full @xmath100-module @xmath101 in @xmath99 ( i.e. , a free @xmath100-module of rank @xmath102 $ ] ) which contains the number @xmath103 and is a ring is called an _ order _ of @xmath99 . note that every @xmath104-basis of @xmath101 is simultaneously a @xmath7-basis of @xmath99 , whence @xmath105 in particular . it turns out that among the various orders of @xmath99 there is one _ maximal order _ which contains all the other orders , namely the ring of integers @xmath106 in @xmath99 ; see ( * ? ? ? * chapter 2 , section 2 ) . for cyclotomic fields , one has the following well - known result . * theorem 2.6 and proposition 2.16)[p1 ] for @xmath107 , one has : * @xmath108 is the ring of cyclotomic integers in @xmath76 , and hence its maximal order . * @xmath109 is the ring of integers in @xmath94 , and hence its maximal order . [ cyclosec ] if @xmath110 , then @xmath111 . the assertion follows from similar arguments as in the proof of the special case @xmath112 ; compare ( * ? ? ? vi.3 , corollary 3.2 ) . here , one has to observe @xmath113 and then to employ the identity @xmath114 instead of merely using the multiplicativity of the arithmetic function @xmath77 . [ incl ] let @xmath110 . the following statements are equivalent : * @xmath115 . * @xmath116 , or @xmath117 and @xmath118 . for direction ( ii ) @xmath119 ( i ) , the assertion is clear if @xmath116 . further , if @xmath120 , say @xmath121 for a suitable odd number @xmath122 , and @xmath118 , then @xmath123 ( due to @xmath124 ) . however , fact [ gau ] shows that the inclusion of fields @xmath125 can not be proper since we have , by means of the multiplicativity of @xmath77 , the equation @xmath126 . this gives @xmath127 . for direction ( i ) @xmath119 ( ii ) , suppose @xmath115 . then , lemma [ cyclosec ] implies @xmath128 , whence @xmath129 by fact [ gau ] again . using the multiplicativity of @xmath77 together with @xmath130 for @xmath80 and @xmath131 , we see that , given the case @xmath132 , equation ( [ equation ] ) can only be fulfilled if @xmath117 and @xmath118 . the remaining case @xmath133 is equivalent to the relation @xmath116 . [ unique ] let @xmath110 . the following statements are equivalent : * @xmath134 . * @xmath135 , or @xmath136 is odd and @xmath137 , or @xmath70 is odd and @xmath138 . [ unrem ] corollary [ unique ] implies that , for @xmath139 , one has the identity @xmath134 if and only if @xmath135 . [ unique2 ] let @xmath110 with @xmath140 . then , one has : * @xmath141 * @xmath142 for claim ( a ) , let us suppose @xmath143 first . then , fact [ gau ] and corollary [ cr5 ] imply that @xmath144=[\mathbbm{k}_{n}:\mathbbm{k}]=2 $ ] . note that @xmath111 is a cyclotomic field containing @xmath145 . it follows that either @xmath146 or @xmath147 and hence @xmath148 , since the latter is the only real cyclotomic field . now , this implies @xmath149 ; see also lemma [ phin2p](a ) below . the other direction is obvious . claim ( b ) follows immediately from the part ( a ) . [ phin2p ] consider @xmath77 on @xmath150 . then , one has : * @xmath151 if and only if @xmath152 . * @xmath153 if and only if @xmath154 . the equivalences follow from the multiplicativity of @xmath77 in conjunction with the identity @xmath130 for @xmath80 and @xmath131 . let @xmath155 . by corollary [ cr5 ] , for @xmath92 , the field extension @xmath96 is a galois extension with abelian galois group @xmath156 of order @xmath157 . using lemma [ phin2p ] , one sees that @xmath156 is trivial if and only if @xmath158 , and simple if and only if @xmath159 , with @xmath160 as defined in lemma [ phin2p](b ) . the following notions will be of crucial importance . [ algdeldef ] for a set @xmath161 , we define the following properties : * @xmath162<\infty$ ] . * for all @xmath163 , there is a non - singular affine transformation @xmath164 such that @xmath165 . moreover , @xmath6 is called _ degenerate _ when @xmath166 ; otherwise , @xmath6 is non - degenerate . [ mrs ] if @xmath161 satisfies property ( alg ) , then one has @xmath167<\infty$ ] , i.e. , @xmath168 is a real algebraic number field . before we turn to examples of planar sets @xmath6 having properties ( alg ) and ( aff ) , let us prove the central result of this text , where we use arguments similar to the ones used by gardner and gritzmann in the proof of ( * ? ? ? * theorem 4.1 ) . [ charac ] let @xmath161 be non - degenerate with property ( aff ) . further , let @xmath169 with @xmath170 . the following statements are equivalent : * there is an affinely regular @xmath136-gon in @xmath6 . * @xmath171 . if @xmath6 additionally fulfils property ( alg ) , then it only contains affinely regular @xmath136-gons for finitely many values of @xmath136 . for ( i ) @xmath119 ( ii ) , let @xmath55 be an affinely regular @xmath136-gon in @xmath6 . there is then a non - singular affine transformation @xmath172 with @xmath173 , where @xmath174 is the regular @xmath136-gon with vertices given in complex form by @xmath175 . if @xmath176 , condition ( ii ) holds trivially . suppose @xmath177 . the pairs @xmath178 , @xmath179 lie on parallel lines and so do their images under @xmath46 . therefore , @xmath180 moreover , since @xmath181 and @xmath182 are elements of @xmath183 and since @xmath184 for @xmath15 , we get the relation @xmath185 the pairs @xmath186 , @xmath187 also lie on parallel lines . an argument similar to that above yields @xmath188 by subtracting these equations , one gets the relation @xmath189 whence @xmath190 , the latter being equivalent to the inclusion of the fields @xmath191 . for ( ii ) @xmath119 ( i ) , let @xmath174 again be the regular @xmath136-gon as defined in step ( i ) @xmath119 ( ii ) . since @xmath170 , the set @xmath178 is an @xmath192-basis of @xmath71 . since @xmath6 is non - degenerate , there is an element @xmath193 with non - zero imaginary part . hence , one can define an @xmath192-linear map @xmath194 as the linear extension of @xmath195 and @xmath196 . since @xmath197 is an @xmath192-basis of @xmath71 as well , this map is non - singular . since @xmath171 and since @xmath178 is a @xmath198-basis of @xmath199 ( cf . corollary [ cr5 ] ) , the vertices of @xmath200 , i.e. , @xmath201 , lie in @xmath63 , whence @xmath200 is a polygon in @xmath63 . by property ( aff ) , there is a non - singular affine transformation @xmath202 such that @xmath203 is a polygon in @xmath6 . since compositions of non - singular affine transformations are non - singular affine transformations again , @xmath203 is an affinely regular @xmath136-gon in @xmath6 . for the additional statement , note that , since @xmath6 has property ( alg ) , one has @xmath204<\infty$ ] by remark [ mrs ] . thus , @xmath205 has only finitely many intermediate fields . the assertion now follows immediately from condition ( ii ) in conjunction with corollary [ unique ] , remark [ unrem ] and lemma [ unique2 ] . let @xmath206 be an imaginary algebraic number field with @xmath207 and let @xmath208 be the ring of integers in @xmath206 . then , every translate @xmath6 of @xmath206 or @xmath208 is non - degenerate and satisfies the properties ( alg ) and ( aff ) . to this end , we first show that in both cases one has @xmath209 . if @xmath6 is a translate of @xmath206 , this follows immediately from the calculation @xmath210 if @xmath6 is a translate of @xmath208 , one has to observe that @xmath211 since @xmath207 implies @xmath212 . in the first case , property ( aff ) is evident , whereas , if @xmath6 is a translate of @xmath208 , property ( aff ) follows from the fact that there is always a @xmath104-basis of @xmath208 that is simultaneously a @xmath7-basis of @xmath206 . thus , if @xmath213 is a finite set , then a suitable translate of @xmath214 is contained in @xmath6 , where @xmath215 is defined as the least common multiple of the denominators of the @xmath7-coordinates of the elements of @xmath216 with respect to a @xmath7-basis of @xmath206 that is simultaneously a @xmath104-basis of @xmath208 . hence , for these two examples , property ( aff ) may be replaced by the stronger property * for all @xmath163 , there is a homothety @xmath48 such that @xmath165 . thus , we have obtained the following consequence of theorem [ charac ] . [ numcor ] let @xmath206 be an imaginary algebraic number field with @xmath207 and let @xmath208 be the ring of integers in @xmath206 . let @xmath6 be a translate of @xmath206 or a translate of @xmath208 . further , let @xmath169 with @xmath170 . denoting the maximal real subfield of @xmath206 by @xmath217 , the following statements are equivalent : * there is an affinely regular @xmath136-gon in @xmath6 . * @xmath218 . further , @xmath6 only contains affinely regular @xmath136-gons for finitely many values of @xmath136 . in particular , corollary [ numcor ] applies to translates of imaginary cyclotomic fields and their rings of integers , with @xmath219 for a suitable @xmath92 ; cf . facts [ gau ] and [ p1 ] and also compare the equivalences of corollary [ th1mod ] below . remarkably , there are delone subsets of the plane satisfying properties ( alg ) and ( hom ) . these sets were introduced as _ algebraic delone sets _ in ( * ? ? ? * definition 4.2 ) . note that algebraic delone sets are always non - degenerate , since this is true for all relatively dense subsets of the plane . it was shown in ( * ? ? ? * proposition 4.15 ) that the so - called _ cyclotomic model sets _ @xmath6 are examples of algebraic delone sets ; cf . section 4 of @xcite and ( * ? ? ? * definition 4.6 ) in particular for the definition of cyclotomic model sets . any cyclotomic model set @xmath6 is contained in a translate of @xmath220 , where @xmath92 , in which case the @xmath104-module @xmath220 is called the _ underlying @xmath104-module _ of @xmath6 . with the exception of the crystallographic cases of translates of the square lattice @xmath221 and translates of the triangular lattice @xmath222 , cyclotomic model sets are aperiodic ( i.e. , they have no translational symmetries ) and have long - range order ; cf . * remarks 4.9 and 4.10 ) . well - known examples of cyclotomic model sets with underlying @xmath104-module @xmath220 are the vertex sets of aperiodic tilings of the plane like the ammann - beenker tiling @xcite ( @xmath223 ) , the tbingen triangle tiling @xcite ( @xmath224 ) and the shield tiling @xcite ( @xmath225 ) ; cf . * example 4.11 ) for a definition of the vertex set of the ammann - beenker tiling and see figure [ fig : ab2 ] and ( * ? ? ? * figure 1 ) for illustrations . for further details and illustrations of the examples of cyclotomic model sets mentioned above , we refer the reader to ( * ? ? ? * section 1.2.3.2 ) . clearly , any cyclotomic model set @xmath6 with underlying @xmath104-module @xmath220 satisfies @xmath226 whence @xmath227 . it was shown in ( * ? ? ? * lemma 4.14 ) that cyclotomic model sets @xmath6 with underlying @xmath104-module @xmath220 even satisfy the following stronger version of property ( hom ) above : * for all @xmath228 , there is a homothety @xmath48 such that @xmath165 . this property enables us to prove the following characterization . = 0.57 [ th1mod ] let @xmath229 with @xmath230 . further , let @xmath6 be a cyclotomic model set with underlying @xmath104-module @xmath220 . the following statements are equivalent : * there is an affinely regular @xmath136-gon in @xmath6 . * @xmath231 . * @xmath232 , or @xmath127 . * @xmath232 , or @xmath116 , or @xmath233 with @xmath234 an odd divisor of @xmath70 . * @xmath232 , or @xmath235 . * @xmath236 . direction ( i ) @xmath119 ( ii ) is an immediate consequence of theorem [ charac ] in conjunction with relation . for direction ( ii ) @xmath119 ( i ) , let @xmath174 again be the regular @xmath136-gon as defined in step ( i ) @xmath119 ( ii ) of theorem [ charac ] . since @xmath230 , the sets @xmath178 and @xmath237 are @xmath192-bases of @xmath71 . hence , one can define an @xmath192-linear map @xmath194 as the linear extension of @xmath195 and @xmath238 . clearly , this map is non - singular . since @xmath231 and since @xmath178 is a @xmath198-basis of @xmath199 ( cf . corollary [ cr5 ] ) , the vertices of @xmath200 , i.e. , @xmath201 , lie in @xmath76 , whence @xmath200 is a polygon in @xmath76 . because @xmath6 has property ( ) , there is a homothety @xmath239 such that @xmath240 is a polygon in @xmath6 . since homotheties are non - singular affine transformations , @xmath240 is an affinely regular @xmath136-gon in @xmath6 . as an immediate consequence of lemma [ unique2](b ) , we get the equivalence ( ii ) @xmath241 ( iii ) . conditions ( iii ) and ( iv ) are equivalent by lemma [ incl ] . finally , the equivalences ( iii ) @xmath241 ( v ) and ( ii ) @xmath241 ( vi ) follow immediately from fact [ p1 ] . although the equivalence ( i ) @xmath241 ( iv ) in corollary [ th1mod ] is fully satisfactory , the following consequence deals with the two cases where condition ( ii ) can be used more effectively . [ cor2 ] let @xmath229 with @xmath230 . further , let @xmath6 be a cyclotomic model set with underlying @xmath104-module @xmath220 . consider @xmath77 on @xmath150 . then , one has : * if @xmath152 , there is an affinely regular @xmath136-gon in @xmath6 if and only if @xmath232 . * if @xmath159 , there is an affinely regular @xmath136-gon in @xmath6 if and only if @xmath242 by lemma [ phin2p](a ) , @xmath152 is equivalent to @xmath151 , thus condition ( ii ) of corollary [ th1mod ] specializes to @xmath243 , the latter being equivalent to @xmath244 , which means @xmath176 ; cf . corollary [ cr5 ] . this proves the part ( a ) . by lemma [ phin2p](b ) , @xmath159 is equivalent to @xmath245 . by corollary [ cr5 ] , this shows that @xmath246= \phi(n)/2 \in \mathcal{p}$ ] . hence , by condition ( ii ) of corollary [ th1mod ] , one either gets @xmath247 or @xmath248 . the former case implies @xmath176 as in the proof of the part ( a ) , while the proof follows from lemma [ unique2](a ) in conjunction with corollary [ unique ] in the latter case . as mentioned above , the vertex set @xmath249 of the ammann - beenker tiling is a cyclotomic model set with underlying @xmath104-module @xmath250 . since @xmath251 , corollary [ cor2 ] now shows that there is an affinely regular @xmath136-gon in @xmath249 if and only if @xmath252 ; see figure [ fig : ab2 ] for an affinely regular @xmath3-gon in @xmath249 . the other solutions are rather obvious , in particular the patch shown also contains the regular @xmath253-gon @xmath254 , given by the @xmath253th roots of unity . for further illustrations and explanations of the above results , we refer the reader to ( * ? ? ? * section 2.3.4.1 ) or ( * ? ? ? * section 5 ) . this references also provide a detailed description of the construction of affinely regular @xmath136-gons in cyclotomic model sets , given that they exist . _ discrete tomography _ is concerned with the inverse problem of retrieving information about some finite object from information about its slices ; cf . @xcite and also see the refences therein . a typical example is the _ reconstruction _ of a finite point set from its ( _ discrete parallel _ ) _ @xmath255-rays _ in a small number of directions . in the following , we restrict ourselves to the planar case . [ xray .. ] let @xmath256 , let @xmath257 be a direction and let @xmath258 be the set of lines in direction @xmath60 in @xmath19 . then , the ( _ discrete parallel _ ) _ x - ray _ of @xmath216 _ in direction _ @xmath60 is the function @xmath259 , defined by @xmath260 in @xcite , we studied the problem of _ determining _ convex subsets of algebraic delone sets @xmath6 by @xmath255-rays . solving this problem amounts to find small sets @xmath56 of suitably prescribed @xmath6-directions with the property that different convex subsets of @xmath6 can not have the same @xmath255-rays in the directions of @xmath56 . more generally , one defines as follows . let @xmath261 , and let @xmath262 . further , let @xmath263 be a finite set of directions . we say that @xmath264 is _ determined _ by the @xmath255-rays in the directions of @xmath56 if , for all @xmath265 , one has @xmath266 let @xmath161 be a delone set and let @xmath54 be a set of two or more pairwise non - parallel @xmath6-directions . suppose the existence of a @xmath56-polygon @xmath55 in @xmath6 . partition the vertices of @xmath55 into two disjoint sets @xmath267 , where the elements of these sets alternate round the boundary of @xmath55 . since @xmath55 is a @xmath56-polygon , each line in the plane parallel to some @xmath58 that contains a point in @xmath268 also contains a point in @xmath269 . in particular , one sees that @xmath270 . set @xmath271 and , further , @xmath272 and @xmath273 . then , @xmath216 and @xmath274 are different convex subsets of @xmath6 with the same @xmath255-rays in the directions of @xmath56 . we have just proven direction ( i ) @xmath119 ( ii ) of the following equivalence , which particularly applies to cyclotomic model sets , since any cyclotomic model set is an algebraic delone set by ( * ? ? ? * proposition 4.15 ) . * theorem 6.3)[characungen ] let @xmath6 be an algebraic delone set and let @xmath54 be a set of two or more pairwise non - parallel @xmath6-directions . the following statements are equivalent : * @xmath275 is determined by the @xmath255-rays in the directions of @xmath56 . * there is no @xmath56-polygon in @xmath6 . [ trivrem ] trivially , any affinely regular @xmath136-gon @xmath55 in @xmath6 with @xmath136 even is a @xmath56-polygon in @xmath6 with respect to any set @xmath276 of pairwise non - parallel directions having the property that each element of @xmath56 is parallel to one of the edges of @xmath55 . the set @xmath56 then consists only of @xmath6-directions and , moreover , satisfies @xmath277 . by combining corollary [ th1mod ] , direction ( i ) @xmath119 ( ii ) of theorem [ characungen ] and remark [ trivrem ] , one immediately obtains the following consequence . [ corofin ] let @xmath92 and let @xmath6 be a cyclotomic model set with underlying @xmath104-module @xmath220 . suppose that there exists a natural number @xmath278 such that , for any set @xmath56 of @xmath279 pairwise non - parallel @xmath6-directions , the set @xmath275 is determined by the @xmath255-rays in the directions of @xmath56 . then , one has @xmath280 in the situation of corollary [ corofin ] , the question of existence of a suitable number @xmath278 is a much more intricate problem . so far , it has only been answered affirmatively by gardner and gritzmann in the case of translates of the square lattice ( @xmath281 ) , whence corresponding results hold for all translates of planar lattices , in particular for translates of the triangular lattice ( @xmath282 ) ; cf . * theorem 5.7(ii ) and ( iii ) ) . more precisely , it is shown there that , for these cases , the number @xmath283 is the smallest among all possible values of @xmath279 . it would be interesting to know if suitable numbers @xmath278 exist for all cyclotomic model sets . let us finally note the following relation between @xmath56-polygons and affinely regular polygons . the proof uses a beautiful theorem of darboux @xcite on second midpoint polygons ; cf . @xcite or ( * ? ? ? * proposition 4.2)[uaffine ] if @xmath54 is a finite set of directions , there exists a @xmath56-polygon if and only if there is an affinely regular polygon such that each direction of @xmath284 is parallel to one of its edges . a @xmath56-polygon need not itself be an affinely regular polygon , even if it is a @xmath56-polygon in a cyclotomic model set ; cf . * example 4.3 ) for the case of planar lattices and ( * ? ? ? * example 2.46 ) or ( * ? ? ? * example 1 ) for related examples in the case of aperiodic cyclotomic model sets . i am indebted to michael baake , richard j. gardner , uwe grimm and peter a. b. pleasants for their cooperation and for useful hints on the manuscript . interesting discussions with peter gritzmann and barbara langfeld are gratefully acknowledged .	the affinely regular polygons in certain planar sets are characterized . it is also shown that the obtained results apply to cyclotomic model sets and , additionally , have consequences in the discrete tomography of these sets .
let @xmath0 denote the orientable surface of genus @xmath1 with @xmath5 punctures , and let @xmath6 be the corresponding curve complex . finally , let @xmath7 denote the corresponding @xmath8-skeleton . let @xmath9 be a sequence of train tracks on @xmath0 such that @xmath10 is carried by @xmath11 for each @xmath12 . such a collection of train tracks defines a subset of @xmath13 called a _ nested train track sequence_. a _ train track splitting sequence _ is an important special case of such a sequence , in which @xmath11 is obtained from @xmath14 via one of two simple combinatorial moves , _ splitting _ and _ sliding_. a nested train track sequence is said to have @xmath15-bounded steps if the @xmath16-distance between the vertex cycles of @xmath11 and those of @xmath10 is bounded above by @xmath15 . masur - minsky @xcite showed that any nested train track sequence with @xmath15-bounded steps is a @xmath17-quasigeodesic . our first result provides some effective control on @xmath18 as a function of @xmath1 and @xmath2 : let @xmath19 . there exists a function @xmath20 such that any nested train track sequence with @xmath15-bounded steps is a @xmath21-unparameterized quasi - geodesic of the curve graph @xmath22 , which is @xmath21-quasiconvex . masur - mosher - schleimer @xcite used masur and minsky s result to show that if @xmath23 is any essential subsurface , then a sliding and splitting sequence on @xmath0 maps to a uniform unparameterized quasi - geodesic under the subsurface projection map to @xmath24 . using theorem @xmath25 , we show : there exists a function @xmath26 satisfying the following . suppose @xmath23 is an essential subsurface , and let @xmath9 be a splitting and sliding sequence of birecurrent train tracks on @xmath0 . then @xmath9 projects to an @xmath27-unparameterized quasi - geodesic in @xmath28 . let @xmath29 denote the genus @xmath1 handlebody and let @xmath30 denote the set of _ meridians _ , curves on @xmath31 that bound disks in @xmath29 . also due to masur and minsky @xcite is the fact that any two meridians in @xmath32 can be connected by a @xmath33-bounded nested train track sequence . therefore , as a corollary of theorem @xmath25 , we obtain : there exists a function @xmath34 such that @xmath32 is an @xmath35-quasiconvex subset of @xmath36 . the _ mapping class group _ , denoted @xmath37 , is the group of isotopy classes of orientation preserving homeomorphisms of a surface @xmath38 ( see @xcite for a thorough exposition ) . as an application of corollary @xmath39 , we obtain a more effective approach for detecting when a pseudo - anosov mapping class @xmath40 is generic . here , _ generic _ means that the stable lamination of @xmath40 is not a limit of meridians ; the term generic is warranted by a theorem of kerckhoff @xcite , which states that the set of all projective measured laminations which _ are _ limits of meridians constitutes a measure @xmath41 subset of @xmath42 , the space of all projective measured laminations on a surface @xmath38 . in what follows , let @xmath43 denote distance in @xmath44 ; when there is no confusion , the reference to @xmath38 will be omitted . masur and minsky @xcite showed that @xmath44 is a @xmath45-hyperbolic metric space . using corollary @xmath46 , work of abrams - schleimer @xcite , and the fact that the curve graphs are uniformly hyperbolic ( shown by the author in @xcite , and independently by bowditch @xcite , clay - rafi - schleimer @xcite , and hensel - przytycky - webb @xcite ) , we have : there exists a function @xmath47 such that @xmath48 is a generic pseudo - anosov mapping class if and only if there exists some @xmath49 such that for all @xmath50 , @xmath51 by the argument of abrams - schleimer , it suffices to take @xmath52 , for @xmath45 the hyperbolicity constant of @xmath16 , and @xmath35 as in the statement of corollary @xmath39 . @xmath53 we also note that quasiconvexity of @xmath32 and the fact that splitting sequences map to quasi - geodesics under subsurface projection are main ingredients in the proof due to masur and schleimer @xcite that the disk complex is @xmath45-hyperbolic . thus , the effective control discussed above is perhaps a first step to studying the growth of the hyperbolicity constant of the disk complex . * how the main theorem is proved . * the proof of theorem @xmath25 relies on the ability to control 1 . the hyperbolicity constant @xmath54 of @xmath16 ; 2 . @xmath55 , a bound on the diameter of a set of vertex cycles of a fixed train track @xmath56 ; and 3 . the `` nesting lemma constant '' @xmath57 . as mentioned above , due to work of the author , bowditch , clay - rafi - schleimer and hensel - przytycky - webb , curve graphs are uniformly hyperbolic . furthermore , hensel - przytycky - webb @xcite show that all curve graphs are @xmath58-hyperbolic . regarding @xmath59 , the author @xcite has also shown that for sufficiently large @xmath60 . therefore , all that remains is to analyze the growth of @xmath57 , which we address in section @xmath61 by following masur and minsky s original argument while keeping track of the constants that pop up along the way . however , in order to do this , we have need of an effective criterion for determining when a train track @xmath62 is non - recurrent , which we address in section @xmath63 . * organization of paper . * in section @xmath64 , we review some preliminaries about curve complexes and subsurface projections . in section @xmath65 , we review train tracks on surfaces and bounds on curve graph distance given by intersection number , as obtained in previous work . in section @xmath63 , we obtain an effective way of detecting non - recurrence of train tracks by analyzing the linear algebra of the corresponding branch - switch incidence matrix . in section @xmath61 , we obtain an effective version of masur and minsky s nesting lemma , which is the main tool needed to prove theorem @xmath25 . in section @xmath66 we complete the proofs of theorems @xmath25 , @xmath46 , and corollary @xmath39 . * acknowledgements . * the author would primarily like to thank his adviser , yair minsky , for his guidance and for many helpful suggestions . he would also like to thank ian biringer , catherine pfaff , saul schleimer , and harold sultan for their time and for the many motivating conversations they ve had with the author regarding this work . let @xmath67 , @xmath68 be metric spaces . for some @xmath69 , a relation @xmath70 is a @xmath8-_quasi - isometric embedding _ of @xmath71 into @xmath72 if for any @xmath73 we have @xmath74 since @xmath75 is not necessarily a map , @xmath76 need not be singletons , and the distance @xmath77 is defined to be the diameter in the metric @xmath78 of the union @xmath79 . if the @xmath8-neighborhood of @xmath80 is all of @xmath72 , then @xmath75 is a @xmath8-_quasi - isometry _ between @xmath71 and @xmath72 , and we refer to @xmath71 and @xmath72 as being _ quasi - isometric_. given an interval @xmath81\in \mathbb{z}$ ] , a @xmath8-_quasi - geodesic _ in @xmath71 is a @xmath8-quasi - isometric embedding @xmath82 \rightarrow x$ ] . if @xmath82\rightarrow x$ ] is any relation such that there exists an interval @xmath83 $ ] and a strictly increasing function @xmath84 \rightarrow \left[a , b\right]$ ] such that @xmath85 is a @xmath8-quasigeodesic , we say that @xmath75 is a @xmath8-_unparameterized quasi - geodesic_. in this case we also require that for each @xmath86 $ ] , the diameter of @xmath87\right)$ ] is at most @xmath8 . we will sometimes refer to a quasi - geodesic by its image in the metric space @xmath71 . a simple closed curve on @xmath0 is @xmath88 if it is homotopically non - trivial , and not homotopic into a neighborhood of a puncture . the _ curve complex _ of @xmath0 , denoted @xmath6 , is the simplicial complex whose vertices correspond to isotopy classes of essential simple closed curves on @xmath0 , and such that @xmath89 vertices span a @xmath8-simplex exactly when the corresponding @xmath89 isotopy classes can be realized disjointly on @xmath0 . @xmath90 is made into a metric space by identifying each simplex with the standard euclidean simplex with unit length edges . let @xmath91 denote the @xmath8-skeleton of @xmath90 . @xmath90 is a locally infinite , infinite diameter metric space . by a theorem of masur and minsky @xcite , @xmath90 is @xmath45-@xmath92 for some @xmath93 , meaning that the @xmath45-neighborhood of the union of any two edges of a geodesic triangle contains the third edge . @xmath90 admits an isometric ( but not properly discontinuous ) action of @xmath37 , and it is a flag complex , so that its combinatorics are completely encoded by @xmath44 , the @xmath94 ; note also that @xmath90 is quasi - isometric to @xmath44 , and therefore to study the coarse geometry of @xmath95 it suffices to consider the curve graph . let @xmath96 denote distance in the curve graph . if @xmath97 , we can consider more general combinatorial complexes , which also allow vertices to represent essential arcs connecting punctures , up to isotopy . as such , define @xmath98 , the _ arc and curve complex of @xmath38 _ to be the simplicial complex whose vertices correspond to isotopy classes of essential simple closed curves and arcs on @xmath38 . as with @xmath90 , two vertices are connected by an edge if and only if the corresponding isotopy classes can be realized disjointly , and the higher dimensional skeleta are defined by requiring @xmath98 to be flag . as with @xmath95 , denote by @xmath99 the @xmath8-skeleton of @xmath98 . it is worth noting the @xmath98 is quasi - isometric to @xmath90 , with quasi - constants not depending on the topological type of @xmath38 . let @xmath100 be an essential , embedded subsurface of @xmath38 which is not a peripheral annulus . then there is a covering space @xmath101 associated to the inclusion @xmath102 . while @xmath101 is not - compact , note that the gromov compactification of @xmath101 is homeomorphic to @xmath72 , and via this homeomorphism we identify @xmath103 with @xmath104 then , given @xmath105 , the _ subsurface projection map _ @xmath106 is defined by setting @xmath107 equal to its preimage under the covering map @xmath108 . technically , this defines a map from @xmath109 into @xmath110 since their may be multiple connected components of the pre - image of a curve or arc , but the image of any point in the domain is a bounded subset of the range . thus to make @xmath111 a map we can simply choose some component of this pre - image for each point in the domain , and then extend the map @xmath111 simplicially to the higher dimensional skeleta . given an arc @xmath112 , there is a closely related simple closed curve @xmath113 , obtained from @xmath114 by surgering along the boundary components that @xmath114 meets . more concretely , let @xmath115 denote a thickening of the union of @xmath114 together with the ( at most two ) boundary components of @xmath38 that @xmath114 meets , and define @xmath116 to be the components of @xmath117 . thus we obtain a _ subsurface projection _ map @xmath118 for @xmath119 any essential subsurface . here , a subsurface is _ essential _ if it is not a thrice punctured sphere or an annulus whose core curve is homotopic into a neighborhood of a puncture of @xmath38 . then given @xmath120 , define @xmath121 by @xmath122 in this section , we recall some basic terminology of train tracks on surfaces ; we refer to penner - harer @xcite and mosher @xcite for a more in - depth discussion . a _ train track _ @xmath123 is an embedded @xmath124-complex whose vertices and edges are called _ switches _ and _ branches _ , respectively . branches are smooth parameterized paths with well - defined tangent vectors at the initial and terminal switches . at each switch @xmath125 there is a unique line @xmath126 such that the tangent vector of any branch incident at @xmath125 coincides with @xmath127 . as part of the data of @xmath62 , we choose a preferred direction along this line at each switch @xmath125 ; a half branch incident at @xmath125 is called _ incoming _ if its tangent vector at @xmath125 is parallel to this chosen direction , and is called _ outgoing _ if it is anti - parallel . therefore at each switch , the incident half branches are partitioned disjointly into two orientation classes , the _ incoming germ _ and _ outgoing germ_. the valence of each switch must be at least @xmath65 unless @xmath62 has a connected component consisting of a simple closed curve ; in this case @xmath62 has one bivalent switch for such a component . finally , we require that every complementary component of @xmath128 has a negative generalized euler characteristic , that is @xmath129 for any complementary component @xmath130 ; here @xmath131 is the usual euler characteristic and @xmath132 is the number of cusps on @xmath133 . a _ train path _ is a path @xmath134 \rightarrow \tau$ ] , smooth on @xmath135 , which traverses a switch only by entering via one germ and exiting from the other ; a _ closed train path _ is a train path with @xmath136 . a _ proper closed train path _ is a closed train path with @xmath137 ; here @xmath138 is the unit tangent vector to the path @xmath139 at time @xmath140 . let @xmath141 denote the set of branches of @xmath62 ; then a non - negative , real - valued function @xmath142 is called a _ transverse measure _ on @xmath62 if for each switch @xmath125 of @xmath62 , we have @xmath143 where @xmath144 is the set of incoming branches , and @xmath145 the set of outgoing ones . these are called the _ switch conditions_. @xmath62 is called _ recurrent _ if it admits a strictly positive transverse measure , that is , one that assigns a positive weight to every branch . a switch of @xmath62 is called _ semi - generic _ if exactly one of the two germs of half branches consists of a single half branch . @xmath62 is called semi - generic if all switches are semi - generic , and @xmath62 is _ generic _ if @xmath62 is semi - generic and each switch has degree at most @xmath65 . @xmath62 is called _ large _ if each connected component of its complement is simply connected . any positive scaling of a transverse measure is also a transverse measure , and therefore the set of all transverse measures , viewed as a subset of @xmath146 is a cone over a compact polyhedron in projective space . let @xmath147 denote the projective polyhedron of transverse measures . a projective measure class @xmath148\in p(\tau)$ ] is called a _ vertex cycle _ if it is an extreme point of @xmath147 . it is worth noting that if @xmath62 is any train track on @xmath38 , there exists a generic , recurrent train track @xmath149 such that @xmath150 . a lamination @xmath151 is _ carried _ by @xmath62 if there is a smooth map @xmath152 called the _ carrying map for _ @xmath151 which is isotopic to the identity , @xmath153 , and such that the restriction of the differential @xmath154 to any tangent line of @xmath151 is non - singular . if @xmath155 is any simple closed curve carried by @xmath62 , then @xmath155 induces an integral transverse measure called the _ counting measure _ , for which each branch of @xmath62 is assigned the natural number equaling the number of times the image of @xmath155 under its carrying map traverses that branch . a subset @xmath156 is called a _ subtrack _ of @xmath62 if it is also a train track on @xmath38 . in this case , we write @xmath157 . given any train track @xmath62 with branch set @xmath141 , we can distinguish branches as being one of three types : if @xmath158 and both half branches of @xmath159 are the only half branch in their respective germs , @xmath159 is called _ large_. if both half branches of @xmath159 are in germs containing more than one half branch , @xmath159 is _ small _ ; otherwise @xmath159 is _ mixed _ ( figure @xmath59 ) . is small ; middle : @xmath160 is mixed ; right : @xmath161 is large.,width=336 ] if @xmath162 $ ] is a vertex cycle of @xmath62 , then there is a unique ( up to isotopy ) simple closed curve @xmath163 such that @xmath155 is carried by @xmath62 , and the counting measure on @xmath155 is an element of @xmath162 $ ] . therefore , if @xmath164,[v_{2}]$ ] are two vertex cycles of @xmath62 , we can define the distance @xmath165,[v_{2}])$ ] between them to be the curve graph distance between their respective simple closed curve representatives : @xmath166,[v_{2}]):= d_{\mathcal{c}}(c(v_{1}),c(v_{2}))\ ] ] using this , we can also define the distance between two train tracks @xmath62 and @xmath149 to be the distance between their vertex cycle sets : @xmath167,[v_{\tau ' } ] ) : [ v_{\tau } ] \hspace{1 mm } \mbox{is a vertex cycle of } \hspace{1 mm } \tau \hspace{1 mm } \mbox{and } \hspace{1 mm } [ v_{\tau ' } ] \hspace{1 mm } \mbox{is a vertex cycle of } \hspace{1 mm } \tau ' \right\}\ ] ] a _ nested train track sequence _ is a sequence @xmath9 on @xmath0 of birecurrent train tracks such that @xmath168 is carried by @xmath169 for each @xmath170 . this in turn determines a collection of vertices in @xmath22 , by associating the track @xmath168 with its collection of vertices . given @xmath171 , a nested train track sequence @xmath9 is said to have @xmath15-_bounded steps _ if @xmath172 for each @xmath12 . an important special case is the example of a _ splitting and sliding sequence_. this is any train track sequence where @xmath11 is obtained from @xmath10 via one of two combinatorial moves , _ splitting _ ( figure @xmath64 ) and _ sliding _ ( figure @xmath65 ) . we will need the following theorem , as seen in previous work of the author @xcite : there exists a natural number @xmath173 such that if @xmath174 , the following holds : suppose @xmath175 is any train track and @xmath164,[v_{2}]$ ] are vertex cycles of @xmath62 . then @xmath166,[v_{2}])\leq 3.\ ] ] let @xmath176 denote the set of strictly positive transverse measures on @xmath62 . there @xmath62 is recurrent if and only if @xmath177 . for @xmath62 a large track , a _ diagonal extension _ @xmath178 of @xmath62 is a track such that @xmath179 and and each branch of @xmath180 has the property that its endpoints are incident at corners of complementary regions of @xmath62 . following masur and minsky @xcite , let @xmath181 denote the set of all diagonal extensions of @xmath62 , and define @xmath182 let @xmath183 be the union of @xmath184 over all large , recurrent subtracks @xmath185 : @xmath186 and define @xmath187 define @xmath188 to be the measures in @xmath189 whose restrictions to @xmath62 are strictly positive , and define @xmath190 we conclude this section with the statement of a previous result of the author @xcite which will be heavily relied upon in section @xmath65 . for @xmath191 , there is some @xmath192 such that if @xmath193 , whenever @xmath194 and @xmath195 , @xmath196 where @xmath197 . in the above , @xmath198 is the _ geometric intersection number _ between @xmath199 and @xmath200 , defined by @xmath201 where the minimum is taken over all @xmath202 isotopic to @xmath199 . we can explicitly write down the function @xmath203 from the statement of theorem @xmath204 . @xmath203 is an upper bound on the girth of a finite graph with at most @xmath205 vertices , and average degree larger than @xmath206 . as seen in fiorini - joret - theis - wood @xcite , @xmath207 @xmath208 this upper bound will be used in section @xmath61 . let @xmath209 be a train track with branch set @xmath141 and switch set @xmath210 . label the branches @xmath211 and switches @xmath212 , and identify @xmath213 with real - valued functions over @xmath141 . then associated to @xmath62 is a linear map @xmath214 , and a corresponding matrix in the standard basis defined by , given @xmath215 , the @xmath216 coordinate of @xmath217 is the sum of the incoming weights , minus the sum of the outgoing weights at the @xmath216 switch , @xmath218 . let @xmath219 denote the strictly positive orthant of @xmath213 , the collection of vectors with all positive coordinates . we call @xmath220 the _ incidence matrix _ for @xmath62 . note that @xmath221 is a transverse measure on @xmath62 if and only if @xmath222 ; thus , @xmath62 is recurrent if @xmath223 intersects @xmath219 non - trivially . as mentioned in the proof of lemma @xmath224 of @xcite , if @xmath225 , then there is some @xmath226 such that @xmath227 here , @xmath228 is the minimum over all coordinates of the vector @xmath229 , and @xmath230 is the standard euclidean norm in @xmath231 . the main goal of this section is to effectivize this statement , that is , to obtain explicit control on the size of @xmath45 as a function of @xmath1 and @xmath2 : let @xmath232 be a non - recurrent train track on @xmath0 , and let @xmath233 . then @xmath234 where @xmath235 is the sup norm on @xmath231 . we begin by observing that non - recurrence is equivalent to the existence of `` extra '' branches , ones that must be assigned @xmath41 by any transverse measure : suppose that for each branch @xmath158 , there is some corresponding transverse measure @xmath236 on @xmath62 such that @xmath237 . then @xmath62 is recurrent . @xmath53 therefore , the existence of a branch @xmath159 which is assigned @xmath41 by every transverse measure on @xmath62 is equivalent to @xmath62 being non - recurrent . we will call such a branch _ invisible_. given @xmath238 , the switch condition at @xmath239 represents a row vector of the matrix corresponding to the linear transformation @xmath220 . this is the vector @xmath240 that has @xmath124 s in the coordinates corresponding to the incoming half branches incident to @xmath239 , and @xmath241 s in the coordinates corresponding outgoing half branches incident to @xmath239 . note that @xmath240 could also have a @xmath242 in place of two @xmath124 s , if both ends of a single branch are incident to @xmath239 . let @xmath243 denote the row space of @xmath220 , the vector space spanned by the row vectors . the following is an immediate corollary of lemma @xmath224 : suppose @xmath158 is an invisible branch . then @xmath159 is not contained in a closed train path . @xmath53 for @xmath159 a branch of @xmath62 , let @xmath244 denote the switches of @xmath62 incident to @xmath159 ; thus @xmath245 or @xmath64 . for @xmath246 , consider the pointed universal cover @xmath247 with associated covering projection @xmath248 . we define @xmath249 to be the set of train paths in @xmath250 emanating from @xmath251 that do not traverse any branch which projects to @xmath159 under @xmath252 . let @xmath253 be the subset of the universal cover consisting of points contained in some train path of @xmath254 . any train path emanating from @xmath251 has a natural choice of orientation , by defining its initial point to be @xmath251 . this induces an orientation on any branch @xmath255 contained in @xmath256 . note that this is well - defined because @xmath250 does not contain closed train paths ( proper or otherwise ) . we say that @xmath254 is _ unidirectional _ if , whenever @xmath257 project to the same branch @xmath255 of @xmath62 , the orientations of @xmath255 induced by @xmath258 and @xmath259 agree . given @xmath215 , define the _ deviation _ of @xmath229 at @xmath238 , denoted by @xmath260 , to be the absolute value of the coordinate of @xmath217 corresponding to @xmath239 . it suffices to assume that , for @xmath229 as in the statement of the theorem , @xmath261 we will use this assumption to obtain a contradiction . since @xmath62 is non - recurrent , it must contain an invisible branch @xmath159 . let @xmath262 be the two ( possibly non - distinct ) switches incident to the invisible branch @xmath159 , @xmath263 corresponding lifts . then at least one of @xmath264 is unidirectional . suppose not . then there exist branches @xmath265 such that @xmath266 project to a branch @xmath267 of @xmath62 with opposite orientations , and similarly for @xmath268 . thus , in @xmath62 there exist two train paths starting from @xmath269 and ending at @xmath267 , but which traverse @xmath267 in opposite directions . concatenating these two paths produces a loop in @xmath62 , which is a train path away from @xmath269 . by the same exact argument , there is another loop containing the switch @xmath270 and the branch @xmath271 , which is a train path away from @xmath270 . we can then concatenate these two paths across the branch @xmath159 to obtain a `` dumb - bell '' shaped closed train path , which contains @xmath159 ( see figure @xmath63 ) . this contradicts lemma @xmath272 . is unidirectional , then there exist non - closed train paths starting and ending at @xmath269 and @xmath270 . joining these paths across @xmath159 yields a closed train path containing @xmath159 , outlined in blue above . , width=336 ] therefore , we assume henceforth that @xmath273 is unidirectional ; let @xmath274 be the projection of @xmath256 to @xmath62 . that @xmath275 is unidirectional will allow us to redefine which half branches are incoming and which are outgoing ( without changing the linear algebraic structure of @xmath220 ) such that each branch of @xmath276 is mixed . more concretely , orient each edge @xmath277 by projecting the orientation on @xmath278 down to @xmath255 , where @xmath279 is any branch of @xmath250 with @xmath280 ; unidirectionality implies that this construction is well - defined . then we simply define a half - branch @xmath281 to be outgoing at a switch @xmath239 if the orientation of @xmath282 coming from @xmath255 points away from @xmath239 , and similarly for incoming branches . note that this is well - defined , in that two half - branches incident to the same switch in distinct germs will be assigned opposing directional classes . this rule then defines an assignment of direction for all half branches of @xmath62 as follows . the half branches of @xmath62 which are not contained in @xmath276 can be partitioned disjointly into two subcollections : the _ frontier _ half branches ( those which are incident to a switch contained in @xmath276 ) , and the @xmath283 half branches ( those for which the incident switch is not contained in @xmath276 ) . once directions have been assigned to the half branches of @xmath276 as above , directions for frontier half branches are determined by which germ they belong to at the corresponding switch . for interior half branches , simply assign the original directions coming from @xmath62 . let @xmath284 denote the switches of @xmath62 contained in @xmath276 , and recall that @xmath240 denotes the row vector of @xmath220 corresponding to the switch @xmath238 . the vector @xmath285 is a non - zero integer vector , all of whose coordinates are non - negative . since every branch of @xmath276 is mixed , each component of @xmath286 corresponding to a branch of @xmath276 is @xmath41 . the same is true for any branch not in @xmath276 which does not contain a frontier half - branch . we claim that frontier half branches must be incoming at the switch contained in @xmath287 to which it is incident ; this will imply that @xmath286 takes on a positive value for each component corresponding to a branch containing a frontier half branch . indeed , let @xmath255 be a branch containing a frontier half branch , and let @xmath288 be incident to @xmath255 . @xmath289 implies that there is another branch @xmath282 incident to @xmath239 such that @xmath282 is a branch of @xmath276 and @xmath282 is incoming at @xmath239 . thus if @xmath255 were outgoing at @xmath239 , there would exist a train path emanating from @xmath269 which traverses @xmath255 , by contatenating the train path starting at @xmath269 and ending at @xmath282 with the train path connecting @xmath282 to @xmath255 over @xmath239 . this contradicts the assumption that @xmath290 . thus to complete the argument it suffices to show that the collection of frontier half branches is non - empty . recall that @xmath159 is an invisible branch , and is therefore not contained in any closed train path it then follows that the half branch of @xmath159 incident to @xmath269 is frontier . we now use the following elementary fact regarding train tracks on @xmath0 , ( see @xcite for proof ) : let @xmath291 be a train track . then @xmath292 @xmath293 therefore , there are at most @xmath294 row vectors of @xmath220 in the sum @xmath286 . furthermore , since the components of @xmath286 are all non - negative integers , @xmath295 where @xmath296 denotes the standard euclidean dot product . on the other hand , assuming the validity of @xmath297 , one obtains @xmath298 @xmath299 a contradiction . in this section , we will use theorems @xmath204 and @xmath300 to establish the following effective version of masur and minsky s @xcite nesting lemma : there exists a function @xmath301 such that if @xmath178 and @xmath62 are large train tracks and @xmath178 is carried by @xmath62 , and @xmath302 , then @xmath303 when convenient , we will assume our train tracks to be generic ; as mentioned in @xcite , the proof of the nesting lemma in the generic case is easily extendable to the general setting . if @xmath304 , define the _ combinatorial length _ of @xmath305 with respect to @xmath62 , @xmath306 , to be the integral of @xmath305 over @xmath141 , that is @xmath307 we also define @xmath308 where the minimum is taken over all tracks @xmath309 carrying @xmath305 . we will need the following lemma , as seen in hammenstdt @xcite : let @xmath155 be a simple closed curve carried by a train track @xmath62 . then the counting measure on @xmath155 is a vertex cycle of @xmath62 if and only if , for any branch @xmath159 of @xmath62 , the image of @xmath155 under its corresponding carrying map traverses @xmath159 at most twice , and never twice in the same direction . since the vertex cycles are the extreme points of @xmath147 , by the classical krein - milman theorem , any projective transverse measure class can be written as a convex combination of vertex cycles ; that is , given @xmath310 , there exists @xmath311 such that @xmath312 where @xmath313 are the vertex cycles of @xmath62 . any train track on @xmath0 has at most @xmath314 branches , and therefore by lemma @xmath315 , if @xmath62 is any train track and @xmath199 is a vertex cycle , @xmath316 lemma @xmath315 also implies that any train track @xmath62 has at most @xmath317 vertex cycles , since any branch is traversed once , twice , or no times . we therefore conclude that , given @xmath151 as in equation [ 3.1 ] , @xmath318 \max_{i } a_{i}\ ] ] @xmath319 given @xmath320 , there exists a function @xmath321 such that if @xmath322 and @xmath323 , then @xmath324 . suppose @xmath323 . we will abuse notation and refer to the image of @xmath199 under its carrying map by @xmath199 . then every time @xmath199 traverses a branch of @xmath62 , by lemma @xmath315 , it can intersect a vertex cycle at most twice . therefore , if @xmath125 is any vertex cycle of @xmath62 , @xmath325 and hence by theorem @xmath204 , for any @xmath191 and @xmath326 sufficienly large , @xmath327 @xmath328 one needs to be cautious in manipulating the inequality in theorem @xmath204 to obtain equation @xmath329 ; if @xmath330 the direction of the inequality changes and we will not get the desired upper bound on curve graph distance . however , @xmath331 and therefore for sufficiently large @xmath332 this is not an issue . suppose @xmath178 is a large recurrent train track carried by @xmath62 on @xmath0 , and let @xmath333 such that @xmath334 is carried by @xmath149 . then the total number of times , counting multiplicity , that branches of @xmath334 traverse any branch of @xmath335 is bounded above by @xmath336 . the complete argument can be found in masur and minsky s original paper @xcite on the hyperbolicity of the curve complex . for our purposes and for the sake of brevity , it suffices here to simply remark that they show any given branch of @xmath334 can only traverse branches of @xmath335 at most twice . then , since any track has less than @xmath337 branches , the result follows . to prove the following lemma , we use the results from section @xmath63 : there exists @xmath338 with @xmath339 such that if @xmath340 and @xmath178 is large and @xmath62 is generic , @xmath304 and every branch @xmath159 of @xmath341 and @xmath342 of @xmath178 satisfies @xmath343 , then @xmath344 , and @xmath178 is recurrent . we follow masur and minsky s original argument @xcite . the main tools are the elementary moves on train tracks called _ splitting _ and _ shifting _ as introduced in section @xmath65 ( see figures @xmath64 and @xmath65 ) , which can be used to take @xmath62 to a diagonal extension of @xmath178 . in order to do this , we need to move any branch of @xmath341 into a corner of a complementary region of @xmath178 . a split or a shift applied to any such branch either reduces the number of branches of @xmath341 incident to a given branch of @xmath178 , or decreases the distance between a branch of @xmath345 and a corner of a complementary region of @xmath178 . thus , a bounded number of such moves produces a track carried by a diagonal extension of @xmath178 . if a splitting is performed involving a branch @xmath159 of @xmath341 and a branch @xmath155 of @xmath178 , the resulting track contains a new branch @xmath346 of @xmath178 , and we can extend @xmath305 to @xmath346 to be consistent with the switch conditions by assigning @xmath347 . in particular , a sufficient condition for being able to define @xmath305 on the new track is @xmath348 there are at most @xmath337 branches of @xmath341 , and at most @xmath337 branches of @xmath178 or @xmath62 . as earlier mentioned , a splitting move either reduces the number of branches of @xmath341 incident to @xmath178 , or it reduces the number of edges of @xmath178 between a given branch of @xmath345 and a corner that it faces . once a branch of @xmath345 is separated by a corner of a complementary region of @xmath178 by only edges of @xmath178 for which no splitting moves can be performed , a shift move takes such an edge to a corner point . therefore , each edge of @xmath345 is taken to a corner of @xmath178 after no more than @xmath349 shiftings and splittings , and therefore we obtain @xmath149 after at most @xmath350 such moves . now , let @xmath351 , and assume that for this value of @xmath15 , the hypothesis of the statement is satisfied . in light of equation [ 4.5 ] , @xmath305 is definable on the diagonal extension @xmath149 we obtain after splitting and shifting as long as @xmath352 which is precisely what the hypothesis of lemma @xmath353 implies . therefore , @xmath305 is extendable to a diagonal extension of @xmath178 such that all branches receive positive weights , hence @xmath344 . it remains to show that @xmath178 is recurrent ; suppose not . let @xmath354 denote the branch set of @xmath178 . then theorem @xmath300 implies that if @xmath355 is a vector with all positive coordinates , @xmath356 in light of equation @xmath353 , the vector @xmath305 has small deviations , since @xmath305 satisfies the switch conditions on @xmath178 , up to the additive error coming from the weight it assigns to any branch of @xmath341 , which is less than @xmath357 since we assumed that @xmath62 is generic , there are at most two branches of @xmath345 incident to any branch of @xmath178 , and therefore the deviations of @xmath305 are all less than @xmath358 , contradicting theorem @xmath300 . let @xmath320 be given . then there exist functions @xmath359 and @xmath360 satisfying the following : if @xmath178 is large and carried by @xmath62 and @xmath361 such that @xmath149 carries @xmath334 , and if @xmath362 , then any simple closed curve @xmath200 carried on @xmath334 can be written in @xmath363 as @xmath364 , and such that @xmath365 @xmath366 the details of the argument are not entirely relevant for the proof of our main theorem , and can be found in masur - minsky @xcite ; therefore we omit the particulars of the proof , and remark only that in their argument , masur and minsky show that it suffices to take @xmath367 where @xmath368 is the constant from equation [ 3.3 ] , @xmath369 is the constant from the statement of lemma @xmath370 , @xmath371 is a bound on the weights that a vertex cycle can place on any one branch of @xmath334 ( and therefore it suffices to take @xmath372 by lemma @xmath315 ) , and @xmath373 is a bound on the combinatorial length of any vertex cycle on any train track on @xmath0 . putting all of this together , we obtain @xmath374 ( 3(36g+12p-36)^{2})\ ] ] @xmath375 as claimed . they also show that it suffices to take @xmath376 where @xmath377 is a bound on the curve graph distance between any two vertex cycles of the same train track . therefore , by theorem @xmath378 , for sufficiently large @xmath332 , @xmath379 _ proof of lemma @xmath380_. again with concision in mind , we do not include the entirety of masur and minsky s argument ; we simply remark here that in our notation , it suffices to choose @xmath381 here , @xmath369 is as in lemma @xmath370 and is thus bounded above by @xmath382 , @xmath383 , and @xmath384 . thus @xmath385 @xmath386 \cdot ( 36g+12p)\left ( ( 18g+6p)^{18g+6p+2 } \right)^{18g+6p}=:d,\ ] ] and therefore by lemma @xmath387 , for @xmath388 sufficiently large , @xmath389 @xmath390 @xmath391 in this section , we prove the main results : * theorem 1.1 * : _ let @xmath19 . there exists a function @xmath20 such that any nested train track sequence with @xmath15-bounded steps is a @xmath21-unparameterized quasi - geodesic of the curve graph @xmath22 , which is @xmath21-quasiconvex . _ _ proof _ : where possible , we use the same notation that masur and minsky do to avoid confusion . let @xmath45 be the hyperbolicity constant of @xmath44 . by hensel - przytycky - webb @xcite , it suffices to take @xmath392 . let @xmath377 be a bound on the diameter of the set of vertex cycles of a given train track @xmath56 . as mentioned above , for sufficiently large @xmath332 it suffices to take @xmath393 ( see @xcite for a proof of this ) . given a nested train sequence @xmath9 , consider a subsequence @xmath394 such that @xmath395 and such that if @xmath396 is any track not in the subsequence @xmath394 , then there is some @xmath155 for which @xmath397 @xmath408 which implies that @xmath407 is a @xmath409-quasigeodesic . this proves the first part of theorem @xmath25 , with @xmath410 ( we ve shown the sequence to be a @xmath411-quasi - geodesic , but we will need the extra @xmath412 for the quasiconvexity statement ) . we now show @xmath413 is @xmath21-quasiconvex . in any @xmath45-hyperbolic metric space , a geodesic segment @xmath139 connecting the endpoints of a @xmath18-quasigeodesic segment @xmath414 is contained in a @xmath415-neighborhood of @xmath414 , where @xmath416 . @xmath415 is sometimes known as the _ stability constant_. we only give a sketch here ; the main idea of the proof follows an argument on page @xmath419 of ohshika @xcite , and we refer to this for a more complete argument . hyperbolicity of @xmath16 implies the existence of an exponential divergence function ; that is , if @xmath420 are two geodesic rays based at the same point @xmath421 , then there is some exponential function @xmath75 so that for suficiently large @xmath422 ( depending on the choice of geodesic rays ) , the length of any arc outside of a ball of radius @xmath422 centered at @xmath202 , connecting @xmath423 and @xmath424 is at least @xmath425 . let @xmath426 be two elements of a vertex cycle sequence @xmath407 , and let @xmath427 be a geodesic segment connecting them . denote by @xmath428 the @xmath429-quasigeodesic segment obtained by following along the vertex sequence from @xmath202 to @xmath430 . let @xmath431 , and suppose @xmath432 with @xmath433 . let @xmath114 and @xmath159 be two points on @xmath428 whose distance from @xmath239 is @xmath434 and such that @xmath114 and @xmath159 are on different sides of @xmath239 . note that we can assume that such points exist , because the end points of @xmath428 are also the endpoints of @xmath427 , and therefore @xmath239 must be at least @xmath434 from the end points of @xmath428 . let @xmath435 ( resp @xmath342 ) be points located @xmath436 from @xmath239 on either side of @xmath239 on @xmath428 ; if @xmath239 is closer than @xmath436 to one of the endpoints of @xmath428 , simply define @xmath435 ( resp . @xmath342 ) to be this corresponding endpoint of @xmath428 . let @xmath437 be points whose distances are less than @xmath434 from @xmath438 respectively . note that there is an arc @xmath178 joining @xmath430 to @xmath439 , by first connecting @xmath430 to @xmath435 , then @xmath435 to @xmath342 along @xmath428 , and then jumping back over to @xmath427 . thus let @xmath200 be the arc obtained by concatenating the following @xmath61 arcs : the arc along @xmath427 from @xmath114 to @xmath435 , the arc connecting @xmath435 to @xmath430 , the arc along @xmath428 from @xmath430 to @xmath439 , the arc connecting @xmath439 to @xmath342 , and the arc along @xmath427 from @xmath342 to @xmath159 ( see figure @xmath61 ) . we note that the conclusion of lemma @xmath451 is not at all sharp ; indeed , the same argument would have shown that @xmath415 is eventually smaller than @xmath452 for any @xmath453 . however we do not concern ourselves with this , because the contribution to the quasiconvexity of nested sequences coming from @xmath415 will be dominated by a larger term , as will be seen below . we have now shown that the collection of vertices of the sequence @xmath394 is quasiconvex with quasi - convexity constant @xmath454 . it remains to analyze the vertex cycles of tracks that are not in this subsequence . if @xmath125 is such a vertex and @xmath332 is sufficiently large , we know that @xmath125 is within @xmath455 from some vertex of one of the @xmath404 s . in any @xmath45-hyperbolic space , geodesics with nearby end points _ fellow travel _ , in that they remain within a bounded neighborhood of one another , whose diameter depends only on @xmath45 and the distance between endpoints . masur and minsky complete their argument showing the quasiconvexity of @xmath30 by noting that any two disks in @xmath32 can be connected by a path in @xmath32 representing a _ well - nested curve replacement sequence _ , a certain kind of nested train track sequence with @xmath15-bounded steps for which one can take @xmath15 to be @xmath33 . the purpose of this subsection is to prove theorem @xmath46 , which states that the splitting and sliding sequences project to @xmath460-unparameterized quasi - geodesics in the curve graph of any essential subsurface @xmath119 . to do this , we simply follow the original argument of masur - mosher - schleimer @xcite , effectivizing along the way . we first introduce some terminology ; given a subsurface @xmath72 , as in section @xmath64 , let @xmath101 denote the ( non - compact ) covering space of @xmath38 corresponding to @xmath72 . then if @xmath62 is a train track on @xmath38 , let @xmath461 denote the pre - image under the covering projection of @xmath62 to @xmath101 . then let @xmath462 and @xmath463 denote the collection of essential , non - peripheral , simple closed curves ( respectively curves and arcs ) in the gromov compactification of @xmath101 whose interiors are train paths on @xmath461 . let @xmath464 denote the collection of vertex cycles of a track @xmath62 . we first note that any splitting and sliding sequence @xmath9 is a nested train track sequence with @xmath466-bounded steps , for @xmath466 some uniform constant . indeed , if @xmath11 is obtained from @xmath14 by either a splitting or a sliding , any vertex cycle of @xmath11 may intersect a vetex cycle of @xmath14 at most @xmath66 times over any branch of @xmath14 . thus there is some linear function @xmath467 such that @xmath468 for @xmath9 a sliding and splitting sequence on @xmath0 , and @xmath469 ( resp . @xmath470 ) is any vertex cycle of @xmath11 ( resp . @xmath14 ) , and therefore as a consequence of theorem @xmath204 , for sufficiently large @xmath332 , @xmath471 to show that @xmath472 is a @xmath460-unparameterized quasi - geodesic in @xmath24 , we will exhibit a splitting and sliding sequence @xmath473 on @xmath72 such that @xmath474 . then we ll be done by applying theorem @xmath25 to the sequence @xmath475 . given a vertex cycle @xmath199 of @xmath476 , define @xmath477 to be the minimal track carrying @xmath199 ; thus @xmath478 is recurrent by construction , and masur , mosher and schleimer show @xmath478 to be transversely recurrent as well . furthermore , they show that @xmath479 is obtained from @xmath478 by a slide or a split , so long as @xmath480 . therefore @xmath473 constitutes a sliding and splitting sequence of birecurrent train tracks , and thus is a nested train track sequence on @xmath72 with @xmath466- bounded steps . since @xmath478 is a subtrack of @xmath476 , by lemma @xmath315 , any vertex cycle of @xmath478 is a vertex cycle of @xmath476 , and therefore the diameter of @xmath481 is no more than @xmath66 for sufficiently large @xmath332 . since @xmath199 is carried by @xmath476 , it is also carried by @xmath168 . masur , mosher , and schleimer then make use of a lemma which implies the existence of a vertex cycle @xmath482 of @xmath168 which intersects the subsurface @xmath72 essentially . by lemmas @xmath483 and @xmath329 of @xcite , @xmath484 and therefore by lemma @xmath485 and theorem @xmath204 , for @xmath332 sufficiently large , @xmath486 this same argument applies to any vertex cycle of @xmath168 which projects non - trivially to @xmath72 , and thus we conclude that @xmath487 @xmath488 for all @xmath332 sufficiently large . @xmath53 h. masur , y. minsky . _ quasiconvexity in the curve complex_. _ in the tradition of ahlfors and bers , iii_. ( w. abikoff and a. haas , eds . ) , contemporary mathematics * 355 * , amer . math . ( 2004 ) , 309 - 320 .	let @xmath0 denote the genus @xmath1 orientable surface with @xmath2 punctures . we show that nested train track sequences constitute @xmath3-quasiconvex subsets of the curve graph , effectivizing a theorem of masur and minsky . as a consequence , the genus @xmath1 disk set is @xmath4-quasiconvex . we also show that splitting and sliding sequences of birecurrent train tracks project to @xmath3-unparameterized quasi - geodesics in the curve graph of any essential subsurface , an effective version of a theorem of masur , mosher , and schleimer .
entanglement lies at the heart of many aspects of quantum information theory and it is therefore desirable to understand its structure as well as possible . one attempt to improve our understanding of entanglement is the study of our ability to perform information theoretic tasks locally based on anti - de sitter space / conformal field theory ( ads / cft ) conjecture @xcite , such as the strongly coupling , like type ii superconductors @xcite-@xcite . holographic superconductors , quark - gluon plasma , and superconductor / superfluid in condensed matter physics , particularly using qualitative approaches @xcite-@xcite . in @xcite it was shown that entanglement entropy of any entangled cft system could be calculated using the area of the minimal surfaces in bulk . this idea called holographic entanglement entropy ( hee ) and finds a lot of applications @xcite-@xcite , as a geometric approach . in order to address this issues , we consider two possible portions @xmath0(set a ) , @xmath1 ( the complementary set ) of a single quantum system . we make a complete hilbert space @xmath2 . let us to consider the von - neumann entropy @xmath3 , here @xmath4 is the quantum trace of quantum operator @xmath5 over quantum basis @xmath6 . if we compute @xmath7 and @xmath8 , we obtain @xmath9 . a physical meaning is that von - neumann entropies are now more likely to identify it with the surface boundary area @xmath10 @xcite . following to the proposal @xcite , every quantum field theory in @xmath11 dimensional flat boundary has a gravitational dual embedded in @xmath12 bulk . the holographic algorithm tells us that the hee of a region of space @xmath13 and its complement from the @xmath12 geometry of bulk : @xmath14 we should firstly find the minimal @xmath15d mini - super surface @xmath16 . it had been assumed to extend @xmath17 to bulk , but with criteria to keep surfaces with same boundary @xmath18 and @xmath19 . computation of the hee have been initiated @xcite-@xcite . another surprising development is that hee can be used to describe phase transitions in cft systems . here we discuss this issue of observing phase transition in a two dimensional holographic superconductor , and show that in this approach the first order phase transition can be deteced . a toy model for two dimensional holographic superconductors ( 2dhsc ) using the @xmath20 proposed by the following action @xcite-@xcite : @xmath21.\end{aligned}\ ] ] here , @xmath22 defines the three dimensional gravitational constant , the newton constant @xmath23 , @xmath24 is the ads radius , @xmath25 mass of scalar field , and @xmath26 . we adopt the following set of planar coordinates for asymptotically @xmath27 spacetime : @xmath28 to have thermal behavior we choose the temperature for @xmath29 from our @xmath27 black hole following bekenstein - hawking formula : @xmath30 here @xmath31 is the horizon the the blackhole , the largest root of the eq . @xmath32 . to preserve staticity , we select the that the abelien gauge field @xmath33 and scalar field @xmath34 are static ( time independent ) and spherically symmetric : @xmath35 it is useful to write the set of the equations of motion ( eom)s in terms of the new radial coordinate @xmath36 , . using this coordinate the black hole horizon shifted to the @xmath37 and ads boundary locates at @xmath38 . we need to study solutions of the following set of non linear ordinary differential eqs . near @xmath38 : @xmath39+\frac{r_{+}^2\phi}{z^4 } \left[\frac { a^2 e^{\beta } } { f^2}-\frac{m^2}{f}\right]=0~,\label{phiz}\\ & & a ' ' + \frac{a ' } { z } \left[1-\frac{z\beta ' } { 2}\right]-\frac{2 r_{+}^2 a \phi ^2}{z^4f}=0~,\label{az}\\ & & \beta ' - \frac{4 \kappa ^2r_{+}^2}{z^3 } \left[\frac{a ^2 \phi^2 e^{\beta}}{f^2}- \frac{z^4\phi ' ^2}{r_{+}^2}\right]=0\label{betaz } , \\ & & f'-\frac{2 r_{+}^2}{l^2z^3}- \kappa ^2 z e^{\beta } a'^2-\frac{2 \kappa ^2 m^2 r_{+}^2\phi ^2}{z^3 } \\&&\nonumber-\frac{2 \kappa ^2 r_{+}^2}{z^3 } \left[\frac { a ^2 \phi^2 e^{\beta}}{f}+\frac{f \phi ' ^2z^4}{r_{+}^2}\right]=0,\label{fz}\end{aligned}\ ] ] the numerical and analytical solutions of the above eqs . with @xmath40 were presented in ref . we need to specify the bulk system to find minimal area surfaces . the appropriate systems parametrization i d to represent bulk sector in one degree of freedom @xmath41 . using metric given in eq . ( [ g ] ) , the total length ( angle ) @xmath42 and hee are defined by the following integrals : @xmath43 we need to evaluate the sensitivity of ( [ theta0],[shee ] ) in the bulk of acute regimes of temperature.we apply the domain wall approximation analysis @xcite which is based on the idea is proposed to investigate some aspects of the hee along renormalization group ( rg ) trajectories . the @xmath27 is relating to rg flows in @xmath44 . susy deformation of @xmath45 susy - ym theory . ] the geometry ( metric ) which we will use is called here domain wall geometry , in which we suppose that two pats of the ads are connected by a wall with two ads radius @xmath46 . furthermore we suppose that the length scale of ads @xmath24 , is defined as the following : @xmath47 we have a sharp phase transition between two patches of the ads space time , so we locate the dw at the radius @xmath48 . in this approach we define @xmath49 as a possible `` turning '' point of the minimal surface @xmath16 . it is defined by @xmath50 . we will suppose that @xmath51 .. we replaced the integrating out to @xmath52 by integrating out to large positive radius @xmath53 . indeed , we assume that @xmath53 stands out for uv cutoff @xcite . we can rewrite hee as we like : @xmath54,\\ & & \label{sir}s_{ir}=\int_{r_}^{r_{dw}}{\frac{rdr}{\sqrt{f_{ir}(r^2-l_{ir}^2)}}},\\&&\label{suv } s_{uv}=\int_{r_{dw}}^{r_{uv}}{\frac{rdr}{\sqrt{f_{uv}(r^2-l_{uv}^2)}}}.\end{aligned}\ ] ] in both cases ir , uv , the geometry of ads has imposed tight constraints on the metric . furthermore the angle is defined by a two partion integral , @xmath55 : @xmath56,\\ & & \theta_{uv}=\frac{\sqrt{r_{uv}^2-l^2}}{r_{uv}}-\frac{\sqrt{r_{dw}^2-l^2}}{r_{dw}}.\end{aligned}\ ] ] the entangelement entropy can be computed as the following form : @xmath57\end{aligned}\ ] ] the normal phase of the system can even be achieved when we set the scalar field @xmath59 . in this case the ee between @xmath0 and its complement is given by : @xmath60 we rewrite @xmath61 in terms of the coordinate @xmath62 : @xmath63 near the critical point @xmath58 we can approximate the integral as the following : @xmath64 and for the angle , @xmath65 at criticality when @xmath58 we obtain : @xmath66 where @xmath67 we evaluate the hee and angle(length ) numerically with @xmath68 . numeric analysis showed @xcite : when we increase the temperature and dual chemical potential , hee is also increasing smoothly . system becomes a mormal conductor at high temperatures and tends to a superconductor phase when temperature decreases . * if we consider hee in constant temperature ( isothermal curves ) , we showed that there exists at least one lower temperature regime @xmath69 is almost compulsory for superconductivity . furthermore , we conclude that there is no `` confinement / deconfinemnet '' phase transition point in the @xmath29 . * if we fixed the relative chemical potential @xmath70 , we observe that by increasing temperature we have much more amount of hee . the slope of the hee @xmath71 decreases as the relative chemical potential @xmath72 decreases . we interpret this phenomena as emergent of more degrees of freedom ( dof ) at low temperature regimes . * both of the hee and the characterestic length of the entangled system are always increasing with respect to the temperature @xmath73 , and never decreasing for fixed @xmath74 . * hee is linear function of length . one reason is that in small values of belt angle ( small sizes ) the system emerges new extra dof . another reason could be underestood via first law of entanglememt entropy @xcite,@xcite . * when we add the scalar field to the system , @xmath75,if we adjust data as @xmath76 . , we found the critical temperature @xmath77 . it was demonstrated that the system softly has transition from normal phase @xmath78 to the superconductor phase @xmath58 for @xmath79 . * we demonstrated that the slope of the hee with respect to the belt angle @xmath80 remains constant . it means that there is no critical belt length @xmath81 in our @xmath29 system . a numerical study of hee and length shows that these quantities were smooth , and that their behaviors went most smoothly when the phase transition held the same mechanism as the usual . when the temperature of the system @xmath73 rattends to its critical value @xmath74 , we detect a discontinuity in the @xmath82 when the phase is changed , and a first order phase transition may be introduced into the system . by computing the difference between @xmath82 for @xmath83 and @xmath84 at @xmath85 ( scaled the entropy in @xmath86 scale ) we observe that this phase transition is of the first order . indeed , at the critical point @xmath87 and we observe the first order phase transitions from the behavior of the entanglement entropy hee at the critical point @xmath88 . these types of first order phase transitions have been observed recently in literature @xcite . we conclude that the hee is indeed a good probe to phase transition in lower dimensional holographic superconductors . furthermore , it implies that the hee can indicate not only the occurrence of the phase transition , but also we can learn about the order of the phase transition from it . prior work has documented the effectiveness of holographic entanglement entropy in describing phase transition of quantum systems . liu et al . @xcite for example , reports that holographic superconductors in two dimensional systems moving towards the phase transition in lower dimensional quantum field theory systems improved across several studies . however , these studies have either been purely bulk studies or have not focused on thermodynamic whose entanglement entropy was boundary - related . in this study we showed that entanglement entropy predicts a first order phase transition in two dimensional holographic superconductors . s. ryu and t. takayanagi , phys . 96 , 181602 ( 2006 ) . s. ryu and t. takayanagi , j. high energy phys . 0608 , 045 ( 2006 ) . t. nishioka and t. takayanagi , j. high energy phys . 0701 , 090 ( 2007 ) . klebanov , d. kutasov , and a. murugan , nucl . b 796 , 274 ( 2008 ) . a. pakman and a. parnachev , j. high energy phys . 0807 , 097 ( 2008 ) . 13 t. nishioka , s. ryu , and t. takayanagi , j. phys . a 42 , 504008 ( 2009 ) . . hung , r.c . myers , and m. smolkin , j. high energy phys . 1104 , 025 ( 2011 ) . k. bamba , m. jamil , d. momeni and r. myrzakulov , astrophys . space sci . * 344 * , 259 ( 2013 ) doi:10.1007/s10509 - 012 - 1312 - 2 [ arxiv:1202.6114 [ physics.gen-ph ] ] . j. de boer , m. kulaxizi , and a. parnachev , j. high energy phys . 1107 , 109 ( 2011 ) . a. bravetti , d. momeni , r. myrzakulov and h. quevedo , gen . * 45 * , 1603 ( 2013 ) doi:10.1007/s10714 - 013 - 1549 - 2 [ arxiv:1211.7134 [ gr - qc ] ] . n. ogawa and t. takayanagi , j. high energy phys . 1110 , 147 ( 2011 ) . m. jamil , d. momeni and r. myrzakulov , eur . j. c * 72 * , 2075 ( 2012 ) doi:10.1140/epjc / s10052 - 012 - 2075 - 1 [ arxiv:1208.0025 ] . t. albash and c.v . johnson , j. high energy phys . 1202 , 095 ( 2012 ) . a. azadi , d. momeni and m. nouri - zonoz , phys . b * 670 * , 210 ( 2008 ) doi:10.1016/j.physletb.2008.10.054 [ arxiv:0810.4673 [ gr - qc ] ] . myers and a. singh , j. high energy phys . 1204 , 122 ( 2012 ) . t. albash and c.v . johnson , j. high energy phys . 1205 , 079 ( 2012 ) ; arxiv:1202.2605 [ hep - th ] . kuang , e. papantonopoulos , and b. wang , arxiv:1401.5720 [ hep - th ] . t. albash and c. v. johnson , jhep 02 ( 2012 ) 95 . we are indebted to tameem albash for pointing out this comment . m. abramowitz , and i. a. stegun , ( eds . ) . _ handbook of mathematical functions with formulas , graphs , and mathematical tables _ , 9th printing . new york : dover , 1972 .	the purpose of this report is to provide a framework for defining phase transition processes in two dimensional holographic superconductors , and to illustrate how they are useful to be described by holographic entanglement entropy . we study holographic entanglement entropy in a two dimensional fully backrected model for holographic superconductors . we prove that phase transition could be observe using a discontinuty in the first order of entropy .
the electromagnetic interaction is of the local nature , e.g. the interaction hamiltonian in quantum electrodynamics is @xmath1 i.e. the current in a point @xmath2 interacts with an electromagnetic field @xmath3 in the same point . however the process of photon radiation takes place not in one point but in some domain of space - time , in which a photon ( an emitted wave in the classical language ) is originating . the longitudinal dimension of this domain is called _ the formation ( coherence ) length_. it is evident that the minimal size of the formation length is the wave length of the emitted photon . one can estimate the formation length from the phase factor entering the expression for the probability of photon emission . in the quasiclassical operator method ( see eq.([a1 ] ) , appendix a ) this factor is ( in this paper we employ units such that @xmath4 ) @xmath5 , \label{2}\ ] ] where @xmath6 is the energy of the initial ( final ) electron , @xmath7 is the photon energy , @xmath8 , @xmath9 is the wave vector of the photon , @xmath10 is the particle location on a classical trajectory , @xmath11 is the time . the main contribution into the integral over time in the general expression for the probability of radiation ( [ a1 ] ) gives the region @xmath12 taking into account that @xmath13 where @xmath14 is the dielectric constant , @xmath15 is the plasma frequency , @xmath16 is the electron density , we have from eq.([3 ] ) for @xmath17 @xmath18 in this paper we consider the processes at high energies , i.e. the lorentz factor @xmath19 . then from eq.([5 ] ) follows : 1 . ultrarelativistic particle radiates into the narrow cone with the vertex angle @xmath20 along the momentum of the initial particle , the contribution of larger angles is suppressed because of shortening of the formation length ; 2 . the effect of the polarization of a medium described by the dielectric constant @xmath14 manifest itself for soft photons only when @xmath21 , since on the earth @xmath22 ev we have @xmath23 ; 3 . when @xmath24 the formation length is of the order @xmath25 in the classical case @xmath26 one has @xmath27 . so for ultrarelativistic particles the formation length extends substantially . for example for a @xmath28 gev electron emitting a @xmath29 mev photon , @xmath30 m , i.e.@xmath31 interatomic distances . the formation length is important in many electromagnetic processes including creation of electron - positron pair , magnetic bremsstrahlung , transition radiation , @xmath32erenkov radiation . there is also a set of applications involving other forces . there is a number of physical interpretation of the formation length . * the momentum transfer @xmath33 to a nucleus in the process of the photon emission by the particle with the momentum @xmath34 @xmath35 takes the minimal value when it is longitudinal @xmath36 , and in this case @xmath37 , i.e. from the uncertainty principle follows that if the minimal momentum transfer is small , then the formation length is large . * for creation of the electron - positron pair the formation length is the distance required for the final state particles to separate enough that they act as independent particles . * this is the distance over which the amplitudes of several interactions can add coherently to the total probability . * the transverse size of the region , where the radiation is formed @xmath38 is determined by the minimal transverse momentum transfer @xmath39 : @xmath40 . although usually @xmath41 and this dimension is much smaller , we consider below the situation when the constraint of @xmath42 is the main effect . the formation length concept is important because within it the amplitudes can add coherently if a charged particle is moving freely . in the opposite case , when some perturbation is acting on the particle within the formation length , breakdown of the coherence happens and radiation will be suppressed . landau and pomeranchuk were the first to show that if the formation length of the bremsstrahlung becomes comparable to the distance over which the multiple scattering becomes important , the bremsstrahlung will be suppressed @xcite . they considered radiation of soft photons . migdal @xcite , @xcite developed a quantitative theory of this phenomenon . now the common name is the landau- pomeranchuk -migdal ( lpm ) effect . let us estimate a disturbance of the emission process due to a multiple scattering . as it is known , the mean angle of the multiple scattering at some length @xmath43 is @xmath44 where @xmath45 , @xmath46 is the radiation length . since we are interesting in influence of the multiple scattering on the radiation process we put here the formation length ( [ 6 ] ) . one can expect that when @xmath47 this influence will be substantial . from this inequality we have @xmath48 here @xmath49 is the characteristic energy scale , for which the multiple scattering will influence the radiation process for the whole spectrum . it was introduced in @xcite and denoted by @xmath50 . for tungsten we have @xmath51 tev , and similar values for all the heavy elements . for the light elements the energy @xmath49 is much larger . when @xmath52 the multiple scattering will influence emission of soft photons only with energy @xmath53 e.g. for particles with energy @xmath28 gev one has @xmath54 mev . of course , we estimate here the criterion only . for description of the effect one has to calculate the probability of the bremsstrahlung taking into account multiple scattering . side by side with the multiple scattering of an emitting electron one has to take into account also the influence of the medium on the radiated electromagnetic field . as it was shown above eq.([5 ] ) , this effect leads also to suppression of the soft photon emission ( the dielectric suppression , the ter - mikaelian effect , see in @xcite ) . a clear qualitative analysis of different mechanisms of suppression is presented in early papers @xcite,@xcite . more simple derivation of the migdal s results is given in @xcite . the next step in a quantitative theory of the landau - pomeranchuk - migdal ( lpm ) effect was made in @xcite . this theory is based on the quasiclassical operator method in qed developed by the authors ( see @xcite , @xcite , @xcite , @xcite ) . one of the basic equations ( obtained with the use of kinetic equations describing the motion of an electron in the medium in the presence of an external field ) is the schrdinger equation in an external field with imaginary potential ( eq.(3.3 ) of @xcite ; see ( [ a21 ] ) in appendix a ) . new activity with the theory of the lpm effect is connected with a very successful series of experiments @xcite - @xcite performed at slac recently . in these experiments the cross section of the bremsstrahlung of soft photons with energy from 200 kev to 500 mev from electrons with energy 8 gev and 25 gev was measured for a variety in materials with an accuracy of the order of a few percent . both the lpm and the dielectric suppression was observed and investigated . these experiments were the challenge for the theory since in all the published before papers calculations are performed to logarithmic accuracy which is not enough for a description of the new experiments . the contribution of the coulomb corrections ( at least for heavy elements ) is larger then experimental errors and these corrections should be taken into account . in papers @xcite-@xcite the problem was investigated using different approaches . very recently the lpm was studied at cern @xcite using iridium target and electrons with energy 287 gev , 207 gev and 149 gev . for iridium according to eqs.([9 ] ) and ( [ 10 ] ) @xmath49=2.2 tev , @xmath55 37 gev for the highest energy and the lpm suppression of bremsstrahlung was observed in much wider interval of photon energies than at slac . the crossing process for the bremsstrahlung is the pair creation by a photon . the created particles in a medium can undergo the multiple scattering . it should be emphasized that for the bremsstrahlung the formation length ( [ 6 ] ) increases strongly if @xmath26 . just because of this the lpm effect was investigated at slac at a relatively low energy . for the pair creation the formation length @xmath56 attains maximum at @xmath57 and this maximum is @xmath58 . because of this even for heavy elements the effect of multiple scattering on pair creation becomes noticeable at photon energies @xmath59 tev @xcite . starting from these energies one has to take into account the influence of a medium on the pair creation and on the bremsstrahlung hard part of the spectrum in electromagnetic showers being created by the cosmic ray particles of the ultrahigh energies . these effects can be quite significant in the electromagnetic calorimeters operating in the detectors on the colliders in tev range . the nonlinear effects of qed are due to the interaction of a photon with electron - positron field . these processes are the photon - photon scattering , the coherent photon scattering , the photon splitting into two photons , and the coalescence of two photons into photon in the coulomb field . the process of creation of the electron - positron pair by a photon connected with the coherent photon scattering @xcite . in the quasiclassical approximation the amplitude @xmath60 of the coherent photon scattering is described by diagram where the electron - positron pair is created by the initial photon with 4-momentum @xmath61 and then annihilate into the final photon with 4-momentum @xmath62 . for high energy photon @xmath63 this process occurs over a rather long distance , known as the time of life of the virtual state @xmath64 where @xmath65 is the characteristic transverse momentum of the process . when the virtual electron ( or positron ) is moving in a medium it scatters on atoms . the mean square of momentum transfer to the electron from a medium on the distance @xmath66 is @xmath67 where @xmath68 is the charge of nucleus , @xmath69 is the number density of atoms in the medium , @xmath70 is the photon momentum transfer ( @xmath71 ) , @xmath72 is the screening radius of atom . the coherent photon scattering amplitude @xmath60 can be obtained from general formulas for probabilities of electromagnetic processes in the frame of the quasiclassical operator method ( see e.g. @xcite ) . it can be estimated as @xmath73 we use the normalization condition im@xmath74 for the case @xmath75 , where @xmath76 is the total cross section of pair creation by a photon . in the case of small momentum transfer @xmath77 we have in the region of small @xmath78 @xmath79 at an ultrahigh energy it is possible that @xmath80 . in this case the characteristic momentum transfer is @xmath81 , the combination @xmath82 enters in the formation length @xmath83 ( [ 5 ] ) . the self - consistency condition is @xmath84 where @xmath85 is defined in eq.([1.2 ] ) . so using eq.([1.3 ] ) one gets for the estimate of the coherent photon scattering amplitude @xmath60 ( the influence of the multiple scattering manifests itself at the high photon energies such that @xmath86 ) @xmath87 the authors have shown @xcite that an external electromagnetic field can also modify the bremsstrahlung process . we consider the effect of an external field on the photon emission in electron - electron(positron ) collision . this effect is associated as well with a reduction of the formation length of a photon ( either real or virtual ) due to relatively large turn of particle velocity over this length , and the corresponding increase in the vertex angle of the radiation cone . if a photon of energy @xmath88 is emitted by an electron ( positron ) at an angle @xmath89 to its velocity , the formation length of such a photon will be according to eq.([5 ] ) ( we neglect the polarization of a medium ) @xmath90 where @xmath91 defined in eq.([1.5 ] ) . the characteristic emission angle in weak field is @xmath92 , and we can neglect the influence of the external field if @xmath93 substituting eq.([11 ] ) into eq.([12 ] ) , we have the criterion for a field to be weak @xmath94 here we introduced the important parameter @xmath95 , which characterizes the quantum effects in an external field , when @xmath96 we are in the classical domain and with @xmath97 we are already well inside the quantum domain @xmath98 where @xmath99 is an electric ( magnetic ) field , eq.([14 ] ) contains the transverse with respect to particle velocity @xmath100 component of fields , @xmath101 is one of @xmath102 in strong fields , where @xmath103 , with characteristic radiation angles @xmath104 , the effective emission angle @xmath105 is determined by a self - consistency argument : the deviation angle of the particle in the field over the formation length must not exceed @xmath106 , i.e. @xmath107 it can be seen from eq.([16 ] ) that when @xmath103 , neither the characteristic emission angle nor the photon formation length depends on the mass of the radiating particle . a parameter characterizing the effect of an external field on the radiation process was derived in @xcite . it follows from eq.([16 ] ) that , for @xmath108 , the characteristic photon formation length is reduced by factor @xmath109 , and the emission angle is increased by factor @xmath110 . since the relevant parameter @xmath111 depends on the frequency @xmath88 ( via @xmath112 ) , this effect is manifested earlier for soft photons . because of this a specific situation arises in bremsstrahlung in a collision of an electron and a positron(an electron ) in colliding - beam experiment . the point is that the external factors act differently on the radiating particle and on the recoil particle . for the radiating particle the criterion of influence of external factors is the same in scattering by a nucleus as in collision of particles . for the recoil particles the effect turns out to be enhanced by @xmath113 times ( in the laboratory frame ) this is due to the fact that the main contribution to the bremsstrahlung cross section give the virtual photons emitted by recoil particle with very low energy @xmath114 where @xmath7 is the frequency of the real photon . the formation length of the virtual photon is @xmath115 for example , for @xmath116gev and @xmath117mev the formation length is enormous : @xmath118 cm . this is the reason why the field has a significant effect on virtual photon emission in @xmath119 collisions even for moderate field and at relatively low particle energies , when the parameter @xmath120 is small while the value @xmath121 is large . in that event , the picture of emission of a bremsstrahlung photon ( radiation vertex ) does not change directly , but a significant change takes place in the virtual photon spectrum at momentum transfers @xmath122 , increasing the lower bound of the effective momentum transfer and resulting in a corresponding decrease of the cross section . the bremsstrahlung cross section under these conditions was derived to logarithmic accuracy in @xcite using the equivalent photon method . in @xcite the problem was solved to relativistic accuracy ( terms @xmath123 were neglected ) . it follows from eq.([13 ] ) , and was demonstrated in @xcite that in the case @xmath124 there is a change of the radiation vertex describing absorption of a virtual photon and emission of a real one . since for @xmath103 the formation length of a real photon @xmath125 falls off as @xmath126 ( see eq.([16 ] ) ) , the bremsstrahlung cross section falls off in just the same extent . we consider first the case where the formation length is much shorter than the thickness of a target @xmath127 . the derivation of basic formulas for probability of radiation under influence of multiple scattering is given in appendix a. the spectral distribution of the probability of radiation per unit time in this case is given by eq.([a19 ] ) and functions @xmath128 satisfy the eq.([a21 ] ) . introducing new variables @xmath129 where the quantities @xmath72 and @xmath130 are defined in eqs.([a13]),([a14 ] ) , we obtain for the spectral distribution of the probability of radiation per unit time @xmath131 , \label{2.8}\ ] ] where @xmath132 , and the functions @xmath133 now satisfy the equation @xmath134 with the initial conditions @xmath135 . here @xmath136 are defined in eq.([2 ] ) and @xmath137 are defined in eq.([1.2 ] ) . the functions @xmath138 and @xmath139 in eq.([2.8 ] ) are rescaled according with the initial conditions ( factors @xmath140 and @xmath141 , correspondingly ) . note , that it is implied that in formula ( [ 2.8 ] ) subtraction at @xmath142 is made . the potential @xmath143 eq.([2.9 ] ) corresponds to consideration of scattering in the born approximation . the difference of exact ( as a function of @xmath144 ) potential @xmath143 and taken in the born approximation is computed in appendix a of @xcite . the potential @xmath143 with the coulomb corrections taken into account is @xmath145 where @xmath146 , the function @xmath147 is @xmath148= \xi^2\sum_{n=1}^{\infty}\frac{1}{n(n^2+\xi^2 ) } , \label{2.9b}\ ] ] where @xmath149 is the logarithmic derivative of the gamma function . in above formulas @xmath150 is space of the impact parameters measured in the compton wavelengths @xmath151 , which is conjugate to space of the transverse momentum transfers measured in the electron mass @xmath152 . an operator form of a solution of eq . ( [ 2.9 ] ) is @xmath153 where we introduce the following dirac state vectors : @xmath154 is the state vector of coordinate @xmath150 , and @xmath155 . substituting eq.([2.10 ] ) into eq.([2.8 ] ) and taking integral over @xmath11 we obtain for the spectral distribution of the probability of radiation @xmath156 where @xmath157 here and below we consider an expression @xmath158 as a limit : @xmath159 of @xmath160 . let us estimate the effective impact parameters @xmath161 which give the main contribution into the radiation probability . since the characteristic values of @xmath161 can be found straightforwardly by calculation of eq.([2.11 ] ) , we the estimate characteristic angles @xmath162 connected with @xmath161 by an equality @xmath163 . the mean square scattering angle of a particle on the formation length of a photon @xmath66 eq.([5 ] ) has the form @xmath164 where @xmath165 , we neglect here the polarization of a medium . when @xmath166 the contribution in the probability of radiation gives a region where @xmath167 , in this case @xmath168 . when @xmath169 the characteristic angle of radiation is determined by self - consistency arguments ( compare with eq.([16 ] ) ) : @xmath170 it should be noted that when the characteristic impact parameter @xmath161 becomes smaller than the radius of nucleus @xmath171 , the potential @xmath143 acquires an oscillator form ( see appendix b , eq.(b.3 ) in @xcite ) @xmath172 here @xmath173 is the screening radius , @xmath174 is the bohr radius . allowing for the estimates eq.([2.14 ] ) we present the potential @xmath143 eq.([2.9 ] ) in the following form @xmath175 here @xmath176 . so , one can to redefine the parameters @xmath173 and @xmath177 to include the coulomb corrections . the inclusion of the coulomb corrections ( @xmath178 and -1 ) into @xmath179 diminishes effectively the correction @xmath180 to the potential @xmath181 . in accordance with such division of the potential we present the propagators in eq . ( [ 2.11 ] ) as @xmath182 where @xmath183 this representation of the propagator @xmath184 permits one to expand it over the `` perturbation '' @xmath185 . indeed , with an increase of @xmath186 the relative value of the perturbation is diminished @xmath187 since the effective impact parameters diminishes and , correspondingly , the value of logarithm @xmath188 in eq.([2.16 ] ) increases . the maximal value of @xmath188 is determined by the size of a nucleus @xmath171 @xmath189 the value @xmath190 is the important parameter of the radiation theory . the matrix elements of the operator @xmath191 can be calculated explicitly . the exponential parametrization of the propagator is @xmath192 the matrix elements of the operator @xmath193 has the form ( details see in @xcite ) @xmath194 \right\ } , \label{2.24}\end{aligned}\ ] ] where @xmath195 ( see eq.([2.16 ] ) ) . substituting formulas ( [ 2.19 ] ) and ( [ 2.24 ] ) in the expression for the spectral distribution of the probability of radiation ( [ 2.11 ] ) we have @xmath196 } \nonumber \\ & & = r_1\left(\ln p -\psi\left(p+\frac{1}{2 } \right ) \right ) + r_2\left ( \psi\left(p \right)-\ln p+\frac{1}{2p } \right ) \label{2.25}\end{aligned}\ ] ] where @xmath197 , let us remind that @xmath198 is the logarithmic derivative of the gamma function ( see eq.([2.9b ] ) ) . if we substitute in eq.([2.25 ] ) @xmath199 the probability will be transformed into the form containing the real functions only . then it can be written as @xmath200 , \label{2.41}\ ] ] where @xmath201 , \nonumber \\ & & \phi(s)=6s^2\left[\int_{0}^{\infty}\exp ( -sx ) \sin ( sx ) \coth x dx - \frac{\pi}{4}\right ] , \label{2.42}\end{aligned}\ ] ] are the functions introduced by migdal ( see eqs.(46 ) , ( 47 ) in @xcite ) , note that parameter @xmath202 in eq.([2.42 ] ) is two times larger than that used by migdal . if in eqs.([2.41 ] ) , ( [ 2.42 ] ) one omits the coulomb correction , then the probability ( [ 2.41 ] ) coincides formally with the probability calculated by migdal ( see eq.(49 ) in @xcite ) . expanding the expression @xmath203 over powers of @xmath185 we have @xmath204 in accordance with eqs.([2.17 ] ) and ( [ 2.26 ] ) we present the probability of radiation in the form @xmath205 at @xmath206 the expansion ( [ 2.26 ] ) is a series over powers of @xmath207 . it is important that variation of the parameter @xmath161 by a factor order of 1 has an influence on the dropped terms in ( [ 2.26 ] ) only . the probability of radiation @xmath208 is defined by eq.([2.25 ] ) . the term @xmath209 in eq.([2.27 ] ) corresponds to the term linear in @xmath185 in eq.([2.26 ] ) . the explicit formula for the first correction to the probability of radiation @xcite is @xmath210,\quad d_2(\nu_0)=\int_{0}^{\infty}\frac{dz e^{-sz}}{\sinh^3 z } } \nonumber \\ & & \displaystyle{\times \left\{\left[d(z)- \frac{1}{2}g(z)\right]\left(\sin sz+\cos sz\right ) + \frac{\pi}{4}g(z)\left(\cos sz - \sin sz\right)\right\ } } , \nonumber \\ & & \displaystyle{d(z)=(\ln \nu_0 \vartheta(1-\nu_0)-\ln \sinh z -c)g(z ) -2g(z ) \cosh z } , \label{2.30}\end{aligned}\ ] ] where @xmath211 here @xmath212 is the euler dilogarithm ; and @xmath213 the functions @xmath214 and @xmath215 are given in fig . 1 and the functions @xmath216 and @xmath217 are shown in fig . 2 . as it was said above ( see eqs.([2.14 ] ) , ( [ 2.18 ] ) ) , @xmath168 at @xmath218 the logarithmic functions @xmath219 and @xmath190 are defined in eqs.([2.16 ] ) and ( [ 2.18 ] ) . if the parameter @xmath220 , the value of @xmath161 is defined from the eq . ( [ 2.14 ] ) , where @xmath221 , up to @xmath222 . so , we have two representation of @xmath223 depending on @xmath161 : at @xmath224 it is @xmath225 and at @xmath226 it is @xmath227 . the mentioned parameters can be presented in the following form @xmath228 it should be noted that in the logarithmic approximation the parameter @xmath161 entering into the parameter @xmath229 is defined up to the factor @xmath230 . however , we calculated the next term of the decomposition over @xmath180 ( an accuracy up to the `` next to leading logarithm '' ) and this permitted to obtain the result which is independent of the parameter @xmath161 ( in terms @xmath231 ) . our definition of the parameter @xmath161 minimizes corrections to eq.([2.25 ] ) practically for all values of the parameter @xmath161 . the approximate solution of eq.([2.14 ] ) given in eq.([2.30b ] ) has quite good numerical accuracy : it is @xmath232 at @xmath233 and @xmath234 at @xmath235 . the lpm effect manifests itself when @xmath236 so , the characteristic energy @xmath237 is the energy starting from which the multiple scattering distorts the whole spectrum of radiation including its hard part . if the radiation length @xmath46 is taken within the logarithmic approximation the value @xmath237 coincides with @xmath49 eq.([9 ] ) . the difference between these values is @xmath238 . the formulas derived in @xcite,@xcite and written down above are valid for any energy . in fig . 3 the spectral radiation intensity in gold ( @xmath239 tev ) is shown for different energies of the initial electron . in the case when @xmath240 ( @xmath241gev and @xmath242gev ) the lpm suppression is seen in the soft part of the spectrum only for @xmath243 while in the region @xmath244 ( @xmath245tev and @xmath241tev ) where @xmath246 the lpm effect is significant for any @xmath2 . for relatively low energies @xmath241gev and @xmath247gev used in famous slac experiment @xcite , @xcite we have analyzed the soft part of spectrum , including all the accompanying effects : the boundary photon emission , the multiphoton radiation and influence of the polarization of the medium ( see below ) . the perfect agreement of the theory and data was achieved in the whole interval of measured photon energies ( 200 kev@xmath248500 mev ) , see below and the corresponding figures in @xcite,@xcite,@xcite . it should be pointed out that both the correction term with @xmath249 and the coulomb corrections have to be taken into account for this agreement . when a scattering is weak ( @xmath250 ) , the main contribution in eq.([2.30 ] ) gives a region where @xmath251 . then @xmath252 combining the results obtained in eq.([2.33 ] ) we obtain the spectral distribution of the probability of radiation in the case when scattering is weak @xmath253 @xmath254 } \nonumber \\ & & \displaystyle{=\frac{\alpha}{2\pi \gamma^2}\frac{2q}{3}\left[r_1\left ( l_1-\frac{1}{3 } \right)+2r_2\left(l_1+\frac{1}{6 } \right ) \right ] } \nonumber \\ & & \displaystyle{=\frac{4z^2\alpha^3n_a}{3m^2\omega}\bigg[\frac{\omega^2 } { \varepsilon^2 } \left(\ln \left(183z^{-1/3 } \right)-\frac{1}{6 } -f(z\alpha)\right ) } \nonumber \\ & & \displaystyle{+2\left(1+\frac{\varepsilon'^2}{\varepsilon^2 } \right ) \left(\ln \left(183z^{-1/3 } \right)+\frac{1}{12}-f(z\alpha ) \right ) \bigg ] } , \label{2.35}\end{aligned}\ ] ] where @xmath190 is defined in eq.([2.18 ] ) . this expression coincide with the known bethe - maximon formula for the probability of bremsstrahlung from high - energy electrons in the case of complete screening ( if one neglects the contribution of atomic electrons ) written down within power accuracy ( omitted terms are of the order of powers of @xmath255 ) with the coulomb corrections , see e.g. eq.(18.30 ) in @xcite , or eq.(3.83 ) in @xcite . at @xmath256 the function @xmath257 eq.([2.30 ] ) has the form @xmath258 under the same conditions ( @xmath256 ) the function @xmath259 eq.([2.25 ] ) is @xmath260 so , in the region where the lpm effect is strong the probability eq.([2.25 ] ) can written as @xmath261 this means that in this limit the emission probability is proportional to the square root of the density . this fact was pointed out by migdal @xcite , the expression ( [ 2.39a ] ) coincides formally with the probability calculated by migdal ( see eq.(52 ) in @xcite ) with the same remarks which were made after eq.([2.42 ] ) . thus , at @xmath256 the relative contribution of the first correction @xmath209 is defined by @xmath262 where @xmath263 . in this expression the value @xmath264 with the accuracy up to terms @xmath265 does nt depend on the photon energy:@xmath266 . hence we can find the correction to the total probability at @xmath267 . the maximal value of the correction is attained at @xmath268 , it is @xmath269 for heavy elements . in the above analysis we did not consider an inelastic scattering of a projectile on atomic electrons . the potential @xmath270 connected with this process can be found from formula ( [ 2.9a ] ) by substitution @xmath271 ( an analysis of an inelastic scattering on atomic electrons as well as the parameter @xmath272 can be found in @xcite ) . the summary potential including both an elastic and an inelastic scattering is @xmath273 } \nonumber \\ & & \displaystyle{=-q_{ef}\mbox{\boldmath$\varrho$}^2 \big(\ln \gamma^2 \vartheta_{ef}^2 + \ln \frac{\mbox{\boldmath$\varrho$}^2}{4}+2c \big ) } , \label{40a}\end{aligned}\ ] ] where @xmath274}.\ ] ] integrating eq.([2.25 ] ) over @xmath275 we obtain the total intensity of radiation in the logarithmic approximation @xmath276 , \label{2.1i}\end{aligned}\ ] ] where @xmath277 @xmath278 is the radiation length in the logarithmic approximation . the relative energy loss of electron per unit time in terms of the bethe - maximon radiation length @xmath278 : @xmath279 in gold is given in fig . 4 ( curve 1 ) , it reduces by 10% ( 15% and 25% ) at @xmath280 gev ( @xmath281 tev and @xmath282 tev ) respectively , and it cuts in half at @xmath283 tev . this increase of effective radiation length can be important in electromagnetic calorimeters operating in detectors on colliders in tev range as well as in analysis of high - energy cosmic rays . the relative energy loss of electron per unit time in terms of the exact bethe - maximon radiation length @xmath46 in iridium and lead calculated taking into account the correction terms eq.([2.30 ] ) is shown in fig.19 ( see sec.3.10.2 ) . the spectral distribution of bremsstrahlung intensity and the spectral distribution over energy of created electron ( positron ) as well as the reduction of energy loss and the photon conversion cross section was calculated by klein @xcite using the migdal @xcite formulas . as will be explained below ( see sec . 3.7 ) we use more accurate procedure of fine tuning and because of this our calculation in logarithmic approximation differs from migdal one . we calculated also the correction term and include the coulomb corrections . for this reason the results shown here in figs.3 - 4 are more precise than given in @xcite . in the case @xmath240 the correction term to the total intensity of radiation ( or to the radiation length ) can be written as @xcite @xmath284 in the opposite case @xmath267 we have @xcite @xmath285 \label{2.3i}\ ] ] although the coefficients in the last expression are rather large at two first terms of the decomposition over @xmath286 this formula has the accuracy of the order of 10% at @xmath287 . analogous expression for the integral probability of radiation has the form @xmath288 . \label{2.4i}\ ] ] ratio of the main terms of eqs.([2.3i ] ) and ( [ 2.4i ] ) gives the mean energy of radiated photon @xmath289 the probability of the pair creation by a photon can be obtained from the probability of the bremsstrahlung with help of the substitution law : @xmath290 where @xmath7 is the photon energy , @xmath291 is the energy of the particle . making this substitution in eq.([2.11 ] ) we obtain the spectral distribution of the pair creation probability @xcite ( over the energy of the created electron @xmath291 ) @xmath292 where @xmath293 the other notations are given in eqs.([2.12 ] ) , ( [ 2.9 ] ) , ( [ 2.9a ] ) , ( [ 2.9b ] ) . just as in the analysis of the lpm effect for radiation we present the potential eq.([2.9a ] ) in the form @xmath294 and will proceed as we did in eqs.([2.16])- ( [ 2.24 ] ) . substituting the expression ( [ 2.24 ] ) in the formula for the spectral distribution of the pair creation probability ( [ 2.2p ] ) we have @xmath295 } \nonumber \\ & & = s_1\left(\ln p-\psi\left(p+\frac{1}{2}\right ) \right ) + s_2\left(\psi ( p ) -\ln p+\frac{1}{2p}\right ) , \label{2.10p}\end{aligned}\ ] ] where we used the same notations as in eq.([2.25 ] ) . this formula gives the spectral distribution of the pair creation probability in the logarithmic approximation which was used also by migdal @xcite . it should be noted that the parameter @xmath161 entering into the parameter @xmath229 ( see eqs.([2.24 ] ) , ( [ 2.16 ] ) and ( [ 2.14 ] ) ) is defined up to the factor @xmath230 , what is inherent in the logarithmic approximation . however , below we will calculate the next term of the decomposition over @xmath180 ( an accuracy up to the `` next to leading logarithm '' ) and this permits to obtain the result which is independent of the parameter @xmath161 . it will be shown that the definition of the parameter @xmath161 minimizes corrections to eq.([2.10p ] ) practically for all values of the parameter @xmath161 . it should be emphasized also that here the coulomb corrections are included into the parameter @xmath229 in contrast to @xcite . substituting the expansion ( [ 2.26 ] ) in eq.([2.2p ] ) we obtain the decomposition of the probability of the pair creation : @xmath296 the probability of pair creation @xmath297 is defined by eq.([2.10p]).the term @xmath298 in eq.([2.12p ] ) corresponds to the first term ( linear in @xmath185 ) in eq.([2.26 ] ) . the explicit expression for the first correction to the pair creation probability is @xcite @xmath299 , \nonumber \\ & & f_1(z)=\left(\ln \varrho_c^2+\ln \frac{\nu}{i } -\ln \sinh z - c\right)g(z ) - 2\cosh z g(z ) , \nonumber \\ & & f_2(z ) = \frac{\nu}{\sinh z } \left(f_1(z)-\frac{g(z)}{2 } \right ) , \quad t = t_1+t_2 , \label{2.15p}\end{aligned}\ ] ] where the functions @xmath300 and @xmath301 are defined in eq.([2.31a ] ) , @xmath188 is defined in eq.([2.16 ] ) . use of the last representation of function @xmath301 simplifies the numerical calculation . as in radiation case the parameter @xmath229 has two representations : eqs.([2.32a ] ) and ( [ 2.30a ] ) . when the scattering of created particles is weak ( @xmath250 ) , the main contribution in eqs.([2.10p ] ) and ( [ 2.15p ] ) gives the region where @xmath251 and one can expand the functions of @xmath302 ( compare with eq.([2.33 ] ) ) . as result we obtain the spectral distribution of the pair creation probability for @xmath303 : @xmath304 } \nonumber \\ & & \displaystyle{=\frac{4z^2\alpha^3n_a}{3m^2\omega}\bigg\ { \left(\ln \left(183z^{-1/3 } \right ) -f(z\alpha)\right ) \left(1-\frac{31\nu_1 ^ 4}{21 } \right)-\frac{1}{6 } } \nonumber \\ & & \displaystyle{+2\frac{\varepsilon^2+\varepsilon'^2}{\omega^2 } \left[\left(\ln \left(183z^{-1/3 } \right)- f(z\alpha)\right)\left(1-\frac{16\nu_1 ^ 4}{21 } \right)+\frac{1}{12 } \right ] \bigg\ } } , \label{2.22p}\end{aligned}\ ] ] where @xmath190 is defined in eq.([2.16 ] ) . integrating eq.([2.22p ] ) over @xmath291 we obtain @xmath305 , \label{2.23p}\ ] ] where @xmath306 note that @xmath307 is four times larger than @xmath237 eq.([2.30b ] ) , in gold it is @xmath308tev . this is just the value of photon energy starting with the lpm effect becomes essential for the pair creation process in heavy elements . if one omits here the terms @xmath309 and @xmath310 these expressions coincide with the known bethe - maximon formula for the probability of pair creation by a high - energy photon in the case of complete screening ( if one neglects the contribution of atomic electrons ) written down within power accuracy ( omitted terms are of the order of powers of @xmath311 ) with the coulomb corrections , see e.g. eqs.(19.4 ) and ( 19.17 ) in @xcite . the pair creation spectral probability @xmath312 vs @xmath313 in gold is shown in fig . 5 for different energies . it is seen that for @xmath314tev which below @xmath307 the difference with the bethe - maximon probability is rather small . when @xmath315 there is significant difference with the bethe - maximon spectrum increasing with @xmath7 growth . in fig . 5 are shown the curves ( thin lines 2,3,4 ) obtained in logarithmic approximation @xmath316 eq.([2.10p ] ) , the first correction to the spectral probability @xmath317 eq.([2.15p ] ) , curves @xmath318 and the sum of these two contributions : curves @xmath319 . it should be noted that for our definition of the parameter @xmath161 eq.([2.14 ] ) the corrections are not exceed 6% of the main term . the corrections are maximal for @xmath320 . integrating eq.([2.10p ] ) over @xmath321 we obtain the total probability of pair creation in the logarithmic approximation @xmath322 , \label{2.23bp}\end{aligned}\ ] ] where @xmath323 @xmath324 is the bethe - maximon probability of pair photoproduction in the logarithmic approximation . the total probability of pair creation @xmath325 in gold is given in fig . 4 ( curve 2),it reduced by 10% at @xmath326 tev and it cuts in half at @xmath327 tev . at @xmath256 the function @xmath328 ( see eq.([2.15p ] ) ) has the form @xmath329 under the same conditions ( @xmath256 ) the function @xmath330 eq.([2.10p ] ) is @xmath331 so , in the region where the lpm effect is strong the probability eq.([2.10p ] ) can written as @xmath332 this means that in this limit the spectral probability of pair creation is proportional to the square root of the density just as the emission probability eq.([2.39a ] ) . at @xmath256 the relative contribution of the first correction @xmath298 is defined by @xmath333 in this expression the value @xmath264 with the accuracy up to terms @xmath265 does nt depend on the electron energy : @xmath334 . hence we can find the correction to the total probability at @xmath335 . the maximal value of the correction is attained at @xmath336 , it is @xmath269 for heavy elements . in this region @xmath256 or @xmath335 the total probability of pair creation in terms of the bethe - maximon total probability can be presented as @xmath337 . \label{2.33p}\ ] ] the contributions of higher orders of the perturbation theory over the interaction with an electromagnetic field give the electromagnetic radiative corrections to the electron mass and lead to appearance of the anomalous magnetic moment ( amm ) of the electron @xcite . it is known that under influence of an external electromagnetic field these effects in particular the amm of electron are changed essentially @xcite , @xcite . here we consider how the mentioned effects modify in a medium @xcite . the total probability of radiation @xmath338 is connected with imaginary part of the radiative correction to the electron mass according to @xmath339 since @xmath340 in eq.([2.11 ] ) is the analytic function of the potential @xmath341 we have that @xmath342 one can derive this formula considering the self - energy diagram and the corresponding amplitude of forward scattering of electron ( see eqs.(12.18 ) and ( 12.39 ) in @xcite ) . in the presence of a homogeneous external field the hamiltonian @xmath343 ( [ 2.9]),([2.10 ] ) acquires the linear over coordinate term . one can find the explicit form of this term using the eqs.([a17 ] ) and ( [ a21 ] ) of appendix a and carrying out the scale transformation of variables according to eq.([2.7 ] ) : @xmath344 where @xmath345 is the known parameter characterizing quantum effects in the radiation process ( see eq.([14 ] ) ) , @xmath346 . for polarized initial electron ( see ( [ a3a ] ) in appendix a ) have for the addition to @xmath340 in eq.([2.11 ] ) depending on the spin vector @xmath347 : @xmath348 for the radiative correction to the electron mass we have respectively @xmath349 it should be noted that expressions for @xmath340 ( [ 2.11 ] ) and ( [ 2.16a ] ) have universal form while the specific of the particle motion is contained in the propagator @xmath184 through the effective potential @xmath350 where we restrict ourselves to the main term in the decomposition over @xmath185 , see eqs.([2.16 ] ) , ( [ 2.26 ] ) . this means that result has the logarithmic accuracy over the scattering ( but not over an external field ) . with regard for an external field the parameter @xmath161 in eq.([2.32a ] ) is defined by a set of equations : @xmath351=1 \quad { \rm for } \quad 4\frac{\chi^2}{u^2}+ 4ql_1 \geq 1 , \label{2.18a}\end{aligned}\ ] ] where @xmath195 , @xmath352 , @xmath190 , @xmath353 and @xmath188 are defined in eqs.([2.30a ] ) and ( [ 2.32a ] ) . substituting into the depending on the spin term in ( [ 2.16a ] ) we find @xmath354 . \label{2.19a}\end{aligned}\ ] ] thus , in the used approximation ( @xmath355 ) we have for @xmath356 eq.([2.16a ] ) @xmath357 where @xmath358 . within relativistic accuracy ( up to terms of higher order over @xmath359 ) the combination of vectors entering in eq.([2.20a ] ) can be written as ( see eq.(12.19 ) in @xcite and @xcite ) @xmath360 where @xmath361 is the magnetic field in the electron rest frame . here we use that @xmath362 , where @xmath363 is the unit vector in the direction of acceleration ( this vector is used in @xcite ) . in the electron rest frame one can consider the value re @xmath364 depending on the electron spin as the energy of interaction of the amm of electron with the magnetic field @xmath365 ( see eq.(12.23 ) in @xcite and @xcite ) @xmath366 taking into account eqs.([2.4a ] ) , ( [ 2.16a ] ) , ( [ 2.16aa ] ) , ( [ 2.20a ] ) , ( [ 3.1a ] ) , ( [ 3.2a ] ) we obtain the following general expression for the amm of electron moving in a medium in the presence of an external electromagnetic field @xcite : @xmath367 in the absence of scattering ( @xmath368 ) the expression ( [ 3.3a ] ) gets over into the formula for the amm of electron in external field ( see eq.(12.24 ) in @xcite and @xcite , @xcite ) @xmath369dt \label{3.4a}\ ] ] in the weak external field ( @xmath370 ) we obtain the formula for the amm of electron under influence of multiple scattering @xmath371 where @xmath372 , \nonumber \\ & & \beta=\frac{1}{2}\left[\psi\left(\frac{1+x}{2 } \right)- \psi\left(\frac{x}{2 } \right ) \right ] , \label{3.6a}\end{aligned}\ ] ] where @xmath198 is defined in eq.([2.9b ] ) . the dependence of the amm of electron on its energy @xmath291 in gold is shown in fig . it is seen that at energy @xmath373 gev value of amm is 0.85 part of standard quantity ( sq ) of amm ( @xmath374 ) , at energy @xmath375 tev it is 0.77 part of sq and at energy @xmath376 tev it is 0.5 part of sq . actually in all heavy elements the behavior of amm of electron will be quite similar , i.e. the scale of energy where the amm of electron deviates from sq is of order of tev . in the case of weak effect of multiple scattering when @xmath240 , where @xmath237 is defined in eq.([2.30b ] ) we obtain for amm of electron @xmath377 at very high energy @xmath267 the effect of multiple scattering becomes strong . in this case the asymptotic expression for amm of electron is @xmath378 redefinition of the characteristic energy @xmath379 is connected with enlargement of the radiation cone ( in comparison with @xmath359 ) or in another terms with increasing of the characteristic transverse momentum transfers due to the multiple scattering . let us discuss one of possibilities of experimental observation of influence of medium on the amm of electron . at a very high energy where the observation becomes feasible the rotation angle @xmath380 of the spin vector @xmath347 in the transverse to the particle velocity @xmath100 magnetic field @xmath381 depends only on the value of amm and does nt depend on the particle energy : @xmath382 where @xmath383 is the path of electron in the field , @xmath384 oe ( see eq.([15 ] ) ) . the dependence of @xmath264 on energy @xmath291 is found above and shown in fig.6 . since at radiation of hard photons in a medium the picture is quite complicated : energy losses , cascade processes , spin flip and depolarization , it is desirable to measure the particles which do nt radiate photons on the path @xmath383 . the number of such particles @xmath385 is determined by the total probability of radiation in a medium found in @xcite , eqs.(3.12)-(3.14 ) ( see eqs.([2.4i ] ) and ( [ 2.5i ] ) in sec.2.2 above ) : @xmath386 where @xmath387 is the rotation angle for standard value of amm in qed ( @xmath374 ) , @xmath388 is the number of initial electrons . with energy increase the function @xmath389 decreases ( see eq.(43 ) in @xcite ) @xmath390 the crucial part of the experiment is an accuracy of measurement of electron polarization before target and after target . if one supposes that spin rotation angle can be measured with accuracy noticeably better than 1/10 then we can put that @xmath391 . in the gold we find for the energy @xmath392tev and the magnetic field @xmath393oe that @xmath394 , the path of electron in the target is @xmath395 cm and number of electron traversing the target without energy loss is @xmath396 . this estimates show that the measurement of the effect found in this paper will be feasible in the not very distant future . as known , the propagation of electromagnetic wave in a medium is defined by its dielectric tensor @xmath397 . for relatively low frequency @xmath7 ( e.g. visible light ) the dielectric tensor is defined by atomic phenomena . when the frequency of wave is much higher the atomic frequencies , the dielectric tensor has a form @xmath398 where @xmath15 is the plasma frequency , in any medium @xmath399 ev . so that for @xmath400 an influence of atomic phenomena on propagation of electromagnetic wave in a medium becomes small . at very high energy the nonlinear effects of qed enter into game . one of them is the polarization of the vacuum by a photon . in the presence of an external electromagnetic field the polarization of vacuum was considered first in the pioneer papers @xcite . in the strong field this effect can be essential for propagation of high - energy photons @xcite , @xcite . to evaluate the polarization tensor one has to consider the amplitude of photon scattering which included the polarization operator @xcite . as above we use the quasiclassical operator method @xcite , @xcite , @xcite . in this method the mentioned amplitude is described by diagram where the virtual electron - positron pair is first created by the initial photon with 4-momentum @xmath401 and polarization @xmath402 and then annihilate into final photon with 4-momentum @xmath403 and polarization @xmath404 . this corresponds to use of the non - covariant perturbation theory where at high energies ( @xmath405 ) the contribution of this diagram survives only . for this energy of photon this process occurs in a rather long time ( or at a rather long distance ) known as the lifetime of the virtual state @xmath406 where @xmath65 is the characteristic transverse momentum of the process , the system @xmath4 is used . when the virtual electron ( or positron ) is moving in a medium it scatters by atoms and changes the velocity under influence of external electromagnetic field . the mean square of momentum transfer to the electron from a medium and an external field on the distance @xmath66 is @xmath407 where @xmath45 , @xmath68 is the charge of nucleus , @xmath69 is the number density of atoms in the medium , @xmath72 is the screening radius of atom , @xmath408 is the electric field strength transverse to the velocity of particle @xmath409 , @xmath381 is the magnetic field strength ( see eq.([14 ] ) ) . in the case of small momentum transfer @xmath410 the influence of a medium and an external field is weak , in this case @xmath411 . at high energy it is possible that @xmath412 . in this case the characteristic value of the momentum transfer ( giving the main contribution into the spectral probability ) is defined by the value of @xmath413 . the self - consistency condition is @xmath414 here @xmath415 is known parameter characterizing the pair production process in a homogeneous external field @xmath416 . with @xmath81 increase the lifetime of the virtual state eq.([1pr.1 ] ) decreases . we will use the following normalization condition for the amplitude under consideration @xmath417 the amplitude @xmath60 is the contraction of the tensor @xmath418 ( @xmath419 and @xmath420 are the polarization vectors of the initial and the final photons ) and the polarization tensor @xmath421 . we select the basic vectors as @xmath422 since the tensor @xmath421 is invariant under the space inversion then in the selected basic vectors it has the diagonal form @xmath423 \label{1pr.8}\ ] ] in absence of external field it is convenient to describe the process of photon scattering using the helicity polarization vector @xmath424 connected with momentum transfer @xmath425 ( see eq.([3.32sc ] ) in the next subsection ) . in presence of external field and for @xmath426 we choose the polarization vectors in the following way : @xmath427 in terms of helicity amplitudes @xmath428 and @xmath429 the tensor @xmath421 and the corresponding dielectric tensor @xmath430 has a form @xmath431 the polarization tensor @xmath432 is diagonal in the basic vectors @xmath402 and @xmath404 eq.([1pr.7 ] ) . the corresponding mass squared are @xmath433 the probability of pair creation by a photon with polarization @xmath434 is @xmath435= -\frac{1}{\omega}{\rm im}\left[k_{++}^2+\xi_3k_{+-}^2\right ] , \label{2pr.15}\ ] ] where @xmath436 is the stokes parameter . for unpolarized photon one has @xmath437 the method of calculation of the functions @xmath438 and @xmath439 which define the tensor @xmath421 is similar to used in previous subsection ( see details in @xcite ) . the general expressions for photon masses squared under simultaneous influence of multiple scattering in a medium and an external electromagnetic field has a form @xmath440dt . \label{2pr.18}\end{aligned}\ ] ] for @xmath439 we found respectively @xmath441 here @xmath442 , \label{2pr.18a}\end{aligned}\ ] ] where @xmath443 is defined in eq.([1pr.3 ] ) , @xmath353 is defined in eq.([2.9 ] ) , the parameter @xmath161 is defined by the set of equations ( compare with eq.([2.18a ] ) ) : @xmath444 in the absence of external field ( @xmath445 ) we have @xmath446dt \nonumber \\ & & = -\frac{\alpha m^2}{2\pi}\int_{0}^{\omega } \frac{\omega d\varepsilon}{\varepsilon\varepsilon'}\left[s_1\left ( \ln p-\psi\left(p+\frac{1}{2 } \right ) \right ) + s_2 \left(\psi(p)-\ln p + \frac{1}{2p } \right)\right ] , \nonumber \\ & & k_{+-}^2=0 , \label{2pr.20a}\end{aligned}\ ] ] where @xmath447 as above is the logarithmic derivative of the gamma function . subsisting the result obtained into formula ( [ 2.16 ] ) we have the probability of pair creation which agrees with eq.([2.10p ] ) . in the absence of multiple scattering @xmath448 we get @xmath449dt , \nonumber \\ & & k_{+-}^2= -\frac{\alpha m^2}{2\pi}\int_{0}^{\omega } \frac{\omega d\varepsilon}{\varepsilon\varepsilon'}s_3 \kappa'^2\int_{0}^{\infty}e^{-it}\varphi_f tdt , \label{2pr.20}\end{aligned}\ ] ] where @xmath450 for this case the expressions for @xmath451 ( [ 2.14 ] ) after substitution results of eq.([2pr.20 ] ) agree with masses squared of photon calculated in an external electromagnetic field ( see @xcite and references therein ) . in the case when the both characteristic parameters are small ( @xmath452 ) , the main terms of decomposition of the functions @xmath453 are the sum of independent corrections to the photon mass squared both on account of the multiple scattering and an external field @xmath454 , \nonumber \\ \hspace{-16 mm } & & k_{+-}^2= \frac{\alpha m^2}{\pi } \bigg [ i\frac{\sqrt{3}\pi}{16\sqrt{2 } } \frac{\omega}{\omega_f } \exp\left(-\frac{8\omega_f}{3\omega}\right)+\frac{1}{30 } \left(\frac{\omega}{\omega_f}\right)^2 \bigg ] ; \nonumber \\ \hspace{-16 mm } & & \omega \ll \omega_e,\quad \omega_f = m \frac{h_0}{\|{\bf f}\| } , \quad \omega \ll \omega_f , \label{2pr.22}\end{aligned}\ ] ] where @xmath307 is defined in eq.([2.23ap ] ) and @xmath455 in eq.([15 ] ) . the correction @xmath456 follows from the first term of decomposition of the type eq.([2.26 ] ) . let us remind that in gold the value @xmath307 is @xmath457 tev , this is the typical value for the heavy elements . in fig.7 the functions re @xmath438(curve 2 ) and i m @xmath438 ( curve 1 ) are given for the case when the influence of a medium is taken into account only ( eq . ( [ 2pr.20a ] ) ) . the both curves are normalized to the asymptotics given by eq . ( [ 2pr.22 ] ) in the limit @xmath458 . it should be noted that beginning with some photon energy @xmath459 the radiative correction to the value re @xmath460 in the absence of a field ( @xmath461 ) becomes larger than @xmath462 ( see eq.([1pr ] ) ) . the estimate of @xmath463 is following @xmath464 for gold one obtains @xmath465 gev . a propagation of high - energy photons in oriented single crystals is one of interesting applications of the result obtained in this subsection . in this case we have both the dense matter with strong effect of multiple scattering and high fields of crystal axes or planes . as we saw , the lpm effect is most pronounced in the heavy elements . the same is valid for the process under consideration . let the high - energy photon incident on crystal . the angle of incidence is small and such that the distance from axis @xmath150 ( or the distance from plane @xmath2 ) can be considered as a constant on the formation length of process ( the constant field approximation is applicable , see sections 12,15 in @xcite ) . for orientation of a crystal along an axis the ratio of density of atoms in the vicinity of axis @xmath466 to the mean density @xmath69 is @xmath467 where @xmath468 is the amplitude of thermal vibrations of atoms , @xmath469 is the mean distance between atoms which form the axis . this ratio is maximal at @xmath470 . for numerical estimates we use for definiteness the tungsten single crystal . for the axis @xmath471 in w the ratio @xmath472=370 at the room temperature ( @xmath473 ) and @xmath472=1020 at @xmath474 . the effect of multiple scattering becomes strong at characteristic photon energy @xmath475tev and this value is inversely proportional to the density . so we have that @xmath476gev at @xmath473 and @xmath477gev at @xmath474 . it should be noted that within logarithmic accuracy we neglect by relatively small variation of @xmath190 due to substitution the screening radius @xmath478 by the value @xmath479 . it is useful to compare these estimates with known threshold energies @xmath480 at which the probability of pair creation in the field of axis is equal to the probability of the bethe - heitler mechanism , see table 12.1 in @xcite . for photon energy @xmath481 the process of pair creation in the field of axis dominates . in w crystal for @xmath471 axis @xmath482gev at t=293 k and @xmath483gev at t=77 k. it is seen that for these energies the ratio @xmath484 which characterize the strength the lpm effect is of the order of unity . at @xmath485 the maximal value of the parameter @xmath486 which determines the probability of pair creation by a photon in a field is also of the order of unity ( at @xmath487 , see the mentioned table ) . so we reach the conclusion that at some energy ( for axial orientation of crystal ) all the discussed effects are essential simultaneously . the analysis in this situation will be published elsewhere . for example , to calculate the influence of the field of axis on the polarization tensor one have to average the general formula ( [ 2pr.18 ] ) over all values of @xmath150 ( this is integration over @xmath488 with the weight @xmath489 , where @xmath490 is the density of axis in the perpendicular to them plane ) . investigation of the coherent photon scattering ( the delbrck scattering ) has rather long history which can be found in review ( see e.g.@xcite ) . there is a special interest to study the process for heavy elements ( the cross section @xmath491 ) . however , in this case the contributions of higher orders of @xmath144 into the amplitude of photon scattering are very important . this means that one needs the theory which is exact with respect to the parameter @xmath144 . the amplitudes of the coherent photon scattering valid for any @xmath144 for high photon energy @xmath63 and small scattering angle ( or small momentum transfer @xmath70 ) were calculated by several group of authors ( see @xcite , @xcite , @xcite , @xcite , @xcite ) . in the last paper the process of the coherent photon scattering was considered in frame of the quasiclassical operator method which appears to be very adequate for consideration of this problem . in the quasiclassical approximation the amplitude @xmath60 of the coherent photon scattering is described by diagram where the electron - positron pair is created by the initial photon with 4-momentum @xmath492 and then annihilate into the final photon with 4-momentum @xmath493 , so that the photon momentum transfer is @xmath494 . for high energy photon @xmath63 this process occurs over a rather long distance ( see eq.([1.1 ] ) ) . it is convenient to describe the process of photon scattering in terms of helicity amplitudes . we choose the polarization vectors with helicity @xmath495 @xmath496 there are two independent helicity amplitudes : @xmath497 where the first subscript is the helicity of the initial photon and the second is the helicity of the final photon . when the initial photons are unpolarized the differential cross section of scattering summed over final photons polarization contains the combination @xmath498 . \label{3.34sc}\ ] ] the calculation of the amplitudes of photon scattering on the separate atom appears to be most simple in the case when the screening radius of atom @xmath173 is much smaller than the formation length of process ( @xmath499 eq.([10a ] ) ) . the imaginary part of helicity amplitudes in this case can be written in the form @xcite @xmath500 where @xmath501 where the function @xmath502 is defined in eq.([2.9b ] ) . the important property of eq.([3.46sc ] ) is that the dependence on the screening radius @xmath72 originates in it from the born approximation . in this approximation in the case of arbitrary screening the radius @xmath173 enters only in the combination @xmath503 because of this we can extend eq.([3.47sc ] ) on the case of arbitrary screening making the substitution @xmath504 in the case @xmath505 ( the screening radius is very large , or in other words we consider the photon scattering in the coulomb field ) we have to substitute in eq.([3.47sc ] ) @xmath506 in the case of the complete screening ( @xmath507 ) the functions @xmath508 are independent of @xmath2 and @xmath509 and the corresponding integrals are @xmath510 for the scattering amplitudes we have @xmath511 the real part of amplitudes are calculated using dispersion relations . the photon scattering amplitude in this case for arbitrary value of parameter @xmath512 for @xmath75 was found recently in @xcite and for arbitrary @xmath70 in @xcite . when a photon is propagating in a medium it also dissociates with probability @xmath513 into an electron - positron pair . the virtual electron and positron interact with a medium and can scatter on atoms . in this scattering the electron and positron interaction with the coulomb field in the course of the coherent scattering of photon is involved also . there is a direct analogue with the lpm effect : the influence of the multiple scattering on process of the bremsstrahlung and pair creation by a photon in a medium at high energy . however there is the difference : in the lpm effect the particles of electron - positron pair created by a photon are on the mass shell while in the process of the coherent scattering of photon this particles are off the mass shell , but in the high energy region ( this is the only region where the influence of the multiple scattering is pronounced ) the shift from the mass shell is relatively small . to include this scattering into consideration the amplitude of the coherent scattering of photon should be averaged over all possible trajectories of electron and positron . this operation can be performed with the aid of the distribution function averaged over the atomic positions of scatterer in the medium in the same manner as it is done in appendix a @xcite . the photon scattering amplitudes can be written in the form @xmath514 where @xmath515 is the main ( logarithmic ) term and @xmath516 is the first correction to the scattering amplitude . the main terms are @xcite @xmath517 where @xmath518 is defined in eq.([2.16 ] ) . the structure of the amplitude @xmath515 is quite similar to the structure of the pair creation probability eq.([2.10p ] ) under influence of the multiple scattering ( the fact that the amplitude is expressed in terms of the probability is the common feature of the coherent photon scattering description ) . this result is not occasional and connected with origin of coherent scattering of photon as a process which is going through dissociation of a photon into electron - positron pair . however the expression for the basic parameter @xmath519 differs from corresponding expression in eq.([2.10p ] ) . the first corrections to the amplitudes are defined by @xmath520,\quad f_1(v)=1-\frac{\ln ( v+\sqrt{1+v^2})}{v\sqrt{1+v^2 } } , \nonumber \\ & & f_3(\nu_s)=i \nu_s s_3\int_{0}^{\infty}\frac{dz e^{-it}}{\sinh^3z } g(z ) = \frac{s_3}{2}\left[1+\frac{1}{2p}-p~\zeta(2 , p ) \right ] \label{4.18sc}\end{aligned}\ ] ] here the functions @xmath521 , @xmath522 and @xmath300 are defined in eq.([2.15p ] ) , @xmath523 is the generalized riemann zeta function . use of the given representations of functions @xmath524 and @xmath301 in @xmath521 simplifies the numerical calculation . the functions @xmath525 and @xmath526 in eq.([4.17sc ] ) encountered in the radiation theory . in the region of the weak effect of scattering ( @xmath527 ) we obtain @xmath528 , \nonumber \\ \hspace{-10 mm } & & m_{+-}=m_{+-}^{(1)}=i\frac{2z^2 \alpha^3 \omega}{9m^2 } f_1\left(\frac{\beta}{2}\right ) \left(1+i\frac{16\omega}{25 \omega_e}\frac{l_{s1}}{l_1 } - \frac{384}{245}\left(\frac{\omega}{\omega_e}\frac{l_{s1}}{l_1 } \right)^2 \right ) , \label{4.20sc}\end{aligned}\ ] ] here @xmath529 , \label{4.21sc}\ ] ] @xmath307 is defined in eq.([2.23ap ] ) and @xmath190 is defined in eq.([2.16 ] ) . the characteristic energy @xmath307 encountered in analysis of influence of the multiple scattering on the probability of pair photoproduction @xcite , in gold @xmath307=10.5 tev . the amplitudes eq.([4.20sc ] ) coincide with the formulas in absence of multiple scattering if we neglect the terms @xmath530 and @xmath531 . the terms @xmath530 define the appearing real part of the scattering amplitudes while the terms @xmath532 are the corrections to the imaginary part . in the region where the effect of scattering is strong ( @xmath533 ) we obtain @xmath534 , \nonumber \\ & & m_{+- } \simeq ( i-1)\frac{\pi z^2 \alpha^3 \omega}{8\sqrt{2}\delta_s^2 } \frac{1}{\sqrt{l_{s3}}}f_1\left(\frac{\beta}{2}\right),\quad l_{s3}=l_{s2}+\frac{1}{2 } \ln l_{s2 } , \nonumber \\ & & l_{s2}=l_{s1}+\frac{1}{2}\ln \frac{\omega}{\omega_e}= 2\ln(a_{s}\delta_s)+1-f_2\left(\frac{\delta a_{s}}{2}\right)-2 f(z\alpha ) , \nonumber \\ & & \delta_s^4=2\pi z^2 \alpha^2 \omega n_a,\quad \nu_s^2=i\frac{\omega}{\omega_e}\frac{4\varepsilon \varepsilon'}{\omega^2 } \left(l_{s3}+ \frac{1}{2 } \ln \frac{4\varepsilon \varepsilon ' } { \omega^2}\right ) . \label{4.24sc}\end{aligned}\ ] ] the real and imaginary parts of the amplitudes are equal ( if we neglect the term @xmath535 in @xmath428 ) . moreover the amplitudes eq.([4.24sc ] ) do nt depend on the electron mass @xmath152 . in place of it we have the value @xmath536 . the influence of a medium on the process of the coherent photon scattering illustrated in fig . 8 a and b , where i m @xmath428 and re @xmath428 as well as i m @xmath429 and re @xmath429 are given as a function of photon energy @xmath7 in gold . this influence is due to the multiple scattering of electron and positron of the virtual pair on the formation length of the process ( see eqs.([1.1 ] ) and ( [ 1.2 ] ) ) @xmath537 in the region @xmath538 this formation length is independent of @xmath70 and it value is coincide practically with the formation length of pair creation by a photon @xmath539 considered in previous subsection . there is some difference connected with the logarithmic dependence of @xmath540 value on @xmath541 : @xmath542 where @xmath543 and @xmath544 , @xmath72 is the screening radius of atom ( [ 1.2 ] ) . this defines the weak ( logarithmic ) dependence of im@xmath428 on @xmath70 in the region @xmath545 . this can be seen in fig . 8a where the curves 1 and 3 represent behavior of i m @xmath428 for @xmath70=0.4435 @xmath152 and @xmath70=0.0387 @xmath152 respectively . for lower value of @xmath70 the minimal momentum transfer @xmath546 ( [ 1.2 ] ) diminishes thereby the interval of contributing the multiple scattering angles increases , so the multiple scattering affects the photon scattering amplitude at a lower energy ( and smaller formation length ) . because of this the curve 3 is shifted to the left respect the curve 1 . note that the curve 3 ( @xmath547 ) is very similar to the curve 2 in fig . 4 which represents the behavior of the probability of pair photoproduction in gold vs photon energy . the new property of influence of a medium is the appearance of the real part of the coherent photon scattering amplitudes at high energy @xmath7 . in the region @xmath548 the value of re@xmath60 is small accordingly eq.([4.20sc ] ) . in the asymptotic region @xmath335 we have -re @xmath549 i m @xmath60 according to eq.([4.24sc ] ) . this property is seen clearly in figs.8 a and b. so the value of -re @xmath60 is small at low and very high energies of photon . at intermediate energies the value of -re @xmath60 have the maximum at @xmath550220 tev for @xmath5510.4435 @xmath152 and at @xmath55080 tev for @xmath5510.0387 @xmath152 . in fig . 8b the same curves are shown for amplitude @xmath429 . these curves are very similar to curves in fig . 8a . the curves in both figures are normalized to imaginary part of the corresponding amplitude in the absence of the multiple scattering . the ratio these imaginary parts @xmath264 is very small : it is @xmath552 for @xmath70=0.4435 @xmath152 and @xmath553 for @xmath70=0.0387 @xmath152 . when one considers bremsstrahlung of enough soft photons @xmath554 ( see eq.([4 ] ) ) , one has to take into account the effect of a polarization of the medium . this effect diminishes the formation length ( see eq.([5 ] ) ) as well as the probability of radiation ( see @xcite , the qualitative discussion may be found in @xcite ) . for analysis we use the general expression for the probability of radiation , see eq.([a1 ] ) , appendix a. the factor in front of exponent in this expression ( see eq.([a2 ] ) ) contains two terms @xmath555 and @xmath556 , the term @xmath555 is not changed and the term @xmath556 contains combination @xmath557 and its dependence on @xmath558 ( term of the order @xmath140 ) may be neglected also . so , the dependence on @xmath15 manifests itself in the exponent of eq.([a1 ] ) , and respectively in the exponent of ( [ a19 ] ) only : @xmath559 where the term with @xmath560 arises in the case of virtual photon : @xmath561 . performing the substitution @xmath562 in eq.([2.7 ] ) we obtain for the potential ( see eqs.([2.7 ] ) ( [ 2.9 ] ) , ( [ 2.9a ] ) , ( [ 2.16 ] ) ) @xmath563 the substitution eq.([43 ] ) in the expression for the probability of radiation eq.([2.8 ] ) gives @xmath564 the value of the parameter @xmath565 in ( [ 43a ] ) is determined by equation ( compare with eqs.([2.14 ] ) , ( [ 2.30a ] ) ) @xmath566 in the opposite case @xmath567 and this is possible in two intervals of the photon energy @xmath7 : 1 . for @xmath568 when the multiple scattering and effects of the polarization of a medium are weak ; 2 . for @xmath569 when effects of the polarization of a medium become stronger then effects of the multiple scattering ( @xmath570 ) . in an intermediate region we substitute @xmath571 in eq.([44a ] ) . after it we obtain the equation for @xmath161 which coincides with eq.([2.14 ] ) , see also eq.([2.30a ] ) : @xmath572 thus , for @xmath573 we have @xmath574 while for @xmath575 we have @xmath576 the spectral distribution of the probability of radiation eq.([2.35 ] ) with allowance for the polarization of a medium have the form @xmath577 } , \label{45}\ ] ] where @xmath578 in the case @xmath579 , the characteristic momentum transfer in the used units ( @xmath580 ) are defined by value @xmath581 , one can use asymptotic expansion eq.([2.33 ] ) and we have for the spectral distribution of the probability of radiation @xmath582 where @xmath178 is defined in ( [ 2.9b ] ) , @xmath583 . the results obtained agree with given in @xcite where calculations are fulfilled within a logarithmic accuracy and without coulomb corrections . it is seen that the dependence of spectral distribution on photon energy ( @xmath584 ) differs essentially from the bethe - maximon one ( @xmath585 ) , and the probability is independent on the density @xmath586 . in the case under consideration probability of radiation is defined not only by the relative time @xmath587 as in sec . the used radiation theory is formulated in terms of two times ( see eqs.([a1 ] ) - ( [ a3 ] ) of appendix a ) . proceeding from this formulation we can obtain more general expression for the radiation probability . with allowance for the polarization of a medium we have for the spectral distribution of the probability of radiation ( compare with eq.([2.8 ] ) ) @xmath588 } , \label{50}\end{aligned}\ ] ] where @xmath589 here the function @xmath590 describes the density of atoms on the projectile trajectory , @xmath558 is defined in eq.([42 ] ) . the functions @xmath591 satisfy the equation ( [ 2.9 ] ) , but now the potential @xmath341 depends on time @xmath592 so , the functions @xmath593 and @xmath143 in eqs.([51 ] ) and ( [ 52 ] ) depend now on mean density of atoms on the projectile trajectory while the function @xmath590 describes the modulation of the density as a function of time @xmath594 using an operator form of a solution of eq.([52 ] ) ( compare with eq.([2.10 ] ) ) we can present the probability eq.([50 ] ) in the form @xmath595 , \label{53}\end{aligned}\ ] ] where the symbol @xmath596 means the chronological product . note , that in eqs.([50 ] ) and ( [ 53 ] ) it is implied that subtraction is made at @xmath142 , @xmath597 . for the homogeneous target of finite thickness @xmath383 the radiation process in a medium depends on interrelation between @xmath383 and formation length @xmath598 eq.([5 ] ) . in the case when @xmath599 we have the thick target where radiation on the boundary should be incorporated . in the case when @xmath600 we have the thin target where the mechanism of radiation is changed essentially and in the case when @xmath601 we have intermediate thickness . integrals over time in eq.([53 ] ) we present as integrals over four domains : 1 . @xmath602 ; 2 . @xmath603 ; 3 . @xmath604 ; 4 . @xmath605 ; where @xmath606 in two more domains @xmath607 and @xmath608 an electron is moving entirely free and there is no radiation . we consider in this subsection the case , when the thickness of a target @xmath383 is much larger than formation length @xmath66 eq.([5 ] ) or @xmath609 . in this case domain 4 . does nt contribute . the contributions of other domains are @xmath610 where @xmath611 . the term in @xmath612 : @xmath613 describes the probability of radiation considered in previous subsections . all other terms define the probability of radiation of boundary photons . so , making mentioned subtraction we have for the spectral distribution of the probability of radiation of boundary photons @xmath614 in the case when both the lpm effect and effect of polarization of a medium are weak one can decompose combination in @xmath60 in eq.([55 ] ) @xmath615 using this decomposition one can find an estimate for probability of radiation @xmath616 where @xmath617 and @xmath618 are some positive coefficients . so , in the case under consideration the probability of boundary photon radiation is negligible small . in the case when @xmath619 one can omit the potential @xmath341 in eq.([51b ] ) , so that the effect of polarization of a medium is essential , then @xmath620\right\ } \label{51d}\end{aligned}\ ] ] this result is the quantum generalization of the transition radiation probability @xcite . let us estimate the probability of radiation eq.([55 ] ) in the case when both the lpm effect and effect of polarization of a medium are strong @xmath621 . in this case we can neglect the term @xmath622 in the term @xmath623 in @xmath60 eq.([55 ] ) . in the term @xmath624 one can put @xmath625 so that @xmath626 substituting these results into eq.([55 ] ) we have up to logarithmic accuracy @xmath627 . \label{51g}\ ] ] in the case @xmath628 this result agrees with eq.([51d ] ) . the case @xmath629 will be considered below ( see eqs.([z24 ] ) and ( [ z25 ] ) ) . we consider now a situation when the formation length of radiation is much larger than the thickness @xmath383 of a target @xcite @xmath630 where @xmath631 are defined in eqs.([5 ] ) , ( [ 6 ] ) . in the case @xmath632 the radiated photon is propagating in the vacuum and one can neglect the polarization of a medium . the spectral distribution of the probability of radiation from a thin target is @xmath633 \left(1-\exp ( -v(\varrho)t ) \right ) } , \label{72}\ ] ] where @xmath634 is defined in eqs.([2.9 ] ) , ( [ 2.9a ] ) , @xmath635 is the modified bessel function . if @xmath636 one can expand the exponent ( the contribution of the region @xmath637 is exponentially damped because in this region @xmath638 ) . in the first order over @xmath639 using the explicit expression for the potential eq.([2.9a ] ) one obtains in this case the bethe - maximon formula with the coulomb corrections eq.([2.35 ] ) . in the opposite case when the multiple scattering of a particle traversing a target is strong ( @xmath640 , the mean square of multiple scattering angle @xmath641 ) . we present the function @xmath642 ( see eqs.([2.9 ] ) , ( [ 2.9a ] ) and ( [ 2.18 ] ) as @xmath643 where @xmath644 is the lower boundary of values contributing into the integral over @xmath645 . substituting this expression into ( [ 72 ] ) we have the integral @xmath646 \right\ } \equiv \pi j}. \label{74}\ ] ] in this integral we expand the exponent in the integrand over @xmath647 keeping the first term of the expansion . it is convenient to substitute @xmath648 . we find @xmath649\exp(-z ) zdz } , \nonumber \\ & & \displaystyle{j_2=-\frac{1}{kl_t}\int_{0}^{\infty } k_1 ^ 2\left(\sqrt{\frac{z}{k}}\right ) \exp(-z ) \ln z zdz},\quad j = j_1+j_2 . \label{76}\end{aligned}\ ] ] expanding the modified bessel functions @xmath650 at @xmath651 and taking the integrals in the last expression we have @xmath652 in the term with @xmath653 in eq.([72 ] ) the region @xmath654 contributes . so we have @xmath655 substituting found @xmath656 and @xmath657 into eq.([72 ] ) we obtain for the spectral distribution of the probability of radiation in a thin target at conditions of the strong multiple scattering @xmath658 the logarithmic term in this formula is well known in theory of the collinear photons radiation at scattering of a radiating particle on angle much larger than characteristic angles of radiation @xmath659 . it is described with logarithmic accuracy in a quasi - real electron approximation ( see @xcite , appendix b2 ) . the formula ( [ 72 ] ) presents the probability of radiation in the case when the formation length @xmath660 . it is known , see e.g. @xcite , that in this case a process of scattering of a particle is independent of a radiation process and a differential probability of radiation at scattering with the momentum transfer @xmath33 can be presented in the form @xmath661 where @xmath662 is the differential probability of scattering with the momentum transfer @xmath33 which depends on properties of a target . the function @xmath663 is the probability of radiation of a photon with a momentum @xmath9 when an emitting electron acquires the momentum transfer @xmath33 . this probability has a universal form which is independent of properties of a target . for an electron traversing an amorphous medium this fact is reflected in eq.([72 ] ) . indeed , passing on to a momentum space we have @xmath664 \left(1-\exp ( -i{\bf q}\mbox{\boldmath$\varrho$ } ) \right ) } \nonumber \\ & & \hspace{-16mm}\displaystyle { = \frac{\alpha d\omega}{\pi \omega}\left[r_1 f_1\left ( \frac{q}{2}\right)+r_2 f_2\left(\frac{q}{2}\right ) \right ] } , \label{81}\end{aligned}\ ] ] where the functions @xmath525 and @xmath526 are defined in eqs.([4.18sc ] ) and ( [ 4.17sc ] ) . remind that @xmath186 is measured in electron mass . the probability of radiation in this form was found in @xcite . for a differential probability of scattering ( here we consider the multiple scattering ) there is a known formula ( cp eqs.([a12 ] ) , ( [ a17 ] ) , ( [ 2.9 ] ) and ( [ 2.9a ] ) ) @xmath665 where @xmath666 is the cross section of single scattering . the formula ( [ 82 ] ) is the exact solution of the kinetic equation for the pure scattering problem ( without radiation ) . let us consider scattering of relativistic electrons in the case when scattering angles @xmath667 and the condition @xmath668 is fulfilled ( the angle @xmath177 is defined in appendix a , eq.([a12 ] ) ) . for combination @xmath669 in eq.([82 ] ) one can use eq.([73 ] ) . let us remind that the impact parameters @xmath645 measured in the compton wavelengths @xmath151 , which is conjugate to space of the transverse momentum transfers @xmath186 measured in the electron mass @xmath152 . @xmath670 substituting eq.([82a ] ) into eq.([82 ] ) and passing to the variable @xmath671 we obtain @xmath672 solution of scattering problem in angular variables is given in the end of appendix a ( see eq.([a23 ] ) ) . on adopted assumptions ( @xmath673 ) one can expands in eq.([82b ] ) over powers of @xmath647 : @xmath674 it is instructive to compare these results with the classical paper bethe on multiple scattering @xcite . the obtained formula eq.([82c ] ) is consistent with eq.(25 ) of @xcite , however there is difference in evaluation of the coulomb corrections entering into the expression for the function @xmath675 eq.([82a ] ) ( see @xmath676 in eq.(22 ) in @xcite ) . in the last formula in @xcite the coulomb correction is contained in the form @xmath677 ( it should be noted that a ) bethe was referring to the numerical coefficient at @xmath678 calculated by molire @xcite as `` only approximate '' ; b ) the coulomb corrections to the cross sections of bremsstrahlung and pair creation by photon which contain the famous function @xmath178 defined in eq.([2.9b ] ) were found after publication of @xcite ) while in eq.([82a ] ) enters different correction : @xmath679 . to compare these results numerically let us consider case @xmath680 . in this case the correction in @xcite is -@xmath681 while for eq.([82a ] ) we have the correction @xmath682 . besides , for accounting the inelastic scattering on electrons the use of eq.([40a ] ) for the potential @xmath683 gives more accurate results than the simple substitution @xmath684 done in @xcite . the main term of the expansion in eq.([82c ] ) has the gaussian form @xmath685 the next terms contains the corrections to the gaussian form . they were analyzed in detail in @xcite . using the formula ( [ 81 ] ) one can easily obtain to within logarithmic accuracy expressions ( [ 79]),([51 g ] ) . both a radiation of boundary photons and a radiation in a thin target may be considered as a radiation of collinear photons ( see e.g. @xcite ) in the case when an emitting particle deviates at large angle ( @xmath686 ) . using eq.([81 ] ) at @xmath687 we find @xmath688 ; \nonumber \\ & & \int_{}^{}d^2q dw_r(q ) f_s({\bf q } ) \simeq \frac{\alpha d\omega}{\pi \omega } \left[r_1+r_2\left(\ln \overline{q^2 } -1 \right ) \right ] . \label{83}\end{aligned}\ ] ] for a thin target value of @xmath689 is defined by mean square of multiple scattering angle on a thickness of a target @xmath383 , and for boundary photons is the same but on the formation length @xmath66 . however , if we one intends to perform computation beyond a logarithmic accuracy , the method given in this subsection has advantage since there is no necessity to calculate @xmath690 . radiation from a thin target using the path integral method was considered in @xcite and @xcite . in the case when a target has intermediate thickness ( @xmath691 ) the separation of contributions on photon emission inside target and boundary photon emission becomes senseless . we consider this case @xcite neglecting by the correction term @xmath180 . we present the spectral probability of radiation as @xmath692 ; \nonumber \\ & & \hspace{-6 mm } j_1^{(m ) } = j_3^{(m ) } = \int_{0}^{\infty } dt_1 \int_{0}^{t } dt_2 { \rm e}^{-i\left(t_1+\kappa t_2 \right)}\left[\left(t_1+t_2\right)^{-m } -n_1^m \right ] , \nonumber \\ & & \hspace{-6 mm } j_2^{(m ) } = \int_{0}^{t } dt ( t - t){\rm e}^{-i \kappa t } \left[t^{-m}-n_2^m \right ] , \nonumber \\ & & j_4^{(m ) } = \int_{0}^{\infty } dt_1 \int_{0}^{\infty } dt_2 { \rm e}^{-i\left(t_1+t_2+\kappa t \right)}\left[\left(t_1+t_2+t\right)^{-m } -n_4^m \right ] , \nonumber \\ & & \hspace{-6 mm } j_5^{(m)}(t)=2 \int_{0}^{\infty}dt_1\int_{0}^{t}dt_2 \frac{{\rm e}^{-it_1 } } { \left(t_1+t_2 \right)^m}\left({\rm e}^{-it_2}-{\rm e}^{-i\kappa t_2 } \right ) + \int_{0}^{t}dt \frac{(t - t)}{(t - i0)^{m } } \left({\rm e}^{-it}-{\rm e}^{-i\kappa t } \right ) \nonumber \\ & & \hspace{-6 mm } + \int_{0}^{\infty}dt_1\int_{0}^{\infty}dt_2 \frac{{\rm e}^{-i\left(t_1+t_2 \right ) } } { \left(t_1+t_2 + t \right)^m}\left({\rm e}^{-it}-{\rm e}^{-i\kappa t } \right ) . \label{z12}\end{aligned}\ ] ] where @xmath340 is defined in ( [ 51 ] ) , the functions @xmath693 are @xmath694 here @xmath695 and @xmath696 are defined in eq.([43 ] ) and in eq.([2.30b ] ) . we confine ourselves to the case @xmath240 when the lpm effect manifests itself for soft emitted photons ( @xmath26 ) . so we can neglect the terms with @xmath697 in ( [ z12 ] ) . let us consider first the case when lpm effect is weak ( @xmath698 ) . we assume here that condition @xmath699 ( definition of @xmath700 see in eq.([5 ] ) ) is fulfilled , that is in the region where @xmath250 one has @xmath701 and effects of the polarization of a medium are negligible . this is true for high energies ( @xmath702 ) . then for thickness @xmath703 the transverse shift of the projectile due to the multiple scattering in a target as a whole have no influence on coherent effects defined by the phase @xmath704 in the factor @xmath705 . indeed , for the projectile traversing a target in the case @xmath706 an increment of the phase @xmath707 is small @xmath708 the angle of multiple scattering @xmath709 is small also comparing with an characteristic angle of radiation @xmath710 . so , in the case @xmath711 the radiation originates on separate atoms of a target and an interference on target boundaries is defined by the value @xmath712 . at @xmath713 this interference is weak , while at @xmath714 there is a damping of the interference terms due to integration over photon emission angles . expanding over @xmath715 in eq.([z12 ] ) we obtain ( @xmath716 ) : @xmath717 \nonumber \\ & & = \frac{\nu_1 ^ 2t}{3 } \left[1 + 3t\left(\left(1-\frac{t^2}{2 } \right ) { \rm si}(t)-2 t { \rm ci}(t ) + \frac{3}{2 } \sin t - \frac{t}{2 } \cos t \right ) \right ] . \label{z22b}\end{aligned}\ ] ] for case @xmath718 @xmath719 , \label{z22c}\ ] ] and for case @xmath714 @xmath720 thus , in the case @xmath711 the probability of radiation is defined by bethe - maximon formula both for @xmath718 and for @xmath714 . however , for @xmath721 the interference on the target boundaries is essential . if we present as above @xmath722 as product of @xmath723 ( bethe - maximon formula ) and some interference factor , then this factor attains @xmath724 at @xmath725 ( minimum of the interference factor ) and @xmath726 at @xmath727 ( maximum of the interference factor ) . when the parameter @xmath728 is large ( @xmath729 ) the radiation is formed inside a target and the interference terms are damped exponentially . in this case formulae derived for thick target in sec.2.1 are applicable . taking into account the contribution of boundary photons ( see sec.3.2 ) we have @xmath730 where @xmath190 and @xmath715 are defined in eqs.([2.16 ] ) and ( [ 2.30a ] ) . we consider now the case when the lpm effect is strong ( @xmath256 ) and the parameter @xmath718 while the value which characterize the thickness of a target @xmath731 . we find for @xmath632 ( @xmath732 ) @xmath733+i\kappa t\left(\ln \frac{\cosh \nu t}{\tanh \nu t}- \frac{\nu t}{\tanh \nu t } \right ) \nonumber \\ & & + \frac{i\kappa}{\nu}\int_{0}^{\nu t } dt \left(\frac{4t}{\sinh 2t}- \frac{t^2}{\sinh^2 t } \right)\bigg\ } , \quad \nu = \exp\left(i\frac{\pi}{4}\right)\nu_0 . \label{z24}\end{aligned}\ ] ] for a relatively thick target ( @xmath734 ) we have from eq.([z24 ] ) @xmath735 here the terms without @xmath340 are the contribution of boundary photons ( see eq.(4.14 ) in @xcite ) while the term @xmath736 gives the probability of radiation inside target ( with correction @xmath737 but without corrections @xmath738 ) . the relative value of the last corrections at @xmath256 is given by eq.([2.40 ] ) . in the limiting case when a target is very thin and @xmath739 but when @xmath740 we have from eq.([z24 ] ) @xmath741 - 2 + \delta , \nonumber \\ & & \delta= \frac{(\nu_0 t)^4}{180}+\frac{2(\nu_0 t)^2}{45}t\left ( \ln \nu_0 ^ 2 t - c\right)-\frac{\kappa t}{6}(\nu_0 t)^2 . \label{z26}\end{aligned}\ ] ] the terms without @xmath742 in this expression coincide with eq.([79 ] ) ( up to terms @xmath743 ) . in the photon energy region where @xmath744 the contribution of the terms @xmath745 ( @xmath403=1,2,3 ) is very small ( @xmath746 ) and decreases with photon energy reduction ( @xmath747 ) , so that in the spectral distribution of radiation only the terms @xmath748 contribute . we obtain for the function @xmath749 in the case when @xmath750 and when the parameter @xmath751 , which characterizes the mean square angle of the multiple scattering in a target as a whole , has an arbitrary value @xmath752 \nonumber \\ & & \hspace{-4 mm } = 4 \int_{0}^{\infty}dx k_1 ^ 2\left(2\sqrt{x } \right ) \left[1-\exp \left(-x\nu_0 ^ 2 t \right ) \right ] = 2\int_{0}^{\infty}d\varrho \varrho k_1 ^ 2(\varrho)\left[1-\exp \left(-k\varrho^2 \right ) \right ] , \nonumber \\ & & \hspace{-4 mm } 4k=\nu_0 ^ 2 t , \label{z29}\end{aligned}\ ] ] where @xmath753 is the modified bessel function . formula ( [ z29 ] ) corresponds at @xmath716 to result for a thin target obtained above ( see eq.([72 ] ) ) without terms @xmath754 . since the dependence on the parameter @xmath695 is contained in eq.([z29 ] ) as a common phase multiplier @xmath755 , one can write more accurate expression for @xmath749 ( with terms @xmath754 ) using the found results ( see eq.([73 ] ) ) : @xmath756 . \label{z30}\ ] ] for the case @xmath757 it has the form @xmath758 - 2 + \frac{c}{l_t } , \nonumber \\ & & k= \frac{\pi z^2 \alpha^2 n l}{m^2}l_t , \quad l_t= \ln \frac{4a_{s2}^2}{\lambda_c^2 \varrho_t^2}-2c,\quad k(\varrho_t)\varrho_t^2=1 \label{ad31}\end{aligned}\ ] ] in the case when parameter @xmath403 is not very high one has to use an exact expression eq.([72 ] ) . for @xmath759 one can expand the exponent in the integrand of eq.([z30 ] ) . then we find @xmath760 at @xmath632 the spectral distribution of probability is @xmath761 this is the bethe - maximon formula for not very hard photons ( terms @xmath762 are omitted ) . when a photon energy decreases , the parameter @xmath695 increases as well as the combination @xmath763 , while the value @xmath764 decreases @xmath765 . just this value defines an accuracy of eq.([z30 ] ) . using eq.([z12 ] ) at @xmath766 we find for the probability of transition radiation ( @xmath767 ) the following expression @xmath768 + \kappa t { \rm si}(\kappa t ) -4 \sin^2 \frac{\kappa t}{2 } \bigg\}. \label{z34}\end{aligned}\ ] ] in the limiting case @xmath769 the probability eq.([z34 ] ) turns into standard probability of the transition radiation ( [ 51d ] ) with oscillating additions @xmath770 , \nonumber \\ & & j_{tr}=\left(1+\frac{2}{\kappa-1}\right)\ln \kappa-2 . \label{z34a}\end{aligned}\ ] ] note , that there is a qualitative difference in a behaviors of interference terms in eqs.([z22d ] ) and ( [ z34a ] ) . in the former an amplitude of oscillation with @xmath7 increase decreases as @xmath771 while in the latter the corresponding amplitude weakly ( logarithmically ) increases with @xmath7 decrease . from the above analysis follows that in the case when @xmath772 ( @xmath256 ) the spectral distribution of probability of radiation with the polarization of a medium taken into account has the form @xmath773 where @xmath774 is the spectral distribution of probability of radiation in a thin target without regard for the polarization of a medium . in the case @xmath775 the probability @xmath774 is defined by eqs.([77 ] ) , ( [ 79 ] ) and for the case @xmath759 it is defined by eq.([z22f ] ) . more accurate representation of the probability of radiation @xmath774 may be obtained using eq.([z30 ] ) . it follows from eqs.([z34a ] ) and ( [ z35 ] ) that if we make allowance for multiple scattering at @xmath776 this results in decreasing of oscillations of the transition radiation probability by magnitude of the bremsstrahlung probability in a thin target . the radiation of the boundary photons with regard for the multiple scattering was considered in @xcite ( for @xmath777 ) , the effect of polarization of a medium was included into consideration of the problem in @xcite and @xcite . in these papers the probability of radiation of boundary photons ( under condition of applicability of eqs.([z25 ] ) and ( [ z26 ] ) ) was analyzed also to within the logarithmic accuracy ( see eq.(20 ) in @xcite and eq.(15 ) in @xcite ) . this accuracy is insufficient for parameters connected with experiment @xcite-@xcite . for example , for @xmath28 gev and in heavy elements the value @xmath696 equates @xmath695 for @xmath778 . one can see from eqs.([z25 ] ) and ( [ z26 ] ) ) that in this case @xmath779 is nearly completely compensated by constant terms . our results , which are consistent with obtained @xcite , are more accurate and presented in more convenient for application form and the coulomb corrections are included . it should be noted that in the experiments @xcite-@xcite the summary energy of all photons radiated by a single electron is measured . this means that besides mentioned above effects there is an additional `` calorimetric '' effect due to the multiple photon radiation . this effect is especially important in relatively thick used targets . since the energy loss spectrum of an electron is actually measured , which is not coincide in this case with the spectrum of photons radiated in a single interaction , one have to consider the distribution function of electrons over energy after passage of a target @xcite . as it is known , this distribution function is the solution of the corresponding kinetic equation . however , the problem can be simplified essentially if one is interested in the soft part of the energy loss spectrum . just this situation was in the experiments @xcite-@xcite , since the measurements were performed in the region of the photon energies of one to five orders of magnitude lower than the electron energy . we consider first the spectral distribution of the energy loss . the probability of the successive radiation of @xmath586 soft photons with energies @xmath780 by a particle with energy @xmath781 on the length @xmath383 in the energy intervals @xmath782 is given by expression @xmath783 where @xmath555 is the normalization constant , @xmath784 is the differential probability of the photon radiation per unit length , @xmath785 is the differential probability of the photon radiation per length @xmath383 . if the probability @xmath786 does nt depend on the particle energy @xmath291 then integrating eq.([y1 ] ) over all photon energies we obtain @xmath787 the value @xmath555 is defined by the condition that probability of all the possible events with radiation of any number of photons or without photon radiation is equal to unit . @xmath788 using eqs.([y1]),([y3 ] ) we can write the expression for the differential distribution of the energy loss in the form @xmath789 here on the right - hand side we have the sum of the probabilities of radiation of @xmath586 photons with summary energy @xmath7 . using standard parametrization of @xmath742-function @xmath790 we obtain @xmath791 d\omega_1\right\}ds \nonumber \\ & & = \frac{1}{\pi}~{\rm re } \int_{0}^{\infty}\exp\left(ix \right ) \exp\left\{-\int_{0}^{\infty}\frac{dw}{d\omega_1 } \left[1-\exp\left(-ix\frac{\omega_1}{\omega}\right)\right ] d\omega_1\right\}dx \label{y5}\end{aligned}\ ] ] the formula ( [ y5 ] ) was derived by landau @xcite ( see also @xcite ) as solution of the kinetic equation under assumption that energy loss are much smaller than particle s energy ( the paper @xcite was devoted to the ionization energy loss ) . let us notice the following . the energy loss are defined by the hard part of the radiation spectrum . in the soft part of the energy loss spectrum eq.([y5 ] ) the probability of radiation of one hard photon only is taken into account accurately . to calculate the probability of the emission of two and more hard photons one has to take into account step by step the recoil in previous acts of the photon emission . the probability of radiation of two and more hard photons is of the order @xmath792 and so on . thus , the formula ( [ y5 ] ) is applicable for the thin targets and has accuracy @xmath793 . if we want to improve accuracy of eq.([y5 ] ) and for the case of thick targets @xmath794 one has to consider radiation of an arbitrary number of hard photons . this problem is solved in appendix of @xcite for the case when hard part of the radiation spectrum is described by the bethe - maximon formula . in this case the eq.([y5 ] ) acquires the additional factor . as a result we extend this formula on the case thick targets . we will analyze first the interval of photon energies where the bethe - maximon formula is valid . we write it in the form ( within the logarithmic accuracy ) @xmath795 \vartheta(\varepsilon-\omega)= \beta\left(1-x_1 + \frac{3}{4}x_1 ^ 2 \right)\vartheta(1-x_1 ) , \label{y7}\ ] ] where @xmath383 is the thickness of the target , @xmath46 is the radiation length , @xmath796 the integral in the curly brackets in eq.([y5 ] ) is @xmath797 substituting this result into eq.([y5 ] ) we have @xmath798 where @xmath799 is the euler gamma function . if we consider radiation of the one soft photon , we have from eq.([y5 ] ) @xmath800 . thus , the formula ( [ y12 ] ) gives additional `` reduction factor '' @xmath801 which characterizes the distortion of the soft bethe - maximon spectrum due to multiple photons radiation . the derivation of eq.([y12 ] ) is based on expression ( [ y5 ] ) in which the statistical independence of photon radiation acts is assumed ( the poisson distribution ) . this statistical independence is broken when two or more hard photons ( energies of which are of the order of the energy of the initial electron ) are emitted . since the probability of hard photon radiation is of the order of @xmath802 , eq.([y12 ] ) is applicable , strictly speaking , at @xmath803 and it has the accuracy up to factor ( @xmath804 ) . however , the contribution of arbitrary number of hard photons radiated in the beth - maximon region is exactly calculated . this contribution is contained in the factor @xmath805 ( see appendix of @xcite , eqs.(a.15 ) , ( a.16 ) ) . so we obtain the reduction factor valid for @xmath806 @xmath807 in fig . 9 the function @xmath808 is given for electron with the energy @xmath291 = 25 gev . one can see that for @xmath7 = 100 mev the difference between the value @xmath808 and unit is 5 - 6 times greater than value @xmath802 . let us discuss this circumstance . the emission of accompanying photons with energy much less or of the order of @xmath7 ( we consider in this figure the situation connected with slac experiment @xcite-@xcite where @xmath809 ) changes the spectral distribution on quantity order of @xmath802 . however , if one photon with energy @xmath810 is emitted , at least , then photon with energy @xmath7 is not registered at all in the corresponding channel of the calorimeter . since mean number of photons with energy larger than @xmath7 is determined by the expression ( see eqs.([y1])-([y5 ] ) ) @xmath811 the probability of the event when no photon with energy @xmath812 is radiated is defined by @xmath813 . this is just the main factor in the expression for the reduction factor @xmath801 ( [ y13 ] ) . in the case , when radiation is described by the bethe - maximon formula the value @xmath814 increases as a logarithm with @xmath7 decrease ( @xmath815 ) and for large ratio @xmath816 the value @xmath814 is much larger than @xmath802 . thus , amplification of the effect is connected with a large interval of the integration ( @xmath817 ) at evaluation of the radiation probability . the bethe - maximon formula becomes inapplicable for the photon energies @xmath818 is defined in eq . ( [ 2.31b ] ) ) , where lpm effect starts to manifest itself ( see eqs.([2.25 ] ) , ( [ 2.41 ] ) ) . calculating the integral in eq.([y5 ] ) we find for the distribution of the spectral energy loss @xmath819 where @xmath820 is the reduction factor in the photon energy range where the lpm effect is essential , @xmath821 in this expression the terms @xmath738 ( see sec.2.1 ) are not taken into account . in fig . 10 the function @xmath822 is shown for @xmath823100 mev ( @xmath291 = 25 gev , @xmath824 = 228 mev ) . it is seen in fig . 10 that the reduction factor @xmath820 changes appreciably in the region of high photon energies solely . this is due to the fact that here the total probability of photon radiation is finite ( in contract to the bethe - maximon formula ) and the integral which defines this probability converges at @xmath825 . in the above analysis we neglected an influence of the polarization of a medium on the bremsstrahlung this is correct if @xmath826 ( see eqs.([42 ] ) , ( [ 43]),([2.30b ] ) , sec . 2.6 ) where @xmath827 we assume , for definiteness , that @xmath828 . this is true in any case for a dense matter if electron energy @xmath702 . in the opposite case when @xmath829 there is an additional suppression of the bremsstrahlung ( see sec . 2.6 ) : @xmath830 and this contribution into reduction factor ( [ y26 ] ) can be neglected . the main contribution into effect considered for the photon energies such that @xmath831 gives the transition radiation eqs.([51d ] ) , ( [ z34a ] ) . let us note , that for the transition radiation the given above derivation of the formula ( [ y5 ] ) is directly inapplicable . however , for application of formula ( [ y5 ] ) it is enough that acts of radiation of soft photons are statistically independent . taking integrals in eq.([y5 ] ) and taking into account that @xmath832 ( but it may be that @xmath833 ) we obtain @xmath834 \nonumber \\ & & \times \sin \left[2\pi \eta \left(\sigma+\psi(1 - 2\eta \sigma ) \right)\right ] \gamma(1 - 2\eta \sigma ) , \label{y38}\end{aligned}\ ] ] where @xmath835 , @xmath805 is defined in eq.([y13 ] ) , @xmath836 in the case @xmath837 we have @xmath838 . \label{y39}\end{aligned}\ ] ] the case of a thin target , where the photon formation length @xmath839 @xmath383 is the target thickness , should be analyzed separately . we consider situation when effects of the polarization of a medium are weak ( @xmath840 ) in this case the scattering takes place on a target as a whole during the radiation process and the spectral probability @xmath786 in eqs.([y2])-([y5 ] ) depends on the momentum transfer @xmath841 ( see eq.([81 ] ) ) @xmath842 on the final step one have to average @xmath843 over @xmath33 with the distribution function @xmath844 eq.([82 ] ) . since the expression for the spectral distribution ( [ y41 ] ) has the same form as in the bethe - maximon case , one can use eq . ( [ y12 ] ) for calculation of the reduction factor . as a result we have for the contribution of the thin target region @xmath845\right > \simeq 1-\frac{2\alpha}{\pi } \displaystyle{\frac{\left < f_2 ^ 2\left(\frac{q}{2}\right ) \right > } { \left < f_2\left(\frac{q}{2}\right ) \right>}\ln \frac{\omega_{th}}{\omega } } , \label{y42}\ ] ] where @xmath846 is the boundary energy starting from which the target becomes thin . the lower limit of the applicability of eq.([y42 ] ) is determined from the condition @xmath847 . the expression ( [ y42 ] ) depends essentially on the mean square of the momentum transfer @xmath848 associated with the target thickness . in the case @xmath849 when an influence of the multiple scattering is weak ( @xmath166 ) @xmath850 where @xmath518 is the effective screening radius eq.([2.18 ] ) . when the multiple scattering is strong ( @xmath851 ) , the reduction factor can be calculated using the gaussian distribution for @xmath852 @xmath853 in the first case ( @xmath849 ) one has @xmath854 and one can neglect the correction to unit in eq.([y42 ] ) . in the opposite case ( @xmath855 ) the main contribution into entering mean values of @xmath856 gives values @xmath857 where @xmath858 in this case the effect under consideration may be noticeable . from the above analysis one can find an approximate expression for the reduction factor in the general form . one can present the integral over @xmath859 in eq.([y5 ] ) as @xmath860 d\omega_1 \simeq -\int_{\omega}^{\infty}\frac{dw}{d\omega_1}d\omega_1 + \int_{0}^{\omega}\frac{dw}{d\omega_1 } \left[\exp\left(-ix\frac{\omega_1}{\omega}\right)-1 \right ] d\omega_1 . \label{y3.1}\ ] ] the second integral on the right - hand side of eq.([y3.1 ] ) ( as well as terms omitted in the first integral ) has an order @xmath861 . for the case @xmath803 this value is small @xmath862 let us note that a contribution of the transition radiation can be enlarged @xmath586 times if one makes a target as a collection of @xmath586 plates conserving the total thickness @xmath383 provided that definite conditions are fulfilled for the plate thicknesses and gaps between plates . in that case @xmath863 and above formulae are valid if @xmath864 . expanding the exponential in eq.([y5 ] ) with the second integral of eq.([y3.1 ] ) and integrating over @xmath2 we have @xmath865 d\omega_1 \nonumber \\ & & = \int_{0}^{\omega}\frac{dw}{d\omega_1}\delta\left(1-\frac{\omega_1 } { \omega } \right)d\omega_1 = \omega \frac{dw}{d\omega } \label{y3.4}\end{aligned}\ ] ] thus , the spectral distribution of the energy loss and reduction factor @xmath866 have the following general form @xmath867 \left(1+o\left(\omega \frac{dw}{d\omega } \right ) \right ) \label{y3.5}\ ] ] the radiation from several plates for the relatively hard part of the spectrum in which the bremsstrahlung in the condition of the strong lpm effect dominates was investigated recently in @xcite . a rather curious interference pattern in the spectrum of the radiation was found which depends on a number ( and a thickness ) of plates and the distance between plates ( the polarization of a medium was neglected ) . in the present subsection the probability of radiation in a radiator consisting of @xmath385 plates is considered @xcite . the transition radiation dominates in the soft part of the considered spectrum , while the bremsstrahlung under influence of the strong lpm effect dominates in the hard part . the intermediate region of the photon energies where contributions of the both mentioned mechanisms are of the same order is of evident interest . we analyze this region in detail . in this region effects of the polarization of the medium are essential . here we consider the case when the target consists of n identical plates of thickness @xmath868 with the equal gaps @xmath869 between them . the case @xmath870 will be analyzed where @xmath66 is the characteristic formation length of radiation in absence of a matter ( @xmath871 ) , see eqs.([4])-([6 ] ) @xmath872 where @xmath162 is the characteristic angle of radiation . one can split the spectral distribution of the probability of radiation into two parts @xmath873 where @xmath874 is the spectral distribution of the probability of bremsstrahlung with allowance for the multiple scattering and polarization of a medium and @xmath875 is the probability of the transition radiation for the @xmath385-plate target . for @xmath876 one has : @xmath877 \nonumber \\ & & \times \left[r_1 \left(g_n - g_n(0 ) \right)+ r_2 \left(g_n^2-g_n^2(0 ) \right ) \right ] , \label{x27}\end{aligned}\ ] ] where @xmath878 + i\left(d_n - d_{n-1 } \right)\left(\tau_1+\tau_2\right)- \beta_n^{-1}\tau_1\tau_2 ; \nonumber \\ & & n \geq n = n_2-n_1 + 1 \geq 1;\quad n_1=0,~0 \leq \tau_1 < \infty ; \quad n_2=n-1,~0 \leq \tau_2 < \infty ; \nonumber \\ & & n_1 \geq 1,~ ~0 \leq \tau_1 \leq t;\quad n_2 \leq n-2,~0 \leq \tau_2 \leq t , \nonumber \\ & & \overline{\kappa}=1+\overline{\kappa_0 ^ 2},~ \overline{\kappa_0 ^ 2}=\kappa_0 ^ 2 \frac{l_1}{l_1+l_2 } \label{x28}\end{aligned}\ ] ] note for the subtraction procedure one has that when @xmath879 , the function @xmath880 . the function @xmath881 can be presented in the form @xmath882 for the case @xmath883 one has @xmath884 in above formulae @xmath885 for one plate ( @xmath886 ) we have @xmath887 and @xmath888 in the integral ( [ x27 ] ) we rotate the integration contours over @xmath889 on the angle @xmath890 and substitute variables @xmath891 . then we carry out change of variables @xmath892 . we have @xmath893dz \nonumber \\ \hspace{-8 mm } & & = \frac{\alpha b}{\pi \omega } \cos ( \kappa t_1 ) \int_{0}^{\infty } \exp ( -bx ) \left[r_1f_1\left(\sqrt{\frac{x}{4 } } \right ) + r_2f_2\left(\sqrt{\frac{x}{4 } } \right)\right],\nonumber \\ & & g(x , z)=1+\frac{x}{b}z(1-z ) , \label{x30}\end{aligned}\ ] ] where the functions @xmath894 and @xmath895 are defined in ( [ 4.18sc ] ) and ( [ 4.17sc ] ) . the special case is when a target consists of two plates ( @xmath896 ) . since the formation length is enough long for the soft photons ( @xmath26 ) only , we consider the term with @xmath897 in eq.([x27 ] ) as far as @xmath898 . for the case ( @xmath896 ) the sum in eq.([x27 ] ) consists of three terms : @xmath899 . for the two first @xmath900 and we have from eq.([x29 ] ) @xmath901d\tau_1 \bigg ] . \label{x31}\end{aligned}\ ] ] for the third term @xmath902 ( see eq.([x18 ] ) ) and we have @xmath903 , \label{x32}\end{aligned}\ ] ] so that @xmath904 ^ 2 } \right]dx_2 \bigg ] . \label{x33}\end{aligned}\ ] ] here we rotate the integration contours over @xmath889 on the angle @xmath890 and substitute variables @xmath891 . in the case of the weak multiple scattering ( @xmath905 ) neglecting the effect of the polarization of a medium ( @xmath906 ) and expanding the integrand in eqs.([x31 ] ) and ( [ x33 ] ) over @xmath907 we have for the probability of radiation @xmath908 , \nonumber \\ & & g(t)=t^2 \int_{0}^{\infty } \frac{x^3 \exp ( -x)}{(x^2+t^2)^2 } \bigg[\left(1-\frac{3x^2}{10t^2 } \right ) \cos\left(t-4\arctan\frac{x}{t } \right ) \nonumber \\ & & + \frac{2x}{t } \sin\left(t-4\arctan\frac{x}{t } \right ) \bigg ] . \label{x34}\end{aligned}\ ] ] note that the main term of the decomposition in eq.([x34 ] ) is the bethe - maximon probability of radiation from two plates which is independent of the distance between plates . this means that in the case considered we have independent radiation from each plate without interference in the main order over @xmath907 . the interference effects appear only in the next orders over @xmath907 . in the case of the strong multiple scattering we have @xmath909 . when the formation length is much longer than the target thickness as a whole ( @xmath910 ) we find for the probability of radiation @xmath911 \cos ( 2\kappa t_1 ) \label{x40}\ ] ] in the opposite case @xmath912 neglecting the effect of the polarization of a medium ( @xmath913 ) we have @xmath914 \label{x41}\ ] ] the case of the strong multiple scattering ( @xmath915 ) for the photon energies where the value @xmath916 is of the special interest . in this case we can neglect the polarization of a medium @xmath917 , and disregard the terms @xmath918 in the exponent of the expressions ( [ x31 ] ) and ( [ x33 ] ) since @xmath919 . we obtain as a result @xmath920 where @xmath921 is the integral sine and @xmath922 is the integral cosine . at @xmath714 we have @xmath923 let us turn to the case of large @xmath385 ( @xmath924 ) . if the formation length of the bremsstrahlung is shorter than the distance between plates ( @xmath925 ) the interference of the radiation from neighboring plates takes place . using the probability of radiation from two plates ( [ x34 ] ) we obtain in the case of weak multiple scattering ( @xmath905 ) @xmath926 , \label{x43}\ ] ] where for @xmath714 the function @xmath927 is defined in eq.([x34 ] ) . in the case of strong multiple scattering ( @xmath928 ) and large @xmath340 we have ( compare with eqs.([x42a ] ) and ( [ x42b ] ) ) @xmath929 the behavior of the spectral distribution @xmath930 can be discussed using as an example the case of two plates with the thickness @xmath868 and the distance between plates @xmath931 which was analyzed in detail above . for plates with the thickness @xmath932 and in the energy interval @xmath933 , in which the effects of the polarization of a medium can be discarded , the condition ( [ x5 ] ) is fulfilled only for enough high energy @xmath291 so that @xmath828 ( see eqs.([5 ] ) , ( [ 2.31b ] ) ) . we study the situation when the lpm suppression of the intensity of radiation takes place for relatively soft energies of photons : @xmath934 . for the hard photons @xmath935 the formation length @xmath43 eq.([6 ] ) is much shorter than the plate thickness @xmath936 , the radiation intensity is the incoherent sum of radiation from two plates and it is independent of the distance between plates . in this interval the bethe - maximon formula is valid . for @xmath937 the lpm effect turns on , but when @xmath938 the thickness of plate is still larger than the formation length @xmath939 @xmath940 so that the formation of radiation takes place mainly inside each of plates . with @xmath7 decreasing we get over to the region where the formation length @xmath941 , but effects of the polarization of a medium are still weak ( @xmath933 ) . within this interval ( for @xmath942 ) the main condition ( [ x5 ] ) is fulfilled . to estimate the value @xmath846 we have to take into account the characteristic radiation angles ( @xmath943 in eq.([x5 ] ) ) , connected with mean square angle of the multiple scattering . using the definition of the parameter @xmath944 in eq.([xd1 ] ) ) we have @xmath945 where the parameter @xmath946 is defined by a set of equations ( we rearranged terms in ( [ 73 ] ) ) : @xmath947 here @xmath948 . one can show ( see discussion after eq.(3.6 ) in @xcite ) that @xmath949 . naturally , @xmath950 , and when @xmath951 one has @xmath952 . it is seen from eq.([x48 ] ) that when the value @xmath868 decreases , the region of applicability of results of this subsection grows . so , when @xmath942 the formation length is longer than the thickness of the plate @xmath868 and the coherent effects depending on the distance between plates @xmath869 turn on . for the description of these effects for @xmath953 one can use eq.([x42a ] ) . for @xmath954 one can use the asymptotic expansion ( [ x42b ] ) and it is seen that at @xmath955 the spectral curve has minimum . let us note an accuracy of formulas is better when @xmath7 decreases , and the description is more accurate for @xmath956 . with further decreasing of the photon energy @xmath7 the value @xmath340 diminishes and the spectral curve grows until @xmath721 . when @xmath957 the spectral curve decreases @xmath958 according with eq.([x41 ] ) . so , the spectral curve has maximum for @xmath721 . the mentioned decreasing continues until the photon energy @xmath7 for which @xmath959 . for smaller @xmath7 the thickness of the target is shorter than the formation length . the results of numerical calculations are given in fig . 11 . the formulas ( [ x31 ] ) , ( [ x33 ] ) were used . the spectral curves of energy loss were obtained for the case of two gold plates with the thickness @xmath960 with different gaps @xmath869 between plates . the initial energy of electrons is @xmath961 gev . the characteristic parameters for this case are : @xmath962 at @xmath963mev the radiation process occurs independently from each plate according with theory of the lpm effect given above . the interference pattern appears at @xmath964mev where the formation length is longer than the thickness of one plate and the radiation process depends on the distance between plates @xmath340 . according to eqs.([x42b ] ) , ( [ x43 ] ) the curves 1 - 5 have minimums at @xmath965 which are outside of fig . 11 and will be discussed below . in accord with the above analysis the spectral curves in fig . 11 have the maximums at photon energies @xmath966 . these values ( in mev ) are @xmath967 for curves @xmath968 respectively . at further decrease of @xmath969 the spectral curves diminish according to eq.(([x41 ] ) ) and attain the minimum at @xmath970 . the corresponding values ( in mev ) are @xmath971 for curves @xmath972 . the least value of @xmath973mev has the curve @xmath974 . however , one has to take into account that at @xmath975mev ( @xmath976 ) the contribution of the transition radiation becomes significant . starting from @xmath977mev the contribution of the transition radiation dominates . the approach presented in this subsection is applicable in the interval of photon energies where effects of the polarization of a medium are essential . it includes also the soft part of the lpm effect . for the case given in fig . 11 our results are given up to @xmath978mev . on the other hand , in @xcite the hard part of the lpm effect spectrum was analyzed where one can neglect the effects of the polarization of a medium ( @xmath979mev for the mentioned case ) . although here ( as in @xcite ) in and in @xcite the different methods are used , the results obtained in overlapping regions are in a quite reasonable agreement among themselves . it is interesting to discuss behavior of the spectral curves obtained in @xcite from the point of view of our results . we consider the low value @xmath980 is not very large ) and a situation when corrections to the value @xmath130 in eq.([x48 ] ) which neglected in @xcite are less than @xmath981 . this leads to the difference in results less than @xmath982 . we concentrate on the case of gold target with the total thickness @xmath983 . the case @xmath886 where @xmath984mev , @xmath985mev was considered in detail in @xcite ( see sec . the curves in figures in @xcite are normalized on the bethe - heitler probability of radiation ( no the coulomb corrections ) , i.e. they measured in units @xmath986 . in the region where our results are applicable @xmath987mev in fig . 2 of @xcite @xmath988 one can see plateau the ordinate of which is @xmath982 less than calculated according with eq.([x30 ] ) . the case of two plates ( @xmath989mev , @xmath990mev ) is given in fig . 3 of @xcite . the lengths of the gaps are the same as in our fig . 11 , except @xmath991 . the positions and ordinates of the minimums and the maximums in the characteristic points ( @xmath992 for minimums and @xmath966 for maximums , see above ) as well as behavior of the spectral curves is described quite satisfactory by our formulas ( see e.g. asymptotic eqs.([x40])-([x42b ] ) ) . as it was said above , migdal have calculated the probability of the bremsstrahlung and pair creation in the logarithmic approximation @xcite.this result , now considered as classic one , was confirmed in the most of subsequent calculation . we developed the method of calculation @xcite which goes beyond the logarithmic approximation . the formula ( [ 2.25 ] ) gives the spectral distribution of the probability of radiation which coincides formally with the probability calculated by migdal ( eq.(49 ) in @xcite ) . however , here coulomb corrections are included into parameter @xmath229 in contrast to @xcite , @xcite . as it was noted , the parameter @xmath161 entering into the parameter @xmath229 ( see eqs.([2.24 ] ) , ( [ 2.16 ] ) and ( [ 2.14 ] ) ) is defined up to the factor @xmath993 , what is inherent in the logarithmic approximation . however , we calculated also the next term of the decomposition over @xmath180 ( an accuracy up to the `` next to leading logarithm '' ) and this permits to obtain the result which is independent of the parameter @xmath161 . it was shown that the definition of the parameter @xmath161 minimizes corrections to eqs.([2.25 ] ) and ( [ 2.10p ] ) practically for all values of the parameter @xmath161 . it should be emphasized also that here the coulomb corrections are included into the parameter @xmath229 in contrast to @xcite . moreover , the procedure of fine tuning presented above ( see @xcite ) permits one ( after adding the term @xmath994 ( [ 2.30 ] ) ) to obtain more accurate expression for the probability of radiation , which has power accuracy in the limit of weak multiple scattering , see ( [ 2.35 ] ) , while the probability calculated by migdal @xcite has logarithmic accuracy only in the mentioned limit . the interrelation between the main term @xmath995 and the first correction @xmath994 depends essentially on fine tuning procedure which was chosen , as it was just said , in a such way that minimize the value of the first correction . for example , one can see that the maximal value of @xmath994 is of the order 7 % of the main term for all four spectral curves shown in the fig . this means that in our case there are more rigid requirements on fine tuning procedure than used by migdal . in the approach used by blankenbecler and drell ( bd ) @xcite the scattering process was considered in the second approximation of the eikonal formalism and an analysis of scattering in medium was based on the model of a random medium . the radiation was calculated for all possible paths and averaged . one advantage of this approach is that it treats naturally finite thickness targets . there are a few short - comings in the bd approach . the gaussian distribution of scattered electrons was obtained by migdal @xcite as a solution of the fokker - planck equation , while in bd approach this is an input . actually there are deviations from the gaussian distribution , see end of sec.3.3 . the calculation of the probability @xmath996 is quite unwieldy in bd approach and general formula for @xmath996 ( eq.(5.32 ) in @xcite)is written down in an implicit form only , and the functions @xmath997 and @xmath998 in the mentioned formula have to be calculated in the each case separately . this means that comparison of bd probability with that calculated by migdal can be done in limiting cases only . in the case of weak multiple scattering the probability in @xcite coincides with the bethe - heitler probability taken with logarithmic accuracy ( just as in migdal paper @xcite ) . however , in the limit of the strong lpm effect , where the probability of radiation is given by eqs.([2.25 ] ) and ( [ 2.39 ] ) above and which coincides with obtained by migdal eq.(52 ) in @xcite ( if one omits the coulomb corrections included into the parameter @xmath696 ) , the probability calculated by bd ( the term with @xmath999 ) differs from ( [ 2.39 ] ) by factor @xmath1000=1.0854 . the origin of this discrepancy was not analyzed in @xcite . in the paper of r.baier , dokshitzer , mueller , peigne , and schiff ( bdmps ) @xcite the multiple scattering of high - energy electrons off a large number of scatterers ( presented as a screened coulomb potential ) is analyzed within eikonal formalism . radiation of soft photons only is included using the classical current approach , although this approach gives no significant simplification for the problem under consideration . all the results are obtained within logarithmic accuracy . no general formula of the type of eq.([2.41 ] ) is written down . in the `` bethe - heitler limit '' the result of @xcite does not completely match the standard bethe - heitler formula . since the method used consists in direct summation over scatterers it naturally included the consideration of a target of finite thickness . for thin target ( @xmath1001 ) the result of @xcite is similar to logarithmic term in eq.([77 ] ) where @xmath1002 for soft photons . the explicit formula for radiation spectrum is given in @xcite in the limit of strong lpm effect where it is similar to eq.([2.39a ] ) if @xmath26 and the coulomb corrections are omitted . note that in migdal papers @xcite , @xcite and in our papers the kinetic equation is the basic element of the used methods while in the both papers @xcite and @xcite the consideration does not included the kinetic equation . due to this feature the calculations in the both papers @xcite and @xcite are more cumbersome and no general formulas for radiation spectra are obtained . a path integral method is applied successfully for analysis of multiple scattering ( see e.g. @xcite ) . zakharov @xcite , @xcite is used this method to study of the lpm effect . using a transverse green function based on a path integral he succeeded in reduction of the problem to the two dimensional schrdinger equation , i.e. he re - derived equation a la eq.([2.9 ] ) , which was obtained earlier in @xcite . the probability of radiation was calculated ( see eq.(19 ) in @xcite ) within logarithmic accuracy ( as in migdal papers @xcite and @xcite ) although it is written down in a more cumbersome form . the procedure of fine tuning so important for migdal calculation @xcite is oversimplified in @xcite . for earlier attempt to do study of the lpm effect in a such way see @xcite . the formalism used in @xcite has the same region of validity as in migdal s papers @xcite , @xcite and in our papers @xcite , @xcite and @xcite in spite of claims about `` rigorous treatment of the lpm effect '' . the developed formalism allows naturally to consider the lpm effect in finite thickness targets @xcite . as an example the calculation of spectral intensity of radiation is given for electron energy @xmath28 gev and gold target with thickness @xmath1003 . this calculation was carried out numerically with the complete potential ( see eq.([2.9 ] ) ) outside of scope of logarithmic approximation . the coulomb corrections were discarded as well as a multiple photon radiation . after arbitrary diminishing of calculated value by 7@xmath1004 it was found very good agreement with a portion of data ( in the region of the lpm drop up to beginning of plateau ) . actually this diminishing should be attributed to coulomb corrections found in @xcite ( for details see @xcite and below in sec . 3.10 ) . more late results of zakharov see in @xcite . we consider the spectral intensity of radiation for the energy of the initial electrons when the lpm suppression of the intensity of radiation takes place for relatively soft energies of photons : @xmath934 : @xmath1005 see eqs.([2.9 ] ) , ( [ 2.16 ] ) , ( [ 2.32a])-([2.31b ] ) . this situation corresponds to the experimental conditions @xcite . the ratio of the thickness of the target @xmath383 and the formation length of radiation @xmath1006 ( [ 5 ] ) is an important characteristics of the process . in eq.([5 ] ) both the multiple scattering and the polarization of a medium are taken into account . this ratio may be written as @xmath1007 } , \nonumber \\ & & \displaystyle{t=\frac{l \omega}{2\gamma^2},\quad \omega_p=\omega_0 \gamma,\quad t_c \equiv t(\omega_c ) \simeq \frac{2\pi}{\alpha}\frac{l}{l_{rad } } } , \label{101}\end{aligned}\ ] ] where we put that @xmath1008 . below we assume that @xmath1009 which is true under the experimental conditions . if @xmath1010 then at @xmath1011 a target is thick and one has the lpm suppression for @xmath937 . there are two opportunities depending on the minimal value of the parameter @xmath802 : @xmath1012 if @xmath1013 then for photon energies @xmath1014 it will be @xmath463 such that @xmath1015 and for @xmath1016 the thickness of a target becomes smaller than the formation length of radiation so that for @xmath1017 the spectral distribution of the radiation intensity is described by eqs . ( [ z24 ] ) , ( [ z30 ] ) . under these conditions for @xmath1018 the spectral curve has a plateau @xmath1019 in accordance with eq.([79 ] ) . under conditions @xmath1020 the spectral intensity of radiation is independent of photon energy @xmath7 . it should be noted that due to smallness of the coefficients in expression for @xmath742 eq.([z26 ] ) , such behaviors of the spectral curve begins at @xmath1021 . it will continue until photon energies where one has to take into account the polarization of a medium and connected with it a contribution of the transition radiation . at @xmath1022 a target remains thick for all photon energies and radiation is described by formulae of sections 2.2 and 2.6 . in this case at @xmath1023 and @xmath1024 @xmath1025 the spectral intensity of radiation formed inside a target is given by eqs.([2.25 ] ) and ( [ 2.30 ] ) and the contribution of the boundary photons is given by eq.([51 g ] ) . it is important to include also the contribution of boundary photons . since the contribution into the spectral intensity of radiation from a passage of the electron inside the target ( @xmath736 ) is diminishing and a contribution of the boundary photons is increasing with @xmath7 decrease , the spectral curve has a minimum at @xmath1026 . the value of @xmath1027 may be estimated from equations ( see ( [ 2.39 ] ) , ( [ z24])-([z25 ] ) ) ) @xmath1028 when the value of @xmath1029 is high enough , the solution of eq.([104 ] ) does not satisfy the condition @xmath1030 and in this case the equation ( [ 104 ] ) ceases to be valid . for determination of @xmath1027 in this case we use the behavior of the spectral intensity of radiation at @xmath1031 . in this case the contribution into radiation from inside passage of the target is described by eq.([47 ] ) while the radiation of the boundary photons reduces to the transition radiation and its contribution is given by eq.([51d ] ) . leaving the dominant terms ( @xmath751 is @xmath7 independent ) we have @xmath1032 since the value @xmath1033 is of the order of unity , the solution of ( [ 104 ] ) at @xmath1034 differs only slightly from @xmath1027 . because of this , if the condition @xmath1035 is fulfilled , the position of the minimum is defined by eq.([104 ] ) . the formulas ( [ 104 ] ) and ( [ 105 ] ) show the position of a minimum of the spectral intensity curve which is the sum of contributions of radiation inside a target and boundary radiation . due to the lpm effect the intensity of radiation inside a target diminishes with @xmath7 decreasing . however at low @xmath7 the contribution of the boundary radiation is sharply increasing and this results in appearance of a minimum of spectral curve . eq.([104 ] ) gives its position at small @xmath1029 while eq.([105 ] ) gives minimum position at large @xmath1029 . if @xmath1036 then at @xmath1037 a target is thin and the bethe - maximon spectrum of radiation which is valid at @xmath1038 will be also valid at @xmath937 in accordance with eq.([z30 ] ) since @xmath1039 . this behavior of the spectral curve will continue with @xmath7 decrease until photon energies where a contribution of the transition radiation become essential . in this case the spectral distribution of radiation has the form ( [ z35 ] ) for all @xmath7 @xmath1040 since for soft photons ( @xmath26 ) @xmath1041 \label{z39}\ ] ] and @xmath1042 a contribution of the transition radiation become visible already at @xmath632 . for @xmath1043 the probability of radiation is defined by eqs.([z22b])-([z22d ] ) . in this case a considerable distinction from bethe - maximon formula will be in the region @xmath1044 . as well known , for soft photons ( @xmath26 ) the bethe - maximon formula for the spectral intensity of radiation does nt depend on a photon energy as well . so , the ratio of these spectral intensities ( see eqs . ( [ z30 ] ) and ( [ ad33 ] ) ) is an important characteristics of the phenomenon under consideration : @xmath1045 \label{z41c}\ ] ] for @xmath1046 one has @xmath1047 and for @xmath1048 one has using expression ( [ 77 ] ) @xmath1049 - 2 + \frac{c}{l_t } \label{z41d}\ ] ] for estimates one can put with a good accuracy @xmath1050 and @xmath1051 . there is , in principle , an opportunity to measure the electron energy ( in region of high energies ) using the lpm effect . for this one can measure the spectral curve on a target with thickness a few percent of @xmath46 and compare the result with the theory prediction . existence of the plateau of the spectral curve in a region of photon energies where a target is thin was first found in @xcite within migdal approach ( quantum theory ) . recently this item was discussed in @xcite ( in classical theory ) , @xcite and @xcite . bremsstrahlung or pair creation suppression can be studied with high energy electron or photon beams . because pair creation suppression requires photons with @xmath1052 ( see eq.([2.23ap ] ) ) , beyond the reach of current accelerators , pair creation has been studied only with cosmic rays , with consequently very limited statistics . the best suppression studies have used electron beams at accelerators . besides the lpm effect , these beams have been used to study effect of polarization of a medium and boundary photon emission . the first tests of lpm suppression came shortly after migdal s appeared . these were cosmic rays experiments with high energy photons ( @xmath1053 tev ) and studied the depth of pair conversion in a dense target @xcite-@xcite , see also more recent @xcite . all of the air shower experiments suffered from the poor statistics . uncertainties in the photon spectrum complicated the analysis . so , these experiments are at best qualitative verification of the lpm effect . the first accelerator based study of lpm effect was done in serpukhov @xcite using 40 gev electrons . photons with energy 20 mev @xmath1054 70 mev emitted from carbon , aluminum , lead and tungsten targets were detected in an sodium iodide calorimeter . it were several limitations due to substantial background . the results agreed with migdal s predictions . a precision measurement of lpm suppression , and study of effect of polarization of a medium was performed by the e-146 collaboration at slac using 25 gev and 8 gev electrons . the experimental setup was conceptually similar to the serpukhov experiment , but heavily optimized to minimize background . the target used are listed in table 1 . photons with energy 200 kev @xmath1054 500 mev were detected in a bgo calorimeter.the logarithmic binning of photon spectrum was used with 25 bins per decade . a very large data set was collected @xcite - @xcite . a detailed , high statistics monte carlo simulation was used by e-146 . its main purpose was to understand multi - photon emission ( see sec.3.5 ) . recently new study of lpm effect at higher energies of electrons ( @xmath1055149 , 207 and 287 gev ) , where the effect has influence upon much wider part of spectrum comparing with @xmath105525 gev , was performed in the h2 beam line of the cern sps @xcite . the experimental setup was also conceptually similar to the serpukhov experiment . photons with energy 2 gev @xmath1056 were detected in a lead glass calorimeter . the logarithmic binning of photon spectrum was used with 25 bins per decade as at slac . here we consider the experimental data @xcite-@xcite,@xcite from the point of view of the above analysis . it was shown that the mechanism of radiation depends strongly on the thickness of the target . first , we estimate the thickness of used target in terms of the formation length . from eq.([101 ] ) we have that @xmath1057 the minimal value of the ratio of the thickness of a target to the formation length is given by eq.([102])(@xmath1058 ) . for defined value of @xmath1029 this ratio is least of all for the heavy elements . indeed , the value of @xmath1059 depends weakly on nucleus charge @xmath1060 ev , while @xmath1061 . furthermore , the ratio @xmath1062 decreases with the energy increase . thus , among all the targets with thickness @xmath1063 the minimal value of @xmath1064 is attained for the heavy elements ( au , w , u ) at the initial energy @xmath28 gev . in this case one has @xmath1065 mev , @xmath1066 mev , @xmath1067 . since the parameter @xmath1029 is energy independent and the ratio @xmath1068 , the minimal value @xmath1069 is attained at the initial energy @xmath1070 gev for all the targets with thickness @xmath1063 . so , all such targets can be considered as thick targets at both energies . as an example we calculated the spectrum of the energy losses in the tungsten target with the thickness @xmath1071 mm ( = @xmath1072 ) for both initial energies shown in fig . 12 ( a ) and ( b ) . the characteristic parameters for this case are given in table 2 . we calculated the main ( migdal type ) term eq.([2.25 ] ) , the first correction term eq.([2.30 ] ) taking into account an influence of the polarization of a medium according to eq.([45 ] ) , as well as the coulomb corrections entering into parameter @xmath696 eq.([2.32a ] ) and value @xmath1073 eq.([2.16 ] ) . the contribution of an inelastic scattering of a projectile on atomic electrons was not included into the numerical calculation ( it is @xmath1074 ) , although it can be done using eq.([40a ] ) . we calculated also the contribution of boundary photons ( see eq.(4.12 ) in @xcite ) . here in the soft part of the spectrum @xmath1075 mev for @xmath28 gev ) the transition radiation term eq.([51d ] ) dominates , while in the harder part of the boundary photon spectrum @xmath1076 the terms depending on both the multiple scattering and the polarization of a medium give the contribution ; for @xmath1070 gev one has @xmath1077 kev . all the mentioned contribution presented separately in fig . 12 . under conditions of the experiment the multiphoton reduction of the spectral curve is very essential . the curve `` t '' in the fig . 12 involves the reduction factor which was constructed as the interpolation of eqs.([y13 ] ) , ( [ y26 ] ) and ( [ y39 ] ) with the accuracy up to @xmath1078 . the curves `` t '' in fig . 12 ( a ) and ( b ) are the final theory prediction in units @xmath1079 . experimental data are taken from @xcite and recalculated according with given in mentioned paper procedure : @xmath1080 since photon energies were histogrammed logarithmically , using 25 bins per decade of energy , one has for the coefficient @xmath403 @xmath1081 depending on the normalization point within bin . in the review @xcite it was indicated that @xmath1082 . in more earlier paper @xcite the value @xmath1083 was given and we used it in our paper @xcite and @xcite . the date recalculated with @xmath1084 are also given in fig . it is seen that there is a perfect agreement of the curves t with data for both energies . the gold targets with thickness @xmath1003 and @xmath1085 are an exception and ca nt be considered as thick targets . we calculated energy losses spectra in these targets for both the initial electron energies @xmath1086 and @xmath1087 . the characteristic parameters of radiation for these cases are given in table 3 . in fig . 13(a ) results of calculations are given for target with a thickness @xmath1088 at @xmath1086 . the curves 1,2,3,4 present correspondingly the functions @xmath1089 ( [ z12 ] ) . at @xmath1090 the value @xmath1091 , the interference terms are exponentially small and one can use formulae for a thick target . in this case the parameter @xmath1092 and contribution of boundary photons ( @xmath1093 ) is small ( @xmath1094 , see @xcite , eq.(4.16 ) ) and distinction bethe - maximon formula ( @xmath1095 ) from @xmath1096 is of the order 10% according with asymptotic damping factor @xmath1097 . at @xmath1098 for the case @xmath1048 and @xmath1099 the spectral curve turns into plateau according with discussion in previous subsection . in this photon energy region the parameter @xmath1100 and eq.([z24 ] ) for a target with intermediate thickness describes the spectral probability of radiation with a good accuracy . with the further photon energy decrease one can use limiting formula ( [ z26 ] ) where @xmath1101 . for this case the ratio of ordinate of the plateau to ordinate of the bethe - maximon intensity ( see eq.([z41d ] ) ) is @xmath1102 . note , that in formulae for @xmath1103 the potential @xmath181 eq.([2.16 ] ) is used which does nt include corrections @xmath1104 . these corrections were given above for both thin and thick targets ( see eq.([2.40 ] ) ) . in our case ( @xmath1105 ) the expressions with corrections @xmath738 are given in eqs.([2.40 ] ) and ( [ ad31 ] ) for a thick target and a thin target respectively . taking into account behaviors of correction in the region @xmath1106 ( see curve 2 in fig . 12(a ) ) we construct an interpolation factor ( taking into account the term @xmath738 ) with accuracy of order 1% . the summary curve ( @xmath340 ) in fig . 13(a ) contains this factor . the transition radiation contributes in the region @xmath1107 ( function @xmath1108 , curve 5 in fig . 13(a ) ) . when @xmath632 this curve is described by asymptotic of @xmath1108 eq.([z24 ] ) . the contribution of the transition radiation increases with @xmath7 decrease and for @xmath1109 it describes by eq.([z34 ] ) . the contribution of the multiple scattering diminishes due to interference factor @xmath1110 in eq.([z35 ] ) ( at @xmath1111 ) . the curve @xmath340 in fig . 13(a ) gives the summary contribution of the multiple scattering ( the curve @xmath1112 , where factor @xmath1113 is included ) and the transition radiation ( the curve 5 ) . in fig . 13(b ) results of calculations are given for target with a thickness @xmath1088 at @xmath1087 . the notations are the same as in fig . 13(a ) . in this case the characteristic photon energy @xmath824 is one order of magnitude lower than for @xmath1086 , so that at @xmath1090 the parameter @xmath715 is small ( @xmath1114 ) . because of this the right part of the curve @xmath1112 coincides with a good accuracy with the bethe - maximon formula ( the coulomb corrections are included ) . note , that for this electron energy the effect of recoil ( factor @xmath1113 ) is more essential . strictly speaking , a target with a thickness @xmath1115 at @xmath1087 is not thin target for any photon energy ( @xmath1116 ) . however , for bremsstrahlung this target can be considered as a thin one for @xmath1117 . since the polarization of a medium becomes essential in the same region ( @xmath1118 ) , the interference factor @xmath1110 in eq.([z35 ] ) causes an inflection of the spectral curve @xmath1112 at @xmath1119 . the transition radiation grows from the same photon energy @xmath7 and because of this the total spectral curve @xmath340 has a minimum at @xmath1120 . as far as there is some interval of energies between @xmath700 and @xmath846 ( @xmath1121 ) , this minimum is enough wide . moreover , the value of its ordinate coincide with a good accuracy with ordinate of the plateau of the spectral curve @xmath1112 in fig . 18(a ) because bremsstrahlung on a thin target is independent of electron energy ( [ ad31]).it should be mentioned also that for this thickness the reduction factor due to multiple photon emission ( see eq.([y26 ] ) ) @xmath1122 ( so this is the effect of the same scale ) but this reduction is compensated nearly exactly because of more precise definition of data ( change of a bin width @xmath1123 in @xcite ( used in @xcite ) to a bin width @xmath1124 in @xcite , @xcite , see discussion in sec.3 of @xcite ) . in fig.14(a ) results of calculations are given for target with a thickness @xmath1125 at @xmath1086 . the notations are the same as in fig . for this thickness @xmath1126 and one has a thin target starting from @xmath1127 . so , we have here very wide plateau . the left edge of the plateau is defined by the contribution of transition radiation ( @xmath1128 ) . since in this case @xmath1129 ( see eqs.([z30])-([ad33 ] ) ) , one has to calculate the ordinate of the plateau using the exact formula for a thin target ( [ z30 ] ) . for this case the ratio of ordinate of the plateau to ordinate of the bethe - maximon intensity ( see eq.([z41d ] ) ) is @xmath1130 . the same ordinate has the plateau for electron energy @xmath1087 ( fig . 13(b ) ) . however , a width of the plateau for this electron energy is more narrow ( @xmath1131 ) due to diminishing of the interval between @xmath700 and @xmath824 . for @xmath1132 the formation length of radiation becomes shorter than target thickness ( @xmath1133 ) and the parameter @xmath715 decreases . the value @xmath1134 increases with @xmath7 growth . a target becomes thick and the spectral curves is described by the bethe - maximon formula in fig . 14(b ) starting from photon energy @xmath1135 . in fig . 14(b ) the contributions of separate terms into @xmath1136 are shown as well . their behavior at @xmath1132 is described quite satisfactory by eqs . ( [ z22b])-([z22f ] ) ( see also discussion at their derivation ) . we compared our calculations with experimental data @xcite . the curves @xmath340 in fig . 13 , 14 give theory prediction ( no fitting parameters ! ) in units @xmath1137 . it should be mentioned that in our papers @xcite and @xcite we put @xmath403=0.09 according with the instruction given in paper @xcite . but the multiphoton effects were not included . however , as was stated above one has to use @xmath403=0.096 . so , from the one side for the gold target with the thickness @xmath1138 and energy @xmath291=25 gev the reduction factor @xmath1139 for photons with energy @xmath1140 mev ( plateau region ) . from the other side , use of the coefficient @xmath403 = 0.096 in ( [ d2 ] ) instead of @xmath403 = 0.09 lowers data upon @xmath269 and this is imitate inclusion of the reduction factor . as a result , the excellent agreement of the theory and data noted in @xcite is not broken . it is seen that in fig . 13(a ) there is a perfect agreement of the theory and data . in fig . 13(b ) there is an overall difference : data is order of @xmath982 higher than theory curve . for photon energy @xmath1090 the theory coincides with bethe - maximon formula ( with the coulomb corrections ) applicable for this energy . note that just for this case it was similar problem with normalization of data matching with the migdal monte carlo simulation ( @xmath1141 , see table ii in @xcite ) . for thickness @xmath1142 there is a qualitative difference between our theory prediction and monte carlo simulation in @xcite . there was a number of experimental uncertainties associated with this target . nevertheless , we show data for @xmath1086 which are lying higher than theory curve . we consider now the data @xcite from the point of view of the above analysis @xcite . energy loss spectra were measured in iridium target of 0.128 mm thickness ( 4.36 % @xmath46 ) . basing on eq.([d1 ] ) and following analysis we see that this is a thick target . since the photons with the energy @xmath1143gev were measured , no boundary effects were observed . for ir the characteristic electron energy eq.([2.30b ] ) @xmath237=2.27 tev and the characteristic photon energy eq.([2.31b ] ) @xmath1144 for which the lpm effect well manifests itself is @xmath824(287 gev)=32 gev . the result of calculation for the initial electron energy @xmath291=287 gev is shown in fig . the curves 1 , 2 , 3 are calculated using eqs.([2.35 ] ) , ( [ 2.30 ] ) , ( [ 2.25 ] ) correspondingly . it should be mentioned that the prediction of our theory ( curve 4 ) in the hard end of spectrum coincides with the bethe - maximon curve within the accuracy better than @xmath1145 . for used thickness of target the multi - photon effects are very essential . in this case it is convenient to calculate the reduction factor @xmath866 using eq.([y3.5 ] ) , where the mentioned spectrum eq.([2.25 ] ) is used . the result obtained is in a good agreement with eqs.([y12 ] ) , ( [ y26 ] ) . the reduction factor for this energy is presented by curve 3 in fig.18 . the final prediction with the reduction factor taken into account for the used target thickness is presented by the curve @xmath340 . the data are recalculated according with eqs.([d2 ] ) , ( [ d3 ] ) using the coefficient @xmath1146 . the result of calculation for the initial electron energy @xmath291=207 gev is shown in fig . 16 and the result of calculation for the initial electron energy @xmath291=149 gev is shown in fig . the reduction factors for these energies are presented by the curves 2 and 1 in fig.18 correspondingly . from fig.18 it follows that if the electron energy decreases the reduction factor @xmath866 diminishes in agreement with eqs.([y12 ] ) , ( [ y26 ] ) . it is seen that for energy 287 gev there is the quite satisfactory agreement of theory with data , for energy 207 gev the agreement is somewhat less satisfactory , and for energy 149 gev data are 10 - 15% below the theory . the difference is the same in hard part of spectrum ( @xmath1147 40 gev ) where the bethe - maximon intensity spectrum is valid . the decrease of relative energy losses of electron @xmath1148 in consequence of the lpm effect can be important in electromagnetic calorimeters operating in detectors on colliders in tev range @xcite . this item was discussed in sec.2.2 using the main ( migdal ) term , @xmath279 in gold is given in fig . 4 ( curve 1 ) . an analysis with better accuracy is of evident interest . the relative energy loss of electron per unit time including the contribution of the correction term eq.([2.25 ] ) in terms of the bethe - maximon radiation length @xmath46 eq.([2.2i ] ) for iridium ( @xmath237=2.27 tev ) curve 1 and for lead ( @xmath237=4.385 tev ) curve 2 is shown in fig.19 . the increase of effective radiation length is 13% in ir and 7.5% in pb for the electron energy 1 tev , is 38% in ir and 24.5% in pb for the electron energy 5 tev , and is 58% in ir and 38.5% in pb for the electron energy 10 tev . for both media the corrections attain the maximal value of order 5% at energy @xmath1149 in this section we consider the radiation at head - on collision of high energy electron and positron beams . the properties of photon emission process from a particle are immediately connected with details of its motion . it is convenient to consider the motion and radiation from particles of one beam in the rest frame of other beam ( the target beam ) . in this case the target beam is an ensemble of the coulomb centers . the radiation takes place at scattering of a particle from these centers . if the target consists of neutral particles forming an amorphous medium , a velocity of particle changes ( in a random way ) only at small impact distances because of screening . in the radiation theory just the random collisions are the mechanism which leads to the incoherent radiation . for colliding beams significant contributions into radiation give the large impact parameters ( very small momentum transfers ) due to the long - range character of the coulomb forces . as a result , in the interaction volume , which is determined also by the formation length @xmath66 eqs.([5 ] ) , ( [ 6 ] ) ( in the longitudinal direction ) , it may be large number of target particles . let us note that in the case when the contribution into the radiation is given by impact parameters comparable with the transverse size of target , the number of particles in the interaction volume is determined by the ratio of the radiation formation length to the mean longitudinal distance between particles . however , not all cases of momentum transfer should be interpreted as a result of random collisions . one have to exclude the collisions , which are macroscopic certain events . for elaboration of such exclusion we present the exact microscopic momentum transfer to the target particle in the form : @xmath1150 . here @xmath1151 is the mean value of momentum transfer calculated according to standard macroscopic electrodynamics rules with averaging over domains containing many particles . the longitudinal size of these domains should be large with respect to longitudinal distances between target particles and simultaneously small with respect to the radiation formation length . the motion of particle in the averaged potential of target beam , which corresponds to the momentum transfer @xmath1152 , determines the coherent radiation ( _ the beamstrahlung _ ) . while the term @xmath1153 describes the random collisions which define the process of incoherent radiation ( _ the bremsstrahlung _ ) . such random collisions we will call `` scattering '' since @xmath1154 . one of principal characteristics of particle motion defining the properties of coherent radiation is the ratio of variation of its transverse momentum to the mass during the whole time of passage across the opposite beam @xmath340 @xmath1155 where @xmath1156 is the number of particles in the opposite beam , @xmath1157 and @xmath1158 is its transverse dimensions ( @xmath1159 ) , @xmath1160 is the longitudinal size of opposite beam . the dispersion of particle momentum during time @xmath340 is small comparing with @xmath152 . it attains the maximum for the coaxial beams : @xmath1161 here @xmath1162 is the square of mean angle of multiple scattering , @xmath1163 is the characteristic logarithm of scattering problem ( @xmath1164 ) . this inequality permits one to use the perturbation theory for consideration of bremsstrahlung , and to analyze the beamstrahlung independently from the bremsstrahlung ] ] . another important characteristics of motion is the relative variation of particle impact parameter during time @xmath340 @xmath1165 here @xmath1166 is @xmath2 or @xmath509 . when the disruption parameter @xmath1167 , the collision does nt change the beam configuration and the particle crosses the opposite beam on the fixed impact parameter . if in addition the parameter @xmath1168 ( this situation is realized in colliders with relatively low energies ) then the beamstrahlung process can be calculated using the dipole approximation . the main contribution into the beamstrahlung give soft photons with an energy @xmath1169 in the opposite case @xmath1170 the main part of beamstrahlung is formed when the angle of deflection of particle velocity is of the order of characteristic radiation angle @xmath359 and the radiation formation length @xmath1171 is defined by @xmath1172 and the characteristic photon energy is @xmath1173 here @xmath95 is the invariant parameter eq.([14 ] ) which defines properties of magnetic bremsstrahlung in the constant field approximation ( cfa ) . for applicability of cfa it is necessary that relative variation of @xmath408 was small on the radiation formation length @xmath1171 . as far @xmath1171 is shorter than @xmath1160 in @xmath1170 times the characteristic parameter becomes @xmath1174 to that extent . the condition @xmath1175 is fulfilled in all known cases . the mean number of photons emitted by a particle during the whole time of passage across the opposite beam @xmath340 is @xmath1176 , it include the electromagnetic interaction constant . using the estimate ( [ c2.6 ] ) we get an estimate of relative energy loss @xmath1177 in the case @xmath1178 ( this condition is satisfied in all existing facilities and proposed collider projects ) the soft photons with energy @xmath1179 are mainly emitted . for @xmath1180 the emission probability is exponentially suppressed . so , such photons are emitted in the bremsstrahlung process only . the boundary photon energy @xmath463 , starting from which the bremsstrahlung process dominates , depends on particular parameters of facility . if @xmath1181 the energy is @xmath1182 . the formation length for @xmath1183 is much shorter than @xmath1171 . on this length the particle deflection angle is small comparing with @xmath359 and one can neglect the variation of transverse beam dimensions ( see eq.([c2.8 ] ) ) . this means that all calculations of bremsstrahlung characteristics can be carried out in adiabatic approximation using local beam characteristics @xmath1184 etc , with subsequent averaging of radiation characteristics over time . note that actually we performed a covariant analysis and the characteristic parameters are defined in a laboratory frame . as an example we consider the situation when the configuration of beams does nt change during the beam collision ( the disruption parameter @xmath1185 ) , and the total particle deflection angle during intersection of whole beam is small comparing with the characteristic radiation angle @xmath359 ( the dipole case ) . the target beam in its rest frame is the ensemble of classical potentials centers with coordinates @xmath1186 and the transverse coordinate of emitting particle is @xmath1187 . in the perturbation theory the total matrix element of the radiation process can be written as @xmath1188 we represent the combination @xmath1189 in the form @xmath1190 in the expression eq.([c1.2 ] ) we have to carry out averaging over position of scattering centers . we will proceed under assumption that there are many scattering centers within the radiation formation length @xmath1191 @xmath1192 where for the gaussian distribution @xmath1193 here @xmath1156 and @xmath1160 are introduced in eq.([c2.1 ] ) . note that in the situation under consideration @xmath1194 , where @xmath1195 is the characteristic transverse size of target beam . let us select terms with approximately fixed phase @xmath1196 in the sum with @xmath1197 in eq.([c1.2 ] ) . if the condition ( [ c1.3 ] ) is fulfilled , there are many terms for which the phase variation is small ( @xmath1198 ) . for this reason one can average over the transverse coordinates ( @xmath1199 ) of target particles in eq.([c1.2 ] ) without touching upon the longitudinal coordinates ( @xmath1200 ) @xmath1201 where @xmath1202 here @xmath1203 is the probability density of target particle distribution over the transverse coordinates normalized to unity . in eq.([c1.5 ] ) in the sum with @xmath1204 we add and subtract the terms with @xmath1205 . the first term ( proportional to @xmath1156 ) on the right - hand side of eq.([c1.5 ] ) is the incoherent contribution to radiation ( _ the bremsstrahlung _ ) . the second term gives the coherent part of radiation . for gaussian distribution eq.([c1.4 ] ) performing averaging over the longitudinal coordinate @xmath1206 one has @xmath1207 it was pointed in sec.1.3 that the external factors act differently on the radiating particle and on the recoil particle . this leads to a specific effect in the bremsstrahlung process at electron - electron(positron ) collision . the point is that in the rest frame of recoil particle the formation length is longer by @xmath1208 times than in the laboratory frame . just this length determines the maximal impact parameters at scattering ( e.g. the screening radius ) . the are a few factors which could act on the recoil electron . one of them is the presence of an external magnetic field in the region of particle collision @xcite ( see sec.1.3 ) . another effect can appear due to smallness of the linear interval @xmath383 where the collision occurs in comparison with @xmath1209 eq.([6 ] ) . this was pointed in @xcite . the special experimental study of bremsstrahlung was performed at the electron - positron colliding beam facility vepp-4 of institute of nuclear physics , novosibirsk @xcite . the deviation of the bremsstrahlung spectrum from the standard qed spectrum was observed at the electron energy @xmath1210gev . the effect was attributed to the smallness of the transverse size of the colliding beams . in theory the problem was investigated in @xcite , where the bremsstrahlung spectrum at the collision of electron - electron(positron ) beams with the small transverse size was calculated to within the power accuracy ( the neglected terms are of the order @xmath1211 ) . after the problem was analyzed in @xcite , and later on in @xcite where some of results for the bremsstrahlung found in @xcite were reproduced . it should be noted that in @xcite ( as well as in all other papers mentioned above ) an incomplete expression for the bremsstrahlung spectrum was calculated . one has to perform the subtraction associated with the extraction of pure fluctuation process as it was shown above . an analogous problem encountered in the analysis of incoherent processes in the oriented crystals @xcite,@xcite where the necessity of the subtraction procedure was indicated . without the subtraction the results for the incoherent processes in oriented crystals would be qualitatively erroneous . the correction to photon emission probability due to the small transverse dimensions of colliding beam for unpolarized electrons and photon was calculated in @xcite basing on subtraction procedure as in eq.([c1.5 ] ) . it is obtained after integration over the azimuth angle of the emitted photon @xmath1212 where @xmath89 is the photon emission angle , @xmath8 , @xmath1213\varrho \frac{d\phi(\mbox{\boldmath$\varrho$})}{d\varrho}d^2\varrho \nonumber \\ \hspace{-7 mm } & & f^{(2)}(\omega , \zeta)=\frac{2\eta^2}{\zeta^2}\int \left(\int k_1(\eta \varrho)\frac{\mbox{\boldmath$\varrho$}}{\varrho}w_c({\bf x}-\mbox{\boldmath$\varrho$})d^2\varrho \right)^2w_r({\bf x})d^2x , \label{c3.12}\end{aligned}\ ] ] here @xmath1214 where @xmath1215 is defined in ( [ c1.6 ] ) , @xmath1216 is the same but for the radiating beam , value @xmath546 is defined in c.m.frame of colliding particles . the term @xmath1217 is the subtraction term . the total probability is @xmath1218 , where @xmath1219 is standard qed probability . the analysis in @xcite was based on eqs.([c3.11])-([c3.12 ] ) . we considered in @xcite the actual case of the gaussian beams . the fourier transform was used @xmath1220,~ w_c({\bf q})=\exp\left[-\frac{1}{2}(q_x^2\sigma_x^2+q_y^2\sigma_y^2)\right ] , \label{c4.1}\end{aligned}\ ] ] where as above the index @xmath264 relates to the radiating beam and the index @xmath1221 relates to the target beam , @xmath1222 and @xmath1223 ( @xmath1158 and @xmath1157 ) are the vertical and horizontal transverse dimensions of radiating ( target ) beam . substituting eq.([c4.1 ] ) into eq.([c3.15 ] ) we find @xmath1224;~ \sigma_x^2=\frac{1}{2(\sigma_x^2+\delta_x^2)},~ \sigma_y^2=\frac{1}{2(\sigma_y^2+\delta_y^2)}. \label{c4.2}\ ] ] using the relation @xmath1225 the following expression for the correction to spectrum was found in @xcite starting from eq.([c3.11 ] ) @xmath1226 where ei(@xmath2 ) is the exponential integral function and erfc(@xmath2 ) is the error function . this formula is quite convenient for the numerical calculations . the subtraction term ( @xmath1217 in eq.([c3.12 ] ) ) gives for coaxial beams @xmath1227 where @xmath1228 \label{c4.20}\end{aligned}\ ] ] here the function @xmath1229 is : @xmath1230 . \label{c4.21}\ ] ] in eq.([c4.20 ] ) we introduced the following notations @xmath1231 in the case of narrow beams one has @xmath1232 . in this case of coaxial beams @xmath1233 is @xmath1234+\frac{2}{9}\bigg\ } , \label{c5.7}\ ] ] where @xmath1235 the dimensions of beams in the experiment @xcite were @xmath1236 , so this is the case of flat beams . the estimate for this case gives @xmath1237 . this term is much smaller than other terms in eq.([c5.7 ] ) . this means that for this case the correction to the spectrum calculated in @xcite is very small . the result of calculation and vepp-4 ( inp , novosibirsk ) data are presented in fig.20 where the bremsstrahlung intensity spectrum @xmath1238 is given in units @xmath1239 versus the photon energy in units of initial electron energy ( @xmath275)@xcite . the upper curve is the standard qed spectrum , the three close curves below are calculated using eqs.([c4.9 ] ) and ( [ c4.19 ] ) for the different vertical dimensions of colliding beams ( equal for both colliding beams @xmath1240):@xmath1241 ( bottom ) , @xmath1242 ( middle ) , @xmath1243 ( top ) ( this is just the 1-sigma dispersion for the beams used in the experiment ) . all the theoretical curves are calculated to within the relativistic accuracy ( the discarded terms are of the order @xmath1244 ) . it is seen that the effect of the small transverse dimensions is very essential in soft part of spectrum ( at @xmath1245 the spectral curve diminishes in 25 % ) , while for @xmath1246 the effect becomes negligible . the data measured in @xcite are presented as circles ( experiment in 1980 ) and as triangles ( experiment in 1981 ) with 6 % systematic error as obtained in @xcite ( while the statistical errors are negligible ) . this presentation is somewhat different from @xcite . it is seen that the data points are situated systematically below the theory curves but the difference is not exceed the 2-sigma level @xcite . it should be noted that this is true also in the hard part of spectrum where the beam - size effect is very small . one more measurement of beam - size effect was performed at the hera electron - proton collider ( desy , germany ) @xcite . the electron beam energy was @xmath291=27.5 gev , the proton beam energy was @xmath1247=820 gev . the parameters of beam in this experiment were ( in our notation ) : @xmath1248 , @xmath1249 . since the ratio of the vertical and the horizontal dimensions is not very small , the contribution of subtraction term ( eq.([c4.19 ] ) ) is essential ( more than 10% ) . the bremsstrahlung intensity spectrum @xmath1238 in units @xmath1239 versus the photon energy in the units of initial electron energy ( @xmath275 ) for the hera experiment is given in fig.21 . for details of comparison of experimental data @xcite , @xcite with theory see @xcite , where we discussed also possible use of beam - size effect for linear collider tuning . it should be noted that for linear collider the condition of strong beam - size effect @xmath1250 is fulfilled for the whole spectrum . this can be seen in fig.22 , where the lower curve is calculated using eq.([c4.9 ] ) and the subtraction term is very small since @xmath1251 . as far as the narrow beams are considered in fig.22 , the lower curve is consistent also with eq.([c5.7 ] ) . this curve depends on the energy and the transverse sizes of beams . it will be instructive to remind that the analysis in @xcite ( see eq.(2.8 ) ) and here is valid if @xmath1252 ( see eq.([c2.6 ] ) , @xmath346 ) . the parameter @xmath1253 depends also on number of particles @xmath1156 and the longitudinal beam size . so , for low @xmath1156 fig.22 is valid for any @xmath2 , but for tesla project ( @xmath1253=0.13 ) it holds in hard part of spectrum only . in fact , the probability of incoherent radiation becomes larger than the probability of coherent radiation only at @xmath1254 where the lower curve in fig.22 is certainly applicable . the particle interaction at beam - beam collision in linear colliders occurs in an electromagnetic field provided by the beams . as a result , 1)the phenomena induced by this field turns out to be very essential , 2)the cross section of the main qed processes are modified comparing to the case of free particles . these items were considered by v.m.strahkhovenko and authors @xcite . the magnetic bremsstrahlung mechanism dominates and its characteristics are determined by the value of the quantum parameter @xmath1255 eq.([14 ] ) dependent on the strength of the incoming beam field at the moment @xmath11 ( the constant field limit(cfa ) ) . in the cfa the spectral probability of radiation from an electron per unit time has the form @xcite , eq.(10.25 ) ( see also eq.(4.24 ) in @xcite or eq.(90.23 ) in @xcite ) @xmath1256 where @xmath1257 is the macdonald functions , @xmath1258 , @xmath346 . the corresponding probability for polarized electrons and photons is given in @xcite . for the gaussian beams @xmath1259 here the function @xmath1260 depends on transverse coordinates . it turns out that for the gaussian beams the integration of the spectral probability over time can be carried out in a general form : @xmath1261 , \label{co2.4}\end{aligned}\ ] ] where @xmath1262 . in the case when @xmath1263 the main contribution into the integral ( [ co2.4 ] ) gives the region @xmath1264 . taking the integrals over @xmath1265 we obtain @xmath1266 for round beams the integration over transverse coordinates is performed with the density @xmath1267 the parameter @xmath1268 we present in the form @xmath1269 where @xmath1156 is the number of electron in the bunch . the laplace integration of eq.([co2.5 ] ) gives for radiation intensity @xmath1270 @xmath1271 where @xmath1272 . integration of eq.([co2.4 ] ) over transverse coordinates gives the final result for the radiation intensity . for the round beams it is shown in fig.23 for @xmath1273 , the curve attains the maximum at @xmath1274 . the right slope of the curve agrees with the asymptotic intensity ( [ co2.8 ] ) and the left slope of the curve agrees with the standard classical intensity . @xmath1275 it will be instructive to compare the spectrum in fig.23 , found by means of integration over the transverse coordinates with intensity spectrum which follows from eq.([co2.4 ] ) ( multiplied by @xmath7 ) with averaged over the density eq.([co2.6 ] ) value @xmath1276 eq.([co2.7 ] ) : @xmath1277 . the last spectrum reproduces the spectrum given in fig.23 , in the interval @xmath1278 with an accuracy better than 2% ( near maximum better than 1% ) while for @xmath1279 one can use the classic intensity ( [ co2.8a ] ) and for @xmath1280 the asymptotics ( [ co2.8 ] ) is applicable . for the flat beams ( @xmath1281 ) the parameter @xmath1282 takes the form @xmath1283,\quad \chi_m= \frac{2n_c\alpha\gamma \lambda_c^2}{\sigma_z\sigma_x } , \label{co2.9}\ ] ] here @xmath1284 , @xmath1285 , @xmath1286 and @xmath1287 are the unit vectors along the corresponding axes . the formula ( [ co2.9 ] ) is consistent with given in @xcite . in @xcite the term with @xmath1286 was missed . because of this the numerical coefficients in results for the flat beams are erroneous . to calculate the asymptotics of radiation intensity for the case @xmath1288 one has to substitute @xmath1289^{1/2 } \label{co2.10}\ ] ] into eq.([co2.5 ] ) and take integrals over transverse coordinates @xmath1290 with the weight @xmath1291 integral over @xmath2 can be taken using the laplace method , while for integration over @xmath509 it is convenient to introduce the variable @xmath1292 as a result we obtain for the radiation intensity in the case of flat beams @xmath1293 it is interesting to compare the high - energy end of intensity spectrum at collision of flat beams eq.([co2.13 ] ) with intensity spectrum of incoherent radiation with regard for smallness of the transverse beam sizes considered above . for calculation we use the project tesla parameters @xcite : @xmath1294 gev , @xmath1295 nm , @xmath1296 nm , @xmath1297 mm , @xmath1298 , @xmath1253=0.13 . the result is shown in fig.24 , where the curves calculated according to eq.([co2.13 ] ) , and eq.([c5.7 ] ) . it is seen that for @xmath1299 the incoherent radiation dominates . for @xmath1254 the incoherent radiation may be used for a tuning of beams @xcite . along with the spectral characteristics of radiation the total number of photons emitted by an electron is of evident interest as well as the relative energy losses . we discuss an actual case of flat beams and the parameter @xmath1178 . in this case one can use the classic expression for intensity ( bearing in mind that starting from @xmath1181 the quantum effects become substantial ) . in classical limit the relative energy losses are @xmath1300 using eqs.([co2.3 ] ) , ( [ co2.10 ] ) , and ( [ co2.11 ] ) we get @xmath1301 for mean number of emitted by an electron photons we find @xmath1302 if @xmath1303 one have to use the quantum formulas . for the energy losses one can use the approximate expression ( the accuracy is better than 2% for any @xmath95 ) @xcite @xmath1304^{-2/3}. \label{co2.17}\ ] ] here @xmath95 is the local value , so this expression for @xmath1305 should be integrated over time and averaged over the transverse coordinates . for mean number of photons emitted by an electron there is the approximate expression ( the accuracy is better than 1% for any @xmath1276 ) @xmath1306^{1/6 } } \label{co2.18},\ ] ] where the expression should be averaged over the transverse coordinates . for the project tesla one gets for @xmath1307 according to eq.([co2.15 ] ) , while the correct result from eq.([co2.17 ] ) is @xmath1308 . for mean number of photons emitted by an electron we have correspondingly @xmath1309 while the correct result is @xmath1310 . the probability of pair creation by a photon in the external field can be find from formulas ( [ co1 ] ) using the substitution rule ( see e.g.@xcite , @xcite ) : @xmath1311 . performing these substitutions we obtain ( see also eq.(3.50 ) in @xcite ) @xmath1312 here @xmath291 is the energy of the created electron , @xmath1313 is the energy of created positron . there are different mechanisms of electron - positron pair creation 1 . real photon radiation in the field and pair creation by this photon in the same field of the opposite beam . this mechanism dominates in the case @xmath1314 . 2 . direct electroproduction of electron - positron pair in the field through virtual photon . this mechanism is also essential in the case @xmath1314 . mixed mechanism(1):photon is radiated in the bremsstrahlung process , i.e. incoherently , and the pair is produced by photon in an external field . mixed mechanism(2 ) : photon is radiated via the beamstrahlung mechanism , and pair is produced in the interaction of this photon with individual particles of oncoming beam , i.e. in interaction with potential fluctuations . incoherent electroproduction of pair . in the actual case @xmath96 mixed and incoherent mechanisms mostly contribute . we start with the mixed mechanism ( 2 ) . in the case @xmath1178 the parameter @xmath695 eq.([co4 ] ) containing the energy of emitted photon is also small and the incoherent cross section of pair creation by a photon is weakly dependent on the photon energy . to logarithmic accuracy @xmath1315 if the term @xmath1316 being neglected , the pair creation probability is factorized ( we discuss coaxial beams of identical configuration ) . the number of pairs created by this mechanism ( per one initial electron ) is @xmath1317^{1/2 } dw dv \nonumber \\ \hspace{-7mm}&&=0.1167\alpha^3 n r \chi_m \frac{\lambda_c^2}{\sigma_x\sigma_y } \ln \frac{\sigma_y}{\lambda_c}. \label{co3.2}\end{aligned}\ ] ] so for the project tesla parameters the total number of produced pairs by this mechanism is @xmath1318 per bunch in one collision . now we turn over to discussion of incoherent electroproduction of pairs when both intermediate photons are virtual . to within the logarithmic accuracy for any @xmath95 and @xmath695 one can use the method of equivalent photons @xmath1319 where @xmath1320 taking into account that the cross section of pair photoproduction is @xmath1321 we obtain in the main logarithmic approximation for the cross section of the pair electroproduction @xmath1322 where @xmath1323 . if we put @xmath1324 we obtain the standard landau - lifshitz cross section @xmath1325 ( see e.g. @xcite , @xcite ) @xmath1326 with regard for the bounded transverse dimensions of beam and influence of an external field the equivalent photon spectrum changes substantially . for @xmath1327 we have @xmath1328 and under condition @xmath1329 we find @xmath1330 for tesla parameters @xmath1331 then @xmath1332 this cross section is three times smaller than standard @xmath1325 . for the project tesla parameters the number of pairs produced by this mechanism is @xmath1333 in one direction per bunch in one collision , @xmath1334 is the geometrical luminosity per bunch . so the both discussed mechanisms give nearly the same contribution for this project . it should be noted that the above analysis was performed under assumption that the configuration of beams does nt changed during collision , although in the tesla project the disruption parameter @xmath1335 . so we see that the beam - size effect and an external field effects are very essential for pair creation mechanism also . * acknowledgments * the authors are indebted to the russian foundation for basic research supported in part this research by grants 00 - 02 - 18007 , 03 - 02 - 16154 . we would like to thank s. klein for discussions and u.uggerhoj for information about experiment and data . the quasiclassical operator method developed by authors is adequate for consideration of the electromagnetic processes at high energy . the probability of the radiation process has a form ( see @xcite , p.63 , eq.(2.27 ) ) ; the method is given also in @xcite , for classical limit see @xcite @xmath1336 , \label{a1}\ ] ] where @xmath1337 is the 4-momentum of the radiated photon , @xmath1338 , @xmath1339 , @xmath11 is the time , and @xmath10 is the particle location on a classical trajectory , @xmath1340 , @xmath8 , let us remind that we employ units such that @xmath4 . the matrix element @xmath1341 is defined by the structure of a current . for an electron ( spin 1/2 particle ) one has @xmath1342 \simeq \frac{1}{2}\left(1+\frac{\varepsilon}{\varepsilon ' } \right ) { \bf e}^{\ast}\mbox{\boldmath$\vartheta$ } , \nonumber \\ & & { \bf b(t)}=\frac{1}{2\sqrt{\varepsilon \varepsilon ' } } \left[\sqrt{\frac{\varepsilon'+m}{\varepsilon+m } } \left({\bf e}^{\ast } \times { \bf p}(t ) \right ) - \sqrt{\frac{\varepsilon+m}{\varepsilon'+m } } \left({\bf e}^{\ast } \times ( { \bf p}(t)-\hbar { \bf k } ) \right)\right ] \nonumber \\ & & \simeq \frac{\hbar \omega}{2\varepsilon'}\left({\bf e}^{\ast } \times \left(\frac{{\bf n}}{\gamma}- \mbox{\boldmath$\vartheta$}\right ) \right ) , \label{a2}\end{aligned}\ ] ] here @xmath1343 is the initial and final momentum of an electron , @xmath434 is the vector of the polarization of a photon ( the coulomb gauge is used),the four - component spinors @xmath1344 and the two - component spinors @xmath1345 describe the initial ( @xmath1346 ) and final ( @xmath1347 ) polarization of the electron , * v=v(t ) is the electron velocity , @xmath1348 , @xmath1349 is the component of particle velocity perpendicular to the vector @xmath1350 , @xmath1351 is the lorentz factor . the final expressions in ( [ a2 ] ) are given for radiation of ultrarelativistic electrons , they are written down with relativistic accuracy ( terms @xmath659 are neglected ) and in the small angle approximation . if we are not interested in the initial and final particles polarization , then @xmath1352 = \frac{1}{2 } { \cal l}(\mbox{\boldmath$\vartheta$}',\mbox{\boldmath$\vartheta$ } ) , \label{a3}\ ] ] where we have used the notation @xmath1353 . for polarized initial electron we have @xmath1354 where @xmath347 is the vector describing the initial polarization of the electron ( in its rest frame ) , @xmath1355 and @xmath1356 . if the particle moves along a definite trajectory , then substituting the classical values @xmath10 and @xmath1357 in ( [ a1])-([a3 ] ) , we obtain the desired probability for this process . when there is scattering , eq.([a1 ] ) must be averaged over all possible particle trajectories . this operation is performed with the aid of the distribution function , averaged over atomic positions in the scattering medium and satisfying the kinetic equation . we consider here general case with the external field ( particle acceleration ) taken into account . the emission probability per unit time is then ( in this appendix we follow the paper @xcite ) @xmath1358 the distribution function @xmath1359 in ( [ a4 ] ) satisfies the kinetic equation @xmath1360d^3v ' , \label{a5}\end{aligned}\ ] ] where @xmath586 is the number density of atoms in the medium , and @xmath1361 is the scattering cross section . the normalization condition @xmath1362 should be also satisfied , as well as the initial condition for @xmath1363 : @xmath1364 in eq.([a4 ] ) , we integrate over @xmath1365 , taking advantage of the fact that @xmath1366 can only depend on the coordinate difference @xmath1367 : @xmath1368 where @xmath1369 we can put @xmath1370 if we take into account that in ultrarelativistic limit @xmath1371 , @xmath1372 , and the scattering cross section is @xmath1373 making use of these results and eq.([a5 ] ) for @xmath1363 , we obtain the following equation for @xmath1374 @xmath1375 , \label{a11}\end{aligned}\ ] ] where for scattering in the screened coulomb potential the cross section in the born approximation is @xmath1376 here @xmath1377 , @xmath173 is the screening radius ( @xmath1378 , @xmath174 is the bohr radius ) . a similar equation for the @xmath1379-dependence of @xmath1380 can be obtained by letting @xmath1381 and @xmath1382 . if the final state of the charged particle is of no interest , the probability ( [ a7 ] ) must be integrated over @xmath1383 . the resulting emission probability per unit time , normalized to a single particle moving at a speed @xmath100 is then @xmath1384 , \label{a13}\ ] ] where @xmath1385 , and @xmath1386 satisfies the equation @xmath1387 , \label{a14}\end{aligned}\ ] ] here @xmath1388 . the initial conditions for @xmath1389 are @xmath1390 if the scattering cross section @xmath1391 depends solely on the angle difference @xmath1392 , as eq.([a12 ] ) , then eq.([a14 ] ) is most conveniently solved by fourier transforming with respect to the variable @xmath1379 : @xmath1393 from eq.([a14 ] ) , with the initial conditions ( [ a15 ] ) , we have @xmath1394 \varphi_{\mu}({\bf x } , \tau),\quad \sigma({\bf x})= \int_{}^ { } d^2\vartheta \exp ( i\mbox{\boldmath$\vartheta$ } { \bf x } ) \sigma(\mbox{\boldmath$\vartheta$ } ) ; \nonumber \\ & & \varphi_{0}({\bf x } , 0)= \delta({\bf x}),\quad \mbox{\boldmath$\varphi$}({\bf x } , 0 ) = -i\mbox{\boldmath$\nabla$}\delta({\bf x } ) \label{a17}\end{aligned}\ ] ] if the angular distribution of the radiation is of no interest , eq.([a13 ] ) must be integrated over the photon emission angles @xmath1379 . bearing in mind that @xmath1395 , and that @xmath1396 we obtain the following expression for the spectral distribution of emission probability per unit time @xmath1397 . \label{a19}\ ] ] the equation like ( [ a17 ] ) in the classical limit and with @xmath1398 was discussed in @xcite , appendix v. for the cross section ( [ a12 ] ) we have for @xmath1399 ( [ a17 ] ) @xmath1400 where @xmath1401 is the modified bessel function . bearing in mind that contributions to the cross section come from @xmath1402 , and expanding @xmath1401 as a power series in @xmath1403 we obtain the following equation for @xmath1404 from eq.([a17 ] ) , to power - series accuracy : @xmath1405 where @xmath1406 is the euler constant . the `` potential '' @xmath1407 is calculated in the born approximation since eq.([a12 ] ) was used . the coulomb correction to the potential @xmath1407 were calculated in appendix a of @xcite . the eikonal approximation was use which is valid for arbitrary value @xmath144 . substituting this correction we have @xmath1408 where the function @xmath178 is defined in eq.([2.9b ] ) . in absence of radiation and an external field ( b=w=0 ) eqs.([a21 ] ) and ( [ a16 ] ) at the initial condition @xmath1409 give the solution of pure scattering problem ( the initial condition for the distribution function @xmath1410 is @xmath1411 ) : @xmath1412\vartheta d\vartheta \label{a23}\end{aligned}\ ] ] for another form of solution see eq.([82b ] ) . l. d. landau and i. ya . pomeranchuk , dokl.akad.nauk sssr 92 * ( 1953 ) 535 , 735 . see in english in _ the collected papers of l. d. landau _ , pergamon press , 1965 . a. b. migdal , phys . rev . * 103 * ( 1956 ) 1811 . a. b. migdal , sov . jetp * 5 * ( 1957 ) 527 . m. l. ter - mikaelian , _ high energy electromagnetic processes in condensed media _ , john wiley & sons , 1972 . e. l. feinberg and i. ya . pomeranchuk , nuovo cimento , supplement to vol . * 3 * ( 1956 ) 652 . v. m. galitsky and i. i. gurevich , nuovo cimento * 32 * ( 1964 ) 396 . v. n. baier and v. m. katkov , sov . jetp , * 26 * ( 1968 ) 854 ; * 28 * ( 1969 ) 807 . v. n. baier , v. m. katkov and v. s. fadin , _ radiation from relativistic electrons _ ( in russian ) atomizdat , moscow , 1973 . v. n. baier , v. m. katkov and v. m. strakhovenko , sov . jetp * 67 * ( 1988 ) 70 . v. n. baier , v. m. katkov and v. m. strakhovenko , _ electromagnetic processes at high energies in oriented single crystals _ , world scientific publishing co , singapore , 1998 . v. b. berestetskii , e. m. lifshitz and l. p. pitaevskii , _ quantum electrodynamics _ pergamon press , oxford , 1982 . p. l. anthony , r. becker - szendy , p. e. bosted _ et al _ , phys . * 75 * ( 1995 ) 1949 . p. l. anthony , r. becker - szendy , p. e. bosted _ et al _ , phys . * 76 * ( 1996 ) 3550 . p. l. anthony , r. becker - szendy , p. e. bosted _ et al _ , phys.rev.d56 ( 1997 ) 1373 . s. klein , rev . mod . phys . * 71 * ( 1999 ) 1501 . h. d. hansen , u. i. uggerhoj , c.c.biino _ et al _ , phys.rev.lett . * 91 * ( 2003 ) 014801 . v. n. baier and v. m. katkov , phys.rev . * d57 * ( 1998 ) 3146 . v. n. baier and v. m. katkov , _ quantum aspects of beam physics _ , ed.p . chen , world scientific pc , singapore , 1998 , p.525 . v. n. baier and v. m. katkov , phys.rev . * d59 * ( 1999 ) 056003 . v. n. baier and v. m. katkov , phys.rev . * d60 * ( 1999 ) 076001 . v. n. baier and v. m. katkov , phys.rev . * d62 * ( 2000 ) 036008 . r. blancenbeckler and s. d. drell , phys.rev . * d53 * ( 1996 ) 6265 . b. g. zakharov , pisma v zhetf * 63 * ( 1996 ) 906 . b. g. zakharov , pisma v zhetf * 64 * ( 1996 ) 781 . r. baier , yu . l. dokshitzer , a. h. mueller , s. peigne , and d. schiff , nucl b478 * ( 1996 ) 577 . v. n. baier and v. m. katkov , phys.rev . * d63 * ( 2001 ) 116008 . v. n. baier , and v. m. katkov , sov . * 17 * ( 1973 ) 1068 . v. m. katkov , and v. m. strakhovenko , sov . j. nucl.phys . * 25 * ( 1977 ) 660 . j. schwinger , phys.rev . * 73 * ( 1948 ) 416 ; * 76 * ( 1949 ) 790 . v. n. baier , v. m. katkov and v. m. strakhovenko , lett . nuovo cimento * 15 * ( 1976 ) 149 ; sov . * 24 * ( 1976 ) 197 . v. n. baier , and v. m. katkov , phys.lett . a 280 ( 2001 ) 275 . v. n. baier , and v. m. katkov , phys.lett . a 286 ( 2001 ) 299 . v. n. baier , v. m. katkov and v. m. strakhovenko , sov . doklady * 16 * ( 1971 ) 230 . v. i. ritus , sov . jetp * 30 * ( 1970 ) 1181 . j. schwinger , phys.rev . * 75 * ( 1949 ) 1912 ; * 82 * ( 1951 ) 664 . n. b. narozhny , zh . eksp.teor.fiz . , * 55 * ( 1968 ) 714 . a. i. milstein and m. schumacher , phys . rep.c243 ( 1994 ) 183 . m. cheng and t. t. wu , phys.rev . * 182 * ( 1969 ) 1873 ; d * 2 * ( 1970 ) 2444 ; d * 5 * ( 1972 ) 3077 . a. i. milstein and v. m. strakhovenko , sov . jetp * 58 * ( 1983 ) 8 . r. n. lee and a. i. milstein , phys.lett . a 198 ( 1995 ) 217 . r. n. lee , a. i. milstein , and v. m. strakhovenko , jour . exper . theor * 90 * ( 2000 ) 66 . v. m. katkov and v. m. strakhovenko , jour . * 92 * ( 2001 ) 561 . v. n. baier , v. m. katkov , phys.rev . * d63 * ( 2001 ) 116008 , hep - ph/0005215 . h. a. bethe , phys.rev . * 89 * ( 1953 ) 1256 . g. molire , z.naturforsch.2a ( 1947 ) 133 . v. m. katkov , and v. m. strakhovenko , sov . * 21 * ( 1976 ) 658 . tsai , rev . * 46 * ( 1974 ) , 815 . v. n. baier , v. m. katkov and v. m. strakhovenko , nucl . b328 * ( 1989 ) 387 . r. blankenbecler , phys.rev * d55 * ( 1997 ) 2441 . v. n. baier , v. s. fadin , v. a. khoze and e.a . kuraev , phys . rep . * 78 * ( 1981 ) 293 . i. i. goldman , sov . jetp * 11 * ( 1960 ) 1341 . f. f. ternovskii , sov . jetp * 12 * ( 1960 ) 123 . v. e. pafomov , sov . jetp * 20 * ( 1965 ) 253 . n. f. shulga and s. p. fomin , sov . jetp * 86 * ( 1998 ) 32 . n. f. shulga and s. p. fomin , nucl.instr.and meth . b * 145 * ( 1998 ) 73 . i. s. gradshtein and i. m. ryzhik , _ table of integrals , series , and products _ , academic press , new york , 1965 . x. artru , g. b. yodth and g. mennesier , phys.rev . * d12 * ( 1975 ) 1289 . v. n. baier and v. m. katkov , phys.lett . a 252 ( 1999 ) 263 . n. f. shulga and s. p. fomin , jetp lett . * 63 * ( 1996 ) 873 . l. d. landau , j.phys . ussr * 8 * ( 1944 ) 201 . r. blankenbecler , phys.rev * d55 * ( 1997 ) 190 . v. n. baier and v. m. katkov , nucl . instr . meth . a 439 ( 2000 ) 189 . n. v. laskin , a. s. mazmanishvili , n. n. nasonov , n. f. shulga , jour . * 89 * ( 1985 ) 763 . v. g. zakharov , phys . atom . nucl.62 ( 1999 ) 1075 . v. n. baier , and v. m. katkov , `` variation of radiation length due to lpm effect '' ( to be published ) m. miezowitz , o. stanisz and w. wolther , nuovo cimento * 5 * ( 1957 ) 513 . p. h. fowler , d. h. perkins and k. pinkau , phil.mag . * 4 * ( 1959 ) 1030 . a. varfolomeev , i. gerasimova , i. i. gurevitch et al . , sov.phys . jetp * 11 * ( 1960 ) 23 . e. lohrmann , phys.rev * 122 * ( 1961 ) 1908 . s. c. strautsz et al . , in _ proc.of the 22th intl.cosmic ray conf._,dublin , ireland , v.4 , p.233 . a. varfolomeev , et al . , sov.phys . jetp * 42 * ( 1976 ) 218 . l. d. landau and e. m. lifshitz , _ classical theory of fields _ 4th english ed . , pergamon elmsford , new york , 1975 . a. e. blinov , a. e. bondar , yu . i. eidelman et al . , _ preprint inp 82 - 15 , novosibirsk 1982 _ ; phys.lett . * b113 * ( 1982 ) 423 . v. n. baier , v. m. katkov , v. m. strakhovenko , _ preprint inp 81 - 59 , novosibirsk 1981 _ ; sov.j.nucl.phys , * 36 * ( 1982 ) 95 . a. i. burov and ya . s. derbenev , _ preprint inp 82 - 07 , novosibirsk 1982_. g. l. kotkin , s. i. polityko , and v. g. serbo , yad.fiz . , * 42 * ( 1985 ) 692 , 925 . + g. l. kotkin et al . , z.phys . , * c39 * ( 1988 ) 61 . v. n. baier and v. m. katkov , phys . rev . * d66 * ( 2002 ) 053009 . k. piotrzkowski , _ * c67 * ( 1995 ) 577 . v. n. baier , v. m. katkov , and v. m. strakhovenko , particle accelerators , * 30 * ( 1990 ) 43 . r. j. noble , nucl.instr.meth . , * a256 * ( 1987 ) 427 . o. napoly , _ proceedings of the 2001 particle accelerator conference _ , chicago , ( 2001 ) p.402 . v. n. baier , v. m. katkov , and v. m. strakhovenko , phys . lett . * a229 * ( 1992 ) 429 . v. a. bazylev , v. v. golovoznin , and a. v. demura , zhurn.teor.i eksp . fiziki , * 92 * ( 1987 ) 1921 . v. n. baier , v. m. katkov , v. m. strakhovenko , phys.stat.sol.(b ) * 149 * ( 1988 ) 403 . v. n. baier and v. m. katkov , _ on coherent radiation in electron - positron colliders _ , proceedings of _ quantum aspects of beam physics _ , ed.p . chen , world scientific pc , singapore , 2003 ( in print ) . * * fig.1 * the functions @xmath214 ( curve 1 ) and @xmath215 ( curve 2 ) in eq.([2.42 ] ) . * * fig.2 * the functions @xmath216 ( curve 1 ) and @xmath217 ( curve 2 ) in eq.([2.30 ] ) . * * fig.3 * the spectral intensity of radiation @xmath1413 in gold in terms of @xmath1414 taken with the coulomb corrections ( see eq.([2.2i ] ) ) . * * curve bh is the bethe - maximon spectral intensity ( see eq.([2.35 ] ) ) ; * * curve 1 is the logarithmic approximation @xmath1415 eq.([2.25 ] ) , curve c1 is the first correction to the spectral intensity @xmath1416 eq.([2.30 ] ) and curve t1 is the sum of the previous contributions for the electron energy @xmath28 gev ; * * curve 2 is the logarithmic approximation @xmath1415 eq.([2.25 ] ) , curve c2 is the first correction to the spectral intensity @xmath1416 eq.([2.30 ] ) and curve t2 is the sum of the previous contributions for the electron energy @xmath1294 gev ; * * curves 3 , c3 , t3 are the same for the electron energy @xmath1417 tev ; * * curves 4 , c4 , t4 are the same for the electron energy @xmath28 tev . * * fig.4 * the relative energy losses of electron per unit time in terms of the bethe - maximon radiation length @xmath278 : @xmath279 in gold vs the initial energy of electron ( curve 1 ) and the total pair creation probability per unit time @xmath325 ( see eq.([2.23bp]))in terms of the bethe - heitler total probability of pair creation @xmath324 ( see eq.([2.23p ] ) ) in gold vs the initial energy of photon ( curve 2 ) . * * fig.5 * the pair creation spectral probability @xmath1418 in gold in terms of the exact total bethe - maximon probability taken with the coulomb corrections ( see eq.([2.23p ] ) ) . * * curve bh is the bethe - maximon spectral probability ( see eq.([2.22p ] ) ) ; * * curve t1 is the total contribution ( the sum of the logarithmic approximation @xmath316 ( [ 2.10p ] ) and the first correction to the spectral probability @xmath317 ( [ 2.15p ] ) ) for the photon energy @xmath1419 tev ; * * curve 2 is the logarithmic approximation @xmath316 ( [ 2.10p ] ) , curve c2 is the first correction to the spectral probability @xmath317 ( [ 2.15p ] ) ) and curve t2 is the sum of the previous contributions for the photon energy @xmath1420 tev ; * * curves 3 , c3 , t3 are the same for the photon energy @xmath1421 tev ; * * curves 4 , c4 , t4 are the same for the photon energy @xmath1422 tev ; * * fig.6 * anomalous magnetic moment(amm ) of electron in units @xmath1423 ( @xmath264 in eq.([3.6a ] ) ) in gold vs electron energy in tev . * * fig.7 * the functions re @xmath438(curve 2 ) and i m @xmath438 ( curve 1 ) versus the photon energy taken in units @xmath307 ( because of this the curves are universal ) for the case when the influence of a medium is taken into account only ( eq.([2pr.20a ] ) ) . the both curves are normalized to the asymptotics given by eq.([2pr.22 ] ) in the limit @xmath458 . * * fig.8(a ) the amplitude @xmath428 of the coherent photon scattering in gold under influence of the multiple scattering at the different momentum transfer to the photon @xmath70 in terms of the amplitude im@xmath428 ( [ 3.55sc ] ) calculated for the screened coulomb potential . * curve 1 is im@xmath428 for @xmath5510.4435 @xmath152 . * * curve 2 is re@xmath428 for @xmath5510.4435 @xmath152 . * * curve 3 is im@xmath428 for @xmath5510.0387 @xmath152 . * * curve 4 is re@xmath428 for @xmath5510.0387 @xmath152 . * * fig.8(b ) the same as in fig.1 but for the amplitude @xmath429 in terms of the amplitude im@xmath429 ( [ 3.55sc ] ) calculated for the screened coulomb potential . * fig.9 * the reduction factor @xmath801 for the energy of the initial electron @xmath28 gev . the curves 1,2 and 3 are respectively for @xmath802 = 0.01 , 0.03 and 0.06 . * * fig.10 * the reduction factor @xmath820 for energy of the initial electron @xmath291=25 gev in tungsten ( @xmath824=228 mev ) . the curves 1 , 2 and 3 are respectively for @xmath802=0.01 , 0.03 and 0.06 . * * fig.11 * the energy losses spectrum @xmath1424 in units @xmath1425 , in the target consisting of two gold plates with thickness @xmath960 for the initial electrons energy @xmath291=25 gev . * * curve 1 is for distance between plates @xmath1426 ; * * curve 2 is for distance between plates @xmath1427 ; * * curve 3 is for distance between plates @xmath1428 ; * * curve 4 is for distance between plates @xmath1429 ; * * curve 5 is for distance between plates @xmath1430 . * * fig.12 * the energy losses @xmath1424 in tungsten with thickness @xmath1431 in units @xmath1425 , ( ( a ) is for the initial electrons energy @xmath291=25 gev and ( b ) is for @xmath291=8 gev ) . the coulomb corrections and the polarization of a medium are included . * * curve 1 is the contribution of the main term describing lpm effect ; * * curve 2 is the correction term ; * * curve 3 is the sum of the previous contributions ; * * curve 4 is the contribution of the boundary photons ; * * curve 5 is the sum of the previous contributions ; * * curve t is the final theory prediction with regard for the reduction factor ( the multiphoton effects ) . + experimental data from fig.9 of @xcite . * * fig.13 * the energy losses @xmath1432 in gold with thickness @xmath1433 in units @xmath1425 , ( ( a ) is for the initial electrons energy @xmath291=25 gev and ( b ) is for @xmath291=8 gev ) . the coulomb corrections and the polarization of a medium are included . * * curve 1 is the contribution of the term @xmath1434 ; * * curve 2 is the contribution of the term @xmath1435 ; * * curve 4 is the contribution of the term @xmath1436 , all ( [ z12 ] ) ; * * curve s is the sum of the previous contributions @xmath1437 ; * * curve 5 is the contribution of the boundary photons , see eq.(2.20 ) , @xcite ; * * curve t is the total prediction for the radiation energy losses . + experimental data from fig.12 of @xcite . * * fig.14 * the energy losses @xmath1432 in gold with thickness @xmath1438 in units @xmath1425 , ( ( a ) is for the initial electrons energy @xmath291=25 gev and ( b ) is for @xmath291=8 gev ) . the coulomb corrections and the polarization of a medium are included . * * curve 1 is the contribution of the term @xmath1434 ; * * curve 2 is the contribution of the term @xmath1435 ; * * curve 4 is the contribution of the term @xmath1436 , all ( [ z12 ] ) ; * * curve s is the sum of the previous contributions @xmath1437 ; * * curve 5 is the contribution of the boundary photons see eq.(2.20 ) , @xcite ; * * curve t is the total prediction for the radiation energy losses . + experimental data from fig.13 of @xcite . * * fig.15 * the energy loss @xmath1424 in iridium target with thickness @xmath1439 in units of inverse radiation length @xmath278 ( eq.([2.2i ] ) ) for the initial electrons energy @xmath291=287 gev . the coulomb corrections are included . * * curve 1 is the the bethe - maximon intensity spectrum ; * * curve 2 is the contribution of the main ( migdal ) term describing lpm effect ; * * curve 3 is the correction term ; * * curve 4 is the sum of the previous contributions ; * * curve t is the final theory prediction with regard for the reduction factor ( the multiphoton effects ) . + experimental data from fig.2 of @xcite . * * fig.16 * the same as in fig.15 but for the initial electrons energy @xmath291=207 gev . * * fig.17 * the same as in fig.15 but for the initial electrons energy @xmath291=149 gev . * * fig.18 * the reduction factor for iridium target with thickness @xmath1439 ( 4.36 % @xmath46 ) versus @xmath275 . the curves 1 , 2 , 3 are for energies 149 gev , 207 gev and 287 gev correspondingly . * * fig.19 * the relative energy losses of electron per unit time including the contribution of the correction term eq.([2.25 ] ) in terms of the bethe - maximon radiation length @xmath46 eq.([2.2i ] ) : @xmath1440 in iridium ( curve 1 ) and in lead ( curve 2 ) vs the initial energy of electron . * * fig.20 * the bremsstrahlung intensity spectrum @xmath1441 in units @xmath1239 versus the photon energy in units of initial electron energy ( @xmath275 ) for vepp-4 experiment . the upper curve is the standard qed spectrum , the three close curves below are calculated for the different vertical dimensions of colliding beams ( equal for two colliding beams @xmath1240):@xmath1241 ( bottom ) , @xmath1242 ( middle ) , @xmath1243 ( top ) . the data measured in @xcite are presented as circles ( the experiment in 1980 ) and as triangles ( the experiment in 1981 ) with 6 % systematic error as obtained in @xcite . * * fig.21 * the bremsstrahlung intensity spectrum @xmath1441 in units @xmath1239 versus the photon energy in units of initial electron energy ( @xmath275 ) for the hera experiment . the upper curve is the standard qed spectrum , the two close curves below are calculated with the beam - size effect taken into account : the bottom curve is for @xmath1442 ; while the top curve is for @xmath1443 . the data taken from fig.5c in @xcite * * fig.22 * the bremsstrahlung intensity spectrum @xmath1441 in units @xmath1239 versus the photon energy in units of initial electron energy ( @xmath275 ) for linear collider with beam energy @xmath242gev . the upper curve is the standard qed spectrum . the curve below is calculated with the beam - size effect taken into account for @xmath1444 . * * fig.23 * spectral intensity of radiation of round beams in units @xmath1445 for @xmath1446=0.13 calculated according to eqs.([co2.4]),([co2.7 ] ) . * * fig.24 * the spectral radiation intensity @xmath1447 of coherent radiation ( fast falling with @xmath275 increase curve ) and of incoherent radiation ( the curve which is almost constant ) in units @xmath1448 for beams with dimensions @xmath1295 nm , @xmath1296 nm , for @xmath1253=0.13 . table 1 the characteristics of the lpm effect + listed are the charge of nucleus @xmath68 , the density @xmath645 , the radiation length @xmath1449,@xmath1450 ( see eq.([2.30b ] ) ) , @xmath1451(see eq.([2.23ap ] ) ) , @xmath824 eq.([2.31b ] ) ) , @xmath700 eq.([42 ] ) ) , @xmath383 is the target thickness in percent of the radiation length @xmath1452 ( all photon energies @xmath824 and @xmath700 are for the electron energy @xmath1453gev ) * 9\|c\| material & z & @xmath1454 & @xmath1455 & @xmath237(tev ) & @xmath307 ( tev ) & @xmath824(mev ) & @xmath700(mev ) & @xmath383 + c & 6 & 2.2 & 18.8 & 144 & 580 & 4.3 & 1.5 & 2.1 , 6 + al & 13 & 2.7 & 8.9 & 68 & 270 & 9.2 & 1.6 & 3.5 , 6 + fe & 26 & 7.87 & 1.76 & 14 & 56 & 46 & 2.7 & 2.8 , 6.1 + pb & 82 & 11.35 & 0.56 & 4.3 & 17.2 & 150 & 3.0 & 2.7 + w & 74 & 19.3 & 0.35 & 2.7 & 10.8 & 230 & 3.9 & 2.7 , 6.4 + u & 92 & 18.95 & 0.32 & 2.5 & 10 & 250 & 3.8 & 2.2 , 4.2 + au & 79 & 19.32 & 0.33 & 2.5 & 10 & 240 & 3.9 & 0.1 , 0.7 + water & - & 1 & 36.1 & 277 & 1100 & 2.6 & - & - + std.rock & 11 & 2.65 & 10.0 & 77 & 300 & 8.1 & - & - + * 9\|c\| @xmath1457 & @xmath1464 & @xmath1465 & @xmath1466 & @xmath1467&@xmath1468 & @xmath1469&@xmath1470 & @xmath1471 + 25 & 240 & 3.9 & 5.82 & 0.96 & 1.6 & 0.75 & 0.12 & 28 + 8 & 25 & 1.3 & 5.82 & 0.96 & 0.76 & 1.6 & 0.25 & 3.0 +	the features of electromagnetic processes are considered which connected with finite size of space region in which final particles ( photon , electron - positron pair ) are formed . the longitudinal dimension of the region is known as the formation length . if some external agent is acting on an electron while traveling this distance the emission process can be disrupted . there are different agents : multiple scattering of projectile , polarization of a medium , action of external fields , etc . the theory of radiation under influence of the multiple scattering , the landau - pomeranchuk - migdal ( lpm ) effect , is presented . the probability of radiation is calculated with an accuracy up to `` next to leading logarithm '' and with the coulomb corrections taken into account . the integral characteristics of bremsstrahlung are given , it is shown that the effective radiation length increases due to the lpm effect at high energy . the lpm effect for pair creation is also presented . the multiple scattering influences also on radiative corrections in a medium ( and an external field too ) including the anomalous magnetic moment of an electron and the polarization tensor as well as coherent scattering of a photon in a coulomb field . the polarization of a medium alters the radiation probability in soft part of spectrum . specific features of radiation from a target of finite thickness include : the boundary photon emission , interference effects for thin target , multi - photon radiation . the theory predictions are compared with experimental data obtained at slac and cern sps . for electron - positron colliding beams following items are discussed : the separation of coherent and incoherent mechanisms of radiation , the beam - size effect in bremsstrahlung , coherent radiation and mechanisms of electron - positron creation . .25 cm .5 cm budker institute of nuclear physics , 630090 novosibirsk , russia 0.5 cm * contents * 1 . introduction 1 . general outlook 2 . processes in medium 3 . radiation in presence of an external field 2 . the effect of a medium in an infinitely thick target 1 . the landau - pomeranchuk - migdal ( lpm ) effect for radiation 2 . integral characteristics of bremsstrahlung 3 . the lpm effect for pair creation 4 . anomalous magnetic moment of the electron in a medium 5 . propagation of high - energy photon in a medium in presence of an external field 6 . an influence of multiple scattering on coherent scattering of photon 7 . the polarization of a medium and bremsstrahlung process 3 . impact of a medium for a target of finite thickness 1 . general consideration 2 . boundary effect for a thick target 3 . a thin target 4 . a target of intermediate thickness 5 . multi - photon effects in energy loss spectra 6 . radiation from a structured target 7 . comparison of the results obtained in different papers 8 . a qualitative behavior of the spectral intensity of radiation 9 . experimental investigation of the lpm effect 10 . discussion of theory and experiment 4 . effects in colliding beams 1 . mechanisms of radiation 2 . beam - size effect in bremsstrahlung 3 . coherent radiation 4 . mechanisms of @xmath0 pair creation 5 .
the coronal magnetic field is the main driver of the eruptive solar activity and ensures the coupling between the corona and the interplanetary medium . one of the manifestation of the solar activity is the acceleration of solar energetic particles ( seps ) , electrons and ions , at energy ranges from few kev to few gev and their release in the heliosphere . in the sun s corona , these energetic particles can be accelerated either at low altitudes inside magnetic reconnection regions , _ i.e. _ in flare locations , or at higher altitudes at the shock driven by coronal mass ejections ( cmes ) . it is generally believed that cross - field diffusion is weak during the interplanetary transport of energetic particles @xcite . according to that , energetic particles can only reach the earth if they are eventually injected along the very field lines connecting the acceleration site and the earth . hereafter , we refer to the latter as `` earth - connected field lines '' . the solar particles events are divided in two categories that depends on the time profile of energetic particle flux measured at earth : the gradual - sep and impulsive - sep . the gradual events are usually considered as being accelerated through the cme - driven shock , implying a large longitudinal injection of energetic particles @xcite . thus , energetic particles eventually reach the `` earth - connected field lines '' @xcite . on the contrary , the impulsive events are accelerated through the flare @xcite and their interplanetary injection may not be straightforward . indeed , according to several studies @xcite , energetic protons and electrons impacting the earth may be related to flares located far from the footprints of the earth - connected field line . moreover , it is noteworthy that the flare - accelerated particles are sometimes detected by different spacecraft in the heliosphere over a wide longitudinal range around earth - connected field lines @xcite . these flare - accelerated particles are believed to be accelerated at low altitude within the corona and not at the cme - driven shock . thereby , their injection into the interplanetary medium should be achieved through a reconnection process between open and closed coronal magnetic field . this reconnection can either be the same as , or different from , the flare reconnection . this paper addresses the following questions : how can magnetic reconnection allow energetic particles that are accelerated in the low solar corona , far from the `` earth - connected field lines '' , to get access to the earth ? do exist dynamics of the magnetic reconnection that would lead to inject flare - accelerated particles over a wide angular distance in the heliosphere ? hereafter , we link the three issues of the open versus closed corona , of particle injection problem and of three - dimensional reconnection , in order to motivate the model that we propose to answer these two questions . the large - scale coronal magnetic field presents two different types of connectivity : closed and open . in the closed magnetic domain , field lines have both footpoints rooted in the photosphere , while for the open domain , a field line is anchored at the photosphere by one footpoint only , its other end extending into the heliosphere . the coupling between the corona and the interplanetary medium is carried out at the transition between open and closed magnetic field . this coupling can either be ideal ( _ e.g. _ when propagating waves carry energy from one domain to another ) or non - ideal ( _ e.g. _ when magnetic reconnection leads to mass and magnetic flux transfer between both domains ) . examples of coronal magnetic configurations displaying such an open / closed transition are : at large - scales , the helmet streamers and plumes / pseudo - streamers ( _ e.g. _ , ) ; at smaller scales , coronal jets ( _ e.g. _ ) . streamers are formed above bipolar regions , where the closed coronal loop system is opened at high altitudes to the interplanetary medium through the solar wind pressure ( _ e.g. _ ) . pseudo - streamers , and jets respectively , are created when a larger active region ( _ see e.g. _ ) or a small magnetic bipole @xcite emerges in an already open coronal magnetic flux tube . it has been shown that , apart from special cases @xcite , a single bipole embedded in an open field most often relates to a magnetic topology that comprises a single magnetic null point in the corona ( _ as in _ ; ; ) . such a null - point is associated to a fan separatrix surface and a singular spine field line @xcite . for a bipole embedded in an open field , the fan forms a dome which encloses an inner connectivity domain . the fan separatrix surface separates closed magnetic field lines from the surrounding open ones , located in the outer connectivity domain . the spine field line crosses the fan surface at the null point . therefore , its outer part is also open . magnetic reconnection at the null point allows the exchange of connectivity between closed and open magnetic field lines . this is the standard `` interchange reconnection mode '' @xcite . in general , this closed to open reconnection allows a direct coupling between the open and the closed corona . in particular , it is a natural mechanism to inject impulsive - sep related energetic particles from the closed corona to the interplanetary medium . the acceleration may occur either in a flare located inside the inner domain , before the null point reconnection itself , or at the null point during the interchange reconnection @xcite . the detection at the earth of such flare - accelerated particles implies that the open coronal flux tube surrounding the outer spine and involved in the interchange reconnection , must be _ a priori _ directly connected to the earth - connected field lines . but , it is not always the case . generally , the geometry of the interplanetary magnetic field ( imf ) is assumed to follow the parker spiral . depending on the solar wind velocity , this implies that the earth - connected field lines are rooted between @xmath2 and @xmath3 in longitude . however , past studies of impulsive - sep events show that energetic particles can very often be detected at the earth even if the associated flare is located away ( up to @xmath4 ) from the earth - connected interplanetary spiral field lines @xcite . this may first imply that the earth - connected field line is not a simple parker spiral starting in the low corona . indeed , using pfss extrapolations showed that open magnetic flux tubes in flaring active regions associated to sep events , can spread in longitude from @xmath5 at the photosphere to more than @xmath3 at the solar source - surface , at the base of the parker spiral . comparing coronal magnetic field , metric radio observations and energetic particles in - situ measurements , they showed that the large longitudinal extensions of open magnetic flux tubes could ensure a direct path to earth for energetic particles , being accelerated far from the earth - connected field lines . even though this effect is surely important , it may not be the only way to inject energetic particles onto the earth - connected field lines . in addition to this first issue , energetic particles of some impulsive sep events can also be detected in a wide flux tube , over more than @xmath6 of longitude from the the parent active region @xcite . these large particle beams can be explained by lateral transport along coronal magnetic loops @xcite or lateral diffusion in open coronal magnetic field @xcite . instead invoking the transport effect of particles , could this large angular spread result from a specific dynamics of injection channel related to the interchange reconnection mode ? for conventional models of interchange reconnection , we expect that particles can only be injected into a narrow bundle of field lines whose connection to the earth is rather unlikely . indeed , the field line that reconnects at the null point and opens in the outer connectivity domain is located where the outer spine was located at the time preceding the reconnection . thus , as the reconnection evolves , the reconnected field lines form an open flux tube concentrated around the initial location of the outer spine . flare - accelerated particles should therefore be only injected into this localized flux tube . even though the open magnetic field can display a rapid ( super - radial ) divergence in the corona ( as easily seen in ; ) , particles are transferred into the open corona along this localized flux tube , implying a narrow localized particle beam . nevertheless , the spine itself can shift its position by several degrees during a long - enough reconnection process ( _ see eg . _ ; ) , and could lead to broaden the open flux tube along which particles would be injected . but , we ignore if this motion is directed towards the earth - connected field line and if the displacement covers a sufficient angular distance . this connectivity problem is illustrated by the theoretical geometry sketched in figure [ f - fig1 ] . this paper addresses the following question : how the dynamics of magnetic reconnection in an open - spine null - point topology can lead to inject energetic particles in the interplanetary medium over a large longitudinal range ? through the invoked mechanism , the angular spreading can lead particles to eventually reach the earth - connected field line , even if the outer spine will never be connected to it . one first clue to answer this question can be obtained from the analysis of a former 3d interchange reconnection model driven by flux emergence @xcite . particle acceleration and transport was there modeled by following test - particles in a fixed magnetic field taken from one snapshot of an mhd simulation . they showed that particles are accelerated at the reconnection current sheet , and are injected along a long narrow layer of open field lines . the spatial extension of this propagation channel could be a way to get energetic particles in a large range of longitudes . one can still ask why such an extended acceleration region occurs in this simulation , and whether or not it is sufficient to explain the large spreading of energetic particle beams . according to and the injection channels are affected by the dynamics of magnetic reconnection . therefore , the time evolution of the magnetic field could play an important role in the extension of particle beams . so as to address all these questions , we performed a new 3d mhd simulation of a coronal null - point topology with an open outer spine , driven by slow photospheric shearing motions , and we investigate the dynamics of the interchange reconnection . the model is not dimensionalized , but the topology being considered can make it relevant to a pseudo - streamer around an active region , or to a small jet inside a coronal hole . after describing the numerical model in section [ s - model ] , we present the results on the magnetic reconnection dynamics in sections [ s - current ] and [ s - slip ] . assuming that the dynamics of the magnetic field can represent the macroscopic evolution of injection and propagation channels , our results provide a new conceptual model to understand how energetic particles could reach the earth - connected interplanetary field line when the reconnection site is initially located far away from it ( figure [ f - fig1 ] ) . we finally discuss the relevance of our model on the interplanetary injection of flare - accelerated particles ( section [ s - disc ] ) . the initial magnetic configuration is potential and is built with sub - photospheric monopoles placed such as to obtain a bipole embedded in an open diverging magnetic field , leadinan asymmetric null - point topology with an open outer spine . the configuration is therefore well - suited to model large - scale pseudo - streamers above active regions , or small - scale jets inside coronal holes . the magnetic field is normalized so that the maximum initial photospheric magnetic field @xmath7= 1 $ ] . the top left panel on figure [ f - fig1a ] displays the initial magnetic topology where grey field lines materialize the null - point separatrices . we use the explicit 3d mhd code described in , to which we added gravity finite-@xmath0 effects . the numerical simulation is performed in non - dimensionalized units and in a cartesian domain in which @xmath8 and @xmath9 are the horizontal axes and @xmath10 the vertical axis , covering the volume defined by : @xmath11 $ ] , @xmath12 $ ] and @xmath13 $ ] . the mesh is non - uniform with @xmath14 grid points , not counting two layers of ghost cells around each face of the domain . the mesh size results from a combination of geometric series in all directions and is adjusted to be the thinnest around the null point position ( @xmath15 ) and at the inversion line ( @xmath16 ) . the mesh is displayed on figure [ f - fig1a ] : on the top right panel , the black grid represents the cells of the mesh at the photospheric boundary , whereas the lower part ( @xmath17 $ ] ) of the vertical mesh , where the fan - spine separatrices are located , is shown on the bottom panel with the white grid . ( the details of the mesh can be seen by zooming in the electronic version of this figure . ) our calculation solves the following set of equations using a fourth - order finite - difference and third - order predictor - corrector scheme : @xmath18 @xmath19 @xmath20 @xmath21 @xmath22 @xmath23 being the mass density , @xmath24 the plasma temperature , @xmath25 the plasma velocity , @xmath26 the magnetic field . @xmath27 is the ratio of specific heats chosen here at @xmath28 . these equations are written in the code in their non - conservative fully developed form . the lower boundary of the numerical box follow the line - tied conditions , whereas the upper boundary and the edges of the box follow the open condition . the operators @xmath29 and @xmath30 correspond respectively to a viscous filter that uses a pseudo - laplacian adapted to the local mesh @xcite , and a standard collisional uniform resistive term . the values of @xmath31 and @xmath32 lead to diffusive speeds such as : @xmath33 and @xmath34 on the scale of the smallest mesh , where the maximum alfvn speed is @xmath35 . @xmath36 and @xmath37 are non - physical explicit diffusive terms which smooth the gradients of density and temperature at a speed of @xmath38 , and help to stabilize the numerical computation . owing to the high - order spatial and temporal scheme of our 3d mhd code , firstly the numerical diffusion is very small , and secondly current sheets always remain resolved on more than one grid point . indeed , the use of the explicit resistive term is mandatory to stabilize the code ( see the appendix of ) . also , the use of five grid points to calculate spatial derivatives imposes that all the gradients developing throughout the simulation are resolved on no less than three grid points ( see the bottom panel of figure [ f - fig1a ] ) . the explicit resistivity in our code is therefore responsible for all non - ideal processes described in this paper . in order to drive magnetic reconnection , we prescribed a slow photospheric flow which moves the positive polarity toward the negative @xmath9 in the area of @xmath39 $ ] and @xmath40 $ ] , where a part of the fan is rooted . the top right panel on figure [ f - fig1a ] displays the location at the photospheric boundary and the shape of the applied shearing motion . a temporal ramp is applied , allowing an initial relaxation phase between @xmath41 and @xmath42 . the photospheric driving is then constant until @xmath43 , after which it is progressively stopped with a similar temporal ramp . the maximal amplitude of the prescribed photospheric velocity is : @xmath44 . considering the large size of the numerical box and the existence of a null point , a finite-@xmath0 stratified atmosphere is required so as to respect the following coronal regime : @xmath45 ( except in a radius of 1 around the null point , where @xmath46 ) and a sub - alfvnic and sub - sonic regime in the whole domain for the coronal perturbations , directly induced by the shearing motions . we define this atmosphere as follows . we first fix the photospheric density @xmath47 $ ] in order to obtain a maximal photospheric alfvn speed @xmath48 . this characteristic speed defines the time unit required for an alfvn wave to travel the distance between the centers of the two polarities of the small bipole , @xmath49 . we then define the gravity and temperature profiles such that the alfvn speed remains roughly constant above the altitude of the null point . there , the density scale height , and therefore the pressure scale height @xmath50 should be half of the magnetic scale height @xmath51 . assuming an hydrostatic atmosphere , and choosing @xmath52 as shown in figure [ f - fig2 ] , we integrate numerically the logarithm of the pressure and deduce the vertical density profile at @xmath41 . even though this atmosphere does not exactly reproduce solar numbers , it still puts the system in a solar - like regime . the photospheric shearing motion of one part of the fan s footpoints leads to a bulging of field lines rooted below the fan surface . this expansion compresses the fan separatrix surface and the inner spine , leading to the development of an over - density ( figure [ f - fig3 ] ) . this also leads to the shearing of the spine ( bottom row , figure [ f - fig3 ] ) , _ i.e. _ a loss of alignment between the inner and the outer spine @xcite . this deformation of the null - point compresses and extends the region of weak @xmath53 , inducing the formation of a narrow and intense current sheet in the vicinity of the null point and along the separatrices @xcite , as seen on figure [ f - fig3 ] . thus , null - point reconnection is expected to occur , leading to a realignment of the inner and the outer spine @xcite . by plotting field lines from fixed footpoints and by integrating these field lines up to their conjugate footpoints at different times , we can follow the evolution of the connectivities during the simulation . in order to follow the field lines that are initially rooted below the fan surface , in the positive polarity advected by the applied flow , we computed analytically the position of their advected footpoints at each times . therefore , we follow the connectivity of these field lines from their footpoints fixed in the applied photospheric flow . we initially plot three groups of field lines : pink , light blue and red , closed below the fan surface , and four groups of field lines ( pink , light blue , yellow and red ) open in the corona ( bottom left panel of figure [ f - fig3 ] ) . in each group , field lines are plotted from fixed footpoints distributed along a small segment . all these segments have the same length and each of them is located at the same distance from the fan s footpoints . the field lines of the same color group are plotted from footpoints uniformly distributed along one segment . as reconnection proceeds , magnetic flux is exchanged between connectivity domains , and the fan surface is displaced and sequentially crosses the fixed footpoint of the plotted field lines . since the plotted field lines are uniformly distributed along each segment , the number of field lines that reconnected per time unit is proportional to the reconnection rate . as the system evolves , the field lines , initially closed in the inner connectivity domain , reconnect at the null point and jump to the outer connectivity domain , opening - up in the corona ( bottom row of figure [ f - fig3 ] ) . on the opposite , the initial open field lines reconnect at the null point and closing - down below the fan surface . this exchange of connectivity between the inner and the outer connectivity domain is , in fact , the interchange reconnection @xcite . during the reconnection through separatrices , the transfer of magnetic flux between the inner and the outer connectivity domain leads to displacements of the separatrices . therefore , the positions of the fan surface and the spine line , both change . even though the deformation of separatrices appears clearly during the evolution ( see the dark blue field lines lines on figure [ f - fig3 ] ) the relative length over which the spine field line shifts by reconnection , as compared to the horizontal scale over which the neighboring open field lines expand with height , is not very large . in itself , this result suggests that the large expansion with height of field lines above flaring active region , as identified by @xcite , may not always allow energetic particles previously accelerated low down in the corona to reach the earth - connected field lines . before field lines reconnect at the null point , they display an apparent continuous slipping motion toward the inner spine ( see figure [ f - fig3 ] as well as the animation on the electronic version ) . figure [ f - fig4 ] also displays the evolution of the connectivities at high altitude , in the whole numerical domain . the four panels show the same groups of field lines , as plotted in figure [ f - fig3 ] , at four different times . the animation shows that , right after having reconnected at the null point and jumped to the outer connectivity domain , field lines are located all along the outer spine in the whole numerical domain . nevertheless , these reconnected field lines do not stay in the vicinity of the outer spine : they very quickly move farther away from it . indeed , after the null point reconnection , field lines display an apparent slipping motion along a specific direction towards the negative @xmath9 . even though the topology is that of a null point , this apparent slipping motion is reminiscent of a slipping reconnection regime occurring through quasi - separatrix layers @xcite . a quasi - separatrix layer ( qsl ) is a purely three dimensional geometrical object , formed by narrow volumes of continuous connectivity but across which the connectivity displays strong gradients @xcite . contrary to the null - point reconnection , the absence of magnetic discontinuity implies that , in a qsl , field lines simply exchange their connectivity with their neighbors and that their footpoints is moved in time across the qsl . thus , continuous reconnection between neighboring field lines leads to a continuous exchange of connectivity and apparent slippage of field lines plotted from fixed footpoints can be noticed . this slipping reconnection regime has been confirmed by several mhd numerical simulations @xcite and supports numerous association of observed flare ribbons with qsl calculated in extrapolated magnetic fields @xcite . since it has recently been shown that this regime can also occur around null points with completely closed magnetic separatrices @xcite , we investigate whether or not the same happens here , in the presence and around open separatrices . the slipping motion of reconnected field lines , before and after the null point reconnection , could result either from slip - running / slipping reconnection through qsls , or from mass motions induced by the propagation of shear alfvn waves injected either by the photospheric driving , by the reconnection jets , or even from numerical errors . note that at late times , the animation shows that the green field lines , rooted in the photospheric driving region far from the null point , also move at the top of the numerical domain ( figure . [ f - fig4 ] ) . nevertheless , these motions are straightforward to interpret . they clearly result from the upward propagation of undulations , that simply correspond to the propagation of a shearing alfvn wave from the line - tied boundary , which is generated by the photospheric driving . we compared the slipping velocities of field lines and the plasma velocities in the vicinity of these reconnected field lines , at the top of the box . we selected the fifth reconnected field line in each of the 3 studied groups of field lines , and we determined for these three selected field lines their slipping velocity at the top of the box for each time interval of the simulation . then , we evaluated the plasma velocity at the top of these three selected field lines . since the mesh of the plotting is much tighter than the mesh of the 3d mhd code , the position of the top of the field lines is never located exactly on one of the mesh points where the mhd variables are integrated . thus , we computed the mean value of the plasma velocity of the four mesh points surrounding the field lines position at the top of the box . the resulting velocities are plotted in figure [ f - fig5 ] . after the field lines have reconnected at the null point , their slipping velocity is greater than the alfvn speed , @xmath54 , and than the fast - mode speed @xmath55 at the top of the box . this leads to a super - alfvnic and super - sonic slipping motion . the slipping velocities of field lines later decrease to @xmath56 and become sub - alfvnic and sub - sonic . meanwhile , the plasma velocity at the top of the box is found to increase during the simulation , in agreement with the upward propagation of the shear alfvn wave from the photospheric driving at the bottom of the box . even though the slipping velocities decrease , they remain much faster than the plasma velocity until @xmath57 ( see figure [ f - fig5 ] ) . after @xmath57 , the temporal evolution of the slipping velocities becomes the same as the evolution of the plasma velocities and their values becomes comparable . therefore , the displacement of field lines at the end of the simulation is indeed due to mass motion , _ i.e. _ ideally evolving field lines , such as the green field lines . at this stage , the analysis suggests that the slipping motions of the reconnected field lines , at least until @xmath57 , are due to magnetic reconnection through a qsl , and at least not to mass motion . the evolution of the slipping velocity of the field lines , successively super - alfvnic then sub - alfvnic , corresponds to a transition from slip - running to slipping reconnection following null - point reconnection , as first put foward by , but for fully closed magnetic fields . if the slipping motion of the reconnected field lines was due to systematic numerical errors throughout the whole integration domain , whether coming from the mhd computation or from the field line integration code , the green field lines , which evolve through the upward propagating shear alfvn wave , should display a slipping motion too . the animation attached to figure [ f - fig4 ] clearly shows , however , that the reconnected field lines slip long before the top of the undulating green lines start to slip , the latter only being due to the alfvn wave passing through the top boundary . this difference indicates that there is no significant systematic numerical error . still , one expects larger errors which could locally affect the reconnected field lines , since those pass through a current sheet , in which sharper magnetic gradients develop . such errors can be estimated with the divergence of the magnetic field , since our mhd code does not enforce the solenoidal condition @xmath58 . it should be noted that all the mhd equations are fully developed in the code , with all the @xmath59 terms being removed , so that these terms neither produce momentum nor induction . in order to quantify the effect of non - zero @xmath59 on the slipping motions , we nevertheless study the dynamics of the field lines with modified magnetic field @xmath60 . at each time and position in the domain , every component of the modified magnetic field was computed by the following taylor - like expansion where @xmath61 is the local size of the mesh interval in the direction @xmath62 : @xmath63 , where @xmath64 . using this modified magnetic field , we re - conduct the same velocity analysis as described above , for the very same reconnected field lines . we find that at the top of the box , the temporal evolution of the slipping velocity and the displacement of the @xmath60 field lines display the same temporal evolution than the @xmath26 field lines . the dynamics of magnetic reconnection for @xmath60 is the same than @xmath26 : the succession of null - point , slip - running and slipping reconnection regimes . however , one notices that reconnection process for @xmath60 starts twice alfvn time later than for @xmath26 . in addition , the slipping velocity of field lines , at a given location at the upper intersection of field line with the boundary , differs between @xmath60 and @xmath26 . indeed , we evaluate that the slipping velocities with @xmath60 depart from those with @xmath26 by @xmath65 on average.this number is presumably due to the strong sensitivity of the integration of field lines within a strongly diverging flux tube such as a quasi - sepatrix layers ( which is described below as the structure along which field lines are slipping ) , even when starting from a region where b has locally been slightly modified to b * even though our mhd code does not ensure the solenoidal condition , and in spite of the fact that the present calculation was ran in single precision , we thus find that the measurable numerical errors do not affect the large - scale dynamics of magnetic reconnection and the observed slipping of the reconnected field lines . in , we had shown that the closed separatrices of a null point were surrounded by a qsl halo . the presence of a qsl , through which continuous magnetic reconnection is known to occur , can be checked by computing the norm @xmath66 @xcite or the squashing degree @xmath67 @xcite . these quantities somehow measure the gradients of field line connectivity , and thus they map the position of the qsl . usually @xmath66 and @xmath67 are calculated in closed field configurations only . here we compute @xmath67 in our open fields in a similar way as how calculated @xmath66 : the upper and bottom boundaries are considered as the two reference planes in which field line footpoints are calculated . even if the lower line - tied boundary appears as more natural reference plane than the upper open frontier at @xmath68 , @xmath67 is uniquely defined for each lines given this choice of computation limit . however , the choice of a different reference plane ( such as a different altitude for the top boundary ) would result in a different value of @xmath67 for a given field line . nevertheless , assuming a monotonous divergence of the magnetic field , the distribution of @xmath67 would remain monotonously homogenous with a different choice of @xmath10 . hence , the occurrence and locations of qsls in open field lines , manifested by open narrow layers of high-@xmath67 , would be independent of the choice for the reference upper plane . figure [ f - fig6 ] shows the location and shape of the high - altitude qsl which surrounds the open spine , as seen from the large @xmath67 isocontours plotted at @xmath69 . initially , the qsl is broad and is almost symmetrically distributed around the spine . at later times , however , the prescribed photospheric motions shear the system , and increase the asymmetry of the null point . following , the parameter @xmath70 which defines the asymmetry of the fan , increases in our simulation from @xmath71 ( at @xmath41 ) to @xmath72 ( at @xmath73 ) . this is associated with the development of a non - symmetric , elongated and thin qsl around the open spine , in a similar way as what we found earlier in an initially asymmetric , but closed , null - point topology @xcite . on figure [ f - fig5 ] , we report the distribution of the high value of q , and thus the location of the qsl , at the top of the box . it is now clear that the intersections of the slipping field lines with the upper boundary of the box are located within the qsl . this finally leads us to interprete the super- ( followed by sub- ) alfvnic field line motions as a substantial amount of slip - running ( followed by slipping ) reconnection , of the field lines which have previously jumped from the inner to the outer connectivity domain by the interchange null - point reconnection . the distribution of q ( figure [ f - fig6 ] ) can also explain why the slipping velocity of some reconnected field lines displays a local minimum around @xmath43 after the slip - running phase ( see figure [ f - fig5 ] ) . a band of low q has formed within a higher q region . noted that field lines slip slower ( res . faster ) in regions of lower ( resp . larger ) @xmath67 . we also find that the brief period of lower field line velocities around @xmath43 is indeed associated with the crossing of this local low-@xmath67 region by the slipping field lines . so as to explain several in - situ measurements in the heliosphere and at earth , we investigated an mhd model for the dynamics of 3d magnetic reconnection to address the interplanetary injection of energetic particles during impulsive - sep events . the model is based on the interchange reconnection mode in an open - spine null - point topology @xcite . this situation is relevant for the two solar magnetic configurations that couple the closed magnetic corona with the open corona and the heliosphere : the large - scale pseudo - streamers above active regions , and the small - scale jets inside coronal holes . in a pseudo - streamer configuration , flare - related particles may have been previously accelerated inside the inner connectivity domain , whereas in a jet configuration , particles are supposed to be directly accelerated at the null - point . in both cases all energetic particles pass through the current sheet formed at the interface between open and closed field regions , before being injected in the interplanetary medium through interchange reconnection . the main result of the mhd simulations is that , before and after the standard interchange reconnection at a coronal null point , the field lines rooted in the photosphere continuously reconnect through slip - running and slipping reconnection regimes . this is due to the presence of a halo of quasi - separatrix layers ( qsls ) that surrounds the fan separatrix surface and spine singular field line originating at the asymmetric null point . in such an `` hybrid topology '' magnetic reconnection does not behave as usual , _ i.e. _ field lines do not exchange their connections instantaneously by pairs . this continuous reconnection associates with a large extension of the current sheet along the fan , exceeding the length of the current sheet which forms at the shearing null point . this large current sheet is a direct result of the development of a qsl around the spine line of a null point becoming increasingly asymmetric . this qsl sustains slip - running and slipping reconnection , hence leads to a large displacement of slipping field lines at high altitudes in the open field domain . this result generalizes to open spine configurations the same mhd process originally discussed for a fully closed magnetic fields @xcite . since slipping reconnection across qsls around an asymmetric ( also called improper ) null point seems to occur for line - tied cases as well as for open boundary conditions , we argue that it must be a generic property of 3d reconnection at null points . as such , the `` torsional fan reconnection mode '' found in asymmetric nulls by may actually be just another formulation of the same physical effect . even though our calculations do not treat the dynamics of the particles themselves , our analyses of the evolution of reconnected field lines can be used as a first proxy to determine the time - evolution of injection and propagation channels for accelerated particles . thereby , we assume that the particle acceleration process , wherever it occurs , can lead to a significant number of particles at sufficiently high energies . in the classical two - dimensional interchange reconnection mode , particles should only be injected into open field lines through the short current sheet that forms around the misaligned separatrices of the null point . thus , particles should only be injected along the few field lines which have reconnected at the null point . the qsl - related field line dynamics in our simulation , however , let us conjecture that during 3d interchange reconnection , particles can also be successively injected along ( initial fast ) slip - running ( and later slow slipping ) field lines . in principle , the association of slip - running and interchange reconnection should allow that particles are accelerated all along the intense current sheet localized in the qsl and not only around the null point . let us illustrate a discretized version of this continuous mechanism : when an initially closed field line @xmath74 reconnects and opens at the null point , a first set of particles @xmath75 are injected along this reconnected open field line . then , as the same open field line @xmath74 slip - runs away from the open spine , another closed field line @xmath76 reconnects and opens at the null , and a second set of particles @xmath77 are injected along this second reconnected field line . in the meantime , the motion of the field line @xmath74 has made it slip along the qsl - related low - altitude current sheet and a new set of particles@xmath78 is injected along the open field line @xmath74 . as this whole process repeats itself @xmath79 times , newer and newer sets of particles are accelerated and injected along the field line @xmath80 being just reconnected at the null point , as well as along all the slip - running field lines @xmath81 that are moving farther and farther away from the open spine . therefore , this mechanism implies that the set of energetic particles would be injected continuously along field lines which are undergoing slip - running / slipping reconnection away from the open spine . we conclude that during the interchange followed by slip - running reconnection , particles could be injected within an extended and narrow flux tube , progressively formed by the field lines slipping along a large - scale open qsl . in a fixed magnetic field configuration , this may explain the elongated shape of particle impacts at the top of the numerical domain , as obtained in the simulations of . in a time - varying system , our analysis implies that particle beams originating from the interchange reconnection should sweep the high altitude domain , along a narrow qsl - related flux tube that extends through the slipping reconnection process . when considering a solar - like configuration such as a large - scale pseudo - streamer or a small - scale jet , the resulting sweeping energetic particle beam could eventually reach the earth - connected parker spiral field line , even if the acceleration site is not initially connected to this very line ( as shown in figure [ f - fig1 ] ) . our model provides a new type of mhd - based particle injection process along a sweeping beam , that may cover a large angular range . around @xmath82 of active regions related to impulsive sep events are close to or included in an open coronal magnetic flux tube @xcite . they are auspicious magnetic configurations for finding pseudo - streamer and jet configuration . in addition , in such a region nearby open coronal magnetic field , when we measured a sep events at the earth , a small jet is usually observed in uv and occurs at a comparable time than type iii bursts @xcite.this suggests that open and close field lines are reconnecting through the interchange mode in a null - point topology . therefore , the null - point configuration embedded in an open magnetic field , as shown in figure [ f - fig1 ] , can be considered as a typical configuration for flare - related impulsive sep events , regardless any cme . this leads us to suggest that the model proposed by this study should be a common injection process . however , one can ask if the angular distance travelled by slipping field lines and thus the sweeping particle beam , in a solar context , is significant ? even though the non - dimensionalized units do not allow to directly compare the results of the simulation with the observations , one can estimate the angular distance , @xmath83 , covered by slipping field lines for typical solar values . associating the geometry of the system to a right triangle , where the spine and the top of the box form the right angle and the slipping field line is the hypothenuse , we evaluate the angle @xmath83 . for an alfvn speed of @xmath84 , and an arbitrary distance unit of @xmath85 for a coronal jet ( resp . @xmath86 for a pseudo - streamer ) , the flux tube formed by the slipping field lines expands over an angular range @xmath87 at the altitude of @xmath88 for a jet ( resp . @xmath89 ) . comparing t the spine field line s motion covering the angular distance @xmath90 , it appears that the slipping reconnection lead to a much more substantial angular spreading of the open coronal flux tube , guiding energetic particles in the heliopshere . in addition , it is noteworthy that @xmath83 is constrained by the simple geometry of the the model used in the 3d - mhd simulation . indeed , the diverging of the magnetic field in 3d is relatively small , but a largest expansion will lead to increase values of @xmath83 . actually , the spatial extension of the open coronal magnetic field being super - radial @xcite , field lines could easily slip over larger distances than in our simulation , implying that the particle beam could sweep over a larger angular range . according to and , energetic particles should not be scattered significantly during their interplanetary propagation , implying that energetic particles propagate in the inner heliosphere along the same field line . thus , the association of slipping and interchange reconnection provides a new way to understand sep connection to the earth and the large angular range over which particles can be detected in the interplanetary medium , suggesting that the cross - field transport of accelerated particles in the coronal magnetic field is not the only way to inject particles in an extended flux tube . this result leads us to believe that the mhd model proposed in this study can be fully relevant to explain several observations such as the measurement of energetic particles over more than @xmath91 of longitude . the model that we propose for sweeping particle beams in the heliosphere is based on the assumption that the macroscopic spatial evolution of propagation channels follows the dynamics of interchanging and slip - running field lines . future work will therefore have to calculate explicitly the particle trajectories in the diffusion region and along the continuously reconnecting field lines in a time - dependent magnetic field . the authors thank the referee for helpful comments , which improved the clarity of the paper . the mhd calculations were done on the quadri - core bi - xeon computers of the cluster of the division informatique de lobservatoire de paris ( dio ) . the work of s.m . is funded by a fellowship of direction gnrale de larmement ( dga ) . financial support by the european commission through the fp6 solaire network ( mtrn - ct-2006 - 035484 ) is gratefully acknowledged .	we present a new model to explain how particles ( solar energetic particles ; seps ) , accelerated at a reconnection site that is not magnetically connected to the earth , could eventually propagate along the well - connected open flux tube . our model is based on the results of a low-@xmath0 resistive magnetohydrodynamics simulation of a three - dimensional line - tied and initially current - free bipole , that is embedded in a non - uniform open potential field . the topology of this configuration is that of an asymmetric coronal null - point , with a closed fan surface and an open outer spine . when driven by slow photospheric shearing motions , field lines , initially fully anchored below the fan dome , reconnect at the null point , and jump to the open magnetic domain . this is the standard interchange mode as sketched and calculated in 2d . the key result in 3d is that , reconnected open field lines located in the vicinity of the outer spine , keep reconnecting continuously , across an open quasi - separatrix layer , as previously identified for non - open - null - point reconnection . the apparent slipping motion of these field lines leads to form an extended narrow magnetic flux tube at high altitude . because of the slip - running reconnection , we conjecture that if energetic particles would be traveling through , or be accelerated inside , the diffusion region , they would be successively injected along continuously reconnecting field lines that are connected farther and farther from the spine . at the scale of the full sun , owing to the super - radial expansion of field lines below @xmath1 , such energetic particles could easily be injected in field lines slipping over significant distances , and could eventually reach the distant flux tube that is well - connected to the earth .
in this paper we use the trade credit network of italian firms to test a model of `` many - to - one '' contagion of economic growth or economic crisis . academic research on inter - firm trade credit networks is still in its infancy , as the data on trade - credit transactions is not easily accessible . an exception is the analysis of the trade credit network of japanese firms , which have been rather extensively characterized by , for example , tamura et al . @xcite , miura et al . @xcite and watanabe et al . . our approach to direct contagion of economic growth rate is built on the assumptions based on the previous results published in various economic literature on balance sheet contagion ( e.g. kiyotaki & moore @xcite , boissay @xcite , petersen & rajan @xcite , economic growth schumpeter @xcite ) , and from our own previously developed methods and models following complex systems approach ( solomon & richmond @xcite , challet et al . @xcite ) . when many firms simultaneously borrow from and lend to each other , and in particular when these firms are speculative and dependent on the credit flow , shocks to the liquidity of some firms may cause the other firms to also get into financial difficulties . as obvious as this argument might sound , it is very difficult to prove the effect of direct contagion in the data , exactly for the reason that the linearity is lost as soon as the firms are simultaneously interacting with each other . the `` many - to - one '' contagion relies on the hypothesis that the change in the annual sales of a supplier follows the change in the * mesoscopically aggregated demand * , i.e. that the yearly growth of sales of a supplier would be proportional to the yearly growth of demand of its customers . assuming that the growth of purchases of all customers of a supplier in a yearly period is known to us , should nt we be able to predict the growth in the sales of the supplier ? indeed we should , as long as : * the linkages between the customers and the supplier are constant over the longer period ( at least two years , as the growth rate of financial indicators reported in balance or profit and loss statements is available on yearly basis ) * the growth of demand ( purchases ) of a customer is assumed to be uniformly distributed among all its suppliers . assuming that the above conditions are satisfied , we compare the prediction of the growth of sales with the real growth as measured from the profit and loss statements , and so we can test our hypotheses of growth contagion from customers to suppliers . the paper is organized as follows . in section [ sec : theory ] we describe the nature and applications of different interaction mechanisms that can be involved in the self - amplifying auto - catalytic loops in supporting both peer interactions and the bi - directional feedback between the micro and the macro structures of the economy . following that , we describe the possible methodologies that could be used in order to empirically show the existence and operation of an auto - catalytic feedback . in section [ sec : data ] , we give the detailed description of the data that were used for the network model and internal properties of the nodes . in that section the reader can also find the definitions of the variables and the algebraic notations . in section [ sec : results ] , we elaborate on the empirical results and provide their interpretation and implications . finally we discuss the results in section [ sec : discussion ] . one of the major issues in economics is to understand how relativelly small and temporary endogeneous changes in technology or wealth distribution may generate macroscopic effects in aggregate productivity , asset prices etc . for this purpose , it is necessary to identify self - amplifying mechanisms , filtering out the main mechanisms that may trigger dynamics leading to a systemic change , from other interactions destined to drown in the noise of local , short lived perturbations . this implies that the vast majority of microeconomic interactions that may affect the macroeconomic / systemic level do so by a kind of auto - catalytic positive feedback loop . this idea was already proposed in various contexts , but was often dismissed in the absence of concrete mechanisms that realize it . below , we list three mechanisms that might effectively amplify microscopic events to macroscopic dynamics in economic systems . in this paper we deal only with the first mechanism , with the second one we dealt in @xcite , and the third mechanism we might tackle in the future . [ [ peer - to - peer - one - to - many - and - many - to - one - interactions - between - firms - in - the - network ] ] ` peer - to - peer ' , ` one to many ' and ` many - to - one ' interactions between firms in the network + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + the domino effect , or contagion can be best understood using powerful mathematical and statistical - mechanics tools developed in percolation theory . they allow the rendition of precise predictions that correspond to real world stylized facts : macroscopic transitions caused by minute parameter changes , fractal spatial and temporal propagation patterns , delays in growth or crisis diffusion between economic sectors or geographical regions . for a formal discussion of social and market percolation models the reader is referred to goldenberg et al . @xcite . in this paper we deal with the many - to - one contagion principle ( in the cases when a supplier has a single customer it reduces to peer - to - peer principle ) . the peers are tied financially , and also physically by the goods they pass . we introduce the assumption that the peers are correlated though their growth rate and we empirically test it . we attempt and succeed to find only partial evidence in our data that the many - to - one is responsible for the propagation of the growth rate on the trade network consisting of suppliers and their customers . this mechanism is only partly responsible for the congation of growth rate , and other mechanisms , such as the following ones , should also be considered . [ [ inter - scale - macro - to - micro - reaction - between - firms - and - the - system ] ] inter scale macro - to - micro reaction between firms and the system + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + for a long while , one of the drawbacks of the dynamical models inspired by physics was the absence of interaction between scales . a model that introduces this interaction was introduced by solomon and golo @xcite . this is a financial model extending the ideas of minsky @xcite , where not only the fate of individual firms ( e.g. failure ) influences the system state ( risk aversion leads to rise in corporate interest rates ) , but also the state of the system is feeds back onto its own components ( e.g. a rise in interest rate leads to more failures ) . together with contagion across the network , the model generates a bounty of predictions that agree with the stylized facts . for example : delaying or arresting the propagation of distress by targeted intervention in key individual components or in system properties ( such as the interest rate ) . this model has been confronted with empirical data in @xcite and provided a significant contribution in the interpretation of the economic collapse in 2008 . [ [ self - interaction - firms - acting - upon - themselves ] ] self - interaction : firms acting upon themselves + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + a model for self - dependent reproduction was proposed by shnerb et al . @xcite , generalizing on the ideas of malthus that proliferation in the microscopic level in stochastic systems leads to the spontaneous emergence of a collection of adaptive objects . the model is analytically tractable by statistical - mechanics method as mentioned above . it generates a host of qualitative ( phase transitions ) and quantitative predictions about the macroscopic behavior of the system . these predictions were precisely confirmed by empirical measurements in many cases : crossing exponentials between decaying and emerging economic sectors after a shock @xcite , identity of the wealth inequality ` pareto - zipf ' exponent @xcite and the market instability exponent @xcite @xcite , etc@xmath0 . the self - interaction mechanisms are not in the scope of this paper , and the microeconomics models of firms are not considered as well . in the analysis of contagion , causality is the key . the mechanisms we sketched in the preceding subsection call for testing , but the analysis should account for causality , rather than correspondence . the microscopic behavior of agents in a network should be revealed in the structure of the _ links _ between the agents and their _ interaction _ patterns . in a financial network , the visible communication between agents is through the invoices they issue and the payments they make . the structure of links between the nodes in a network is formally termed the ` topology of the network ' . the metrics ( indicators ) typically used for measuring topology are for example : node degree being the number of connections from / to each node , clustering being the number of distinct groups of nodes , and connected - component sizes . however , the description of the topology is not the subject of our paper . we aim to justify the existence of the network and the application of the network analysis by trying to measure the significance of the interaction across the network nodes . in our work , we focus on a selection of certain neighborhoods that consist of buyers from a single supplier . thus , the clustering coefficients and component size are used only for validating that the selection process does not destroy the statistical properties of the complete sample . we are taking into account that the properties of firms influence the interaction between them . in financial networks , the strength and the interaction patterns can not be interpreted without understanding the level of financial exposure to risk that the agents ( firms ) are in . the financial risk of providing and extending the trade credit relation is an interplay between a firm s own liquidity position , its social neighborhood ( its industry and its particular buyers and suppliers ) , and systemic effects such as the interest rate . since the risk exposure of a firm can not be directly measured , the ability to meet obligations subject to the environment and the system we have used a quantity available in our data termed the ` rating score ' . this quantity will be defined in the data section below . the data on individual firms come from a dataset ( further on abbreviated as bs ) of italian limited liability companies end - of - year balance sheet and profit & loss statements , which is a part of a proprietary database . the network is assembled from the sales invoices issued by suppliers to their customers when they sell an item . some of these invoices were presented to a bank in order to acquire trade - credit ( tc ) . the borrower in most cases was the seller in the supplier - customer pair . it is these invoices that the bank recorded and which we have been able to analyse . in this study we combined the two datasets ( bs ) and ( tc ) in order to select the suppliers that are most appropriate for this analysis . [ [ balance - sheets - database ] ] balance sheets database + + + + + + + + + + + + + + + + + + + + + + + in italy , all limited liability firms are obliged to submit their annual financial report ( balance sheet ) to the local chamber of commerce . items contained in the firm s balance sheet are assets and liabilities of the firm such as : equity , net - sales , accounts receivable , inventory , bank loans , accounts payable , financial costs , etc @xmath0 . other than balance data , these reports contain financial ratios . financial ratios are ratios of quantities within the balance sheet items , such as acid test or the receivables conversion period and their purpose is to help quickly estimate the financial status of the firm . the balance sheets are collected and stored by an external agency . the firms , no matter whether defaulting or non - defaulting at the end of the period , are ranked with a ` rating score ' ranging from 1 to 9 in increasing order of default probability : 1 is attributed to firms that are predicted to be highly solvent , and 9 identifies firms displaying a serious risk of default . notice that the ranking is an ordinal : firms rated as 9 are not implied to have 9 times the probability of defaulting as compared to firms rated 1 . a good description of the rating score is given in a paper by bottazzi et al @xcite . for the purpose of our analysis , and in accordance to the practitioners behavior , we divide all companies into three groups , using the rating score : one group with an easy access to bank credit ( rating 1 - 3 ) , one with the access to bank credit ( abc ) at medium risk ( rating 4 - 6 ) and the third group at high risk and little or no access to bank credit ( rating 7 - 9 ) . l \| l rating 1 - 3 & abc @xmath1 a + rating 4 - 6 & abc @xmath1 b + rating 7 - 9 & abc @xmath1 c [ [ trade - credit - data ] ] trade credit data + + + + + + + + + + + + + + + + + the trade credit database ( tc ) contains all inter - firm delayed payment transactions during 2007 that were intermediated by a large italian bank . this bank covers about 15% of the entire trade credit in italy , according to the official statistics of the bank of italy . [ [ algebraic - notation ] ] algebraic notation + + + + + + + + + + + + + + + + + + we use the symbols @xmath2 and @xmath3 to mark a supplier and a customer , respectively . transactions between a supplier and a customer will be presented by magnitude and direction of the money flow . a general event of goods or services sold to a customer can be graphically described by @xmath4 . the 2007 tc ( trade credit ) records are invoices that account for goods / services supplied by firm @xmath2 to firm @xmath3 that were presented to the bank for discounting in 2007 . these transactions were booked as accounts receivable in firm @xmath2 s assets , and as accounts payable in firm @xmath3 s liabilities . we can write @xmath5 the sum of all payable trade invoices and cash payments from a customer @xmath3 to a supplier @xmath2 accounting for purchases of goods or services in 2007 . @xmath6 the total invoices used by firm @xmath2 as collateral on a credit line in 2007 . we may also define the following notation from the bs ( balance sheet ) records of firm @xmath2 : @xmath7 : balance sheet item of total sales ( cash and credit ) of firm @xmath2 in year y @xmath8 : balance sheet item of total purchases ( cash and credit ) of firm @xmath2 in year y [ [ testing - for - completeness - of - firm - level - information ] ] testing for completeness of firm - level information + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + to ensure that we have sufficient coverage in the tc database of each firm s sales , we selected the firms @xmath2 for which the expression @xmath9 is greater than a pre - defined value . we call this the ` matching threshold ' or ` matching ' . we vary the threshold between 0 and 1 in order to create the best sample for our empirical analysis . we choose a representative sample with high level of completeness by setting the matching threshold so it is the maximal value that still renders a single large connected component of size @xmath10 , and that the consequent connectivity of firms spans from 1 to approximately the size of the component ( @xmath11 ) . the matching ratio reaches values greater than than 1 ( more than 100% completeness ) , perhaps due to misalignment of the time windows between the trade credit network and balance sheet data . this will be discussed below . the analysis in this paper is , therefore , restricted to the sample of suppliers that have a matching proportion of 0.8 and up to 1.2 ( i.e. 80% to 120% , or formally @xmath12 ) . the reason to prefer a range with a top - cap is that much larger proportion signals to us a mismatch : that the sample of invoices for the selected firm may be incorrect . possibly due to an error in invoice registration . we did not expect to retrieve a large network from the firms exhibiting high matching proportion . several factors reduce the matching . the two most prominent ones are : * trade credit data do not record changes in cash holdings . they do , however , record an accounts receivable on the supplier s side and an accounts payable on the customer side . when accounts receivables are cashed by the customer , a ` sale ' item will be booked to the supplier . now , outstanding invoices and payments made in 2007 may have been booked as accounts receivables in 2006 . we estimate a delay of 3 months on average for discounts and 9 months for payments ( cf . the misalignment of the time frames ) . * the sales ( @xmath13 ) are all sales of the firm including sales performed in other monetary channels . the tc database holds information on invoices only . out of @xmath14 firms that were initially available both as creditors in the tc data ( have incoming links ) and in the bs data ( have sales in @xmath15 ) , only 671 companies fulfill the matching range @xmath12 . lowering this threshold would grow the sample exponentially but at the cost of unavoidably entering suppliers with lesser proportions of their total debt owed . in table [ tab : netparams ] we list some statistics of this network , assuming that the links are directional . out of the total 671 suppliers , 190 companies are of reasonable size in accordance with bank practices ( @xmath16 eur ) . .some basic parameters of the subset network . multi - edge node pairs give the number of node - pairs that link in both directions ( reciprocal ) . comparing the number of customers to the number of links presents a picture of many tree - like subgraphs connected to each other with minimal number of loops . [ cols=">,^",options="header " , ] in our inter - firm network , the in - degree of a supplier is the number of its customers . in a previous study , miura et al . @xcite obtained power laws of the in- and out - degree distributions , with exponents of @xmath17 . in our network , we were able to confirm a similar finding of @xmath18 , i.e a negative exponent @xmath19 . as for the in - degree vs. size , a statistically significant correlation was recovered in our sample with a slope of @xmath20 . in the japanese network of miura et al . , a slope of @xmath17 was obtained . both results appear in figure [ fig : indegree ] . 0.5 ) , title="fig : " ] 0.5 ) , title="fig : " ] in the second observation , the correlation between sales in 2007 and in - degree ( panel [ subfig : indegreevssales ] ) pearson s correlation coefficient is @xmath21 . the broken line marks a net - sales of 1 million eur . the crossing between the broken and the red linear fitting line is characteristic of a firm with @xmath22 . below the dashed line we can find the small companies , and it is clear that most of them also appear left of the crossing point , indicating a client - base smaller than @xmath23 . in order to quantify the relevance of the many - to - one approach , we measure the number of suppliers in our selection that do not have a key customer but are permanently related with a large number of customers . a supplier with a single key - customer is a firm that is to a large extent ( 50 % ) dependent on a single customer . such a seller could be more fragile towards his customer s financial environment than a supplier who has no single key customer . however , the nature of the key - customer relations is not known to us . it could be the nature of small businesses , though it is obvious from figurel [ subfig : indegreevssales ] that the firms that have low in - degree correspond to a very large confidence interval ( meaning that there are firms that have very large sales , in the order of million euro , and yet a single customer ) . this might have a significant ( negative ) impact on the experiments that we are performing . namely , it might indicate that that there are some incidental events of sales that do not reflect the regular trading pattern . the vast number of key - customers in our dataset could also be connected with the fact that the data are preceding the economic crisis and it is probable that an increased number of liquidations , mergers etc happened which can be reflected in the incidental transactions . figure [ fig : truefalsekeycustimers ] displays two subgroups of the suppliers presented in figure [ fig : indegree ] . on the right panel are the suppliers that have a ` key - customer ' , and on the left appear the suppliers that do not have such a customer . we define a key - customer as one that purchases of at least 50% of the supplier s annual sales . out of the 671 firms in the sample there are 414 creditors with a key - customer and 235 without one . having a key - customer is a feature of the supplier and clearly all suppliers that have one customer ( @xmath24 ) qualify for it . we can appreciate that the range of firms that have a key - customer is dominated by suppliers with a small in - degree . however , we also find that suppliers with a large in - degree ( @xmath25 ) have a key - customer . the reason is that in many situations , the payment distribution to a single supplier is fat - tailed , i.e. the largest customer pays an order of magnitude more than the second largest one . a good example of this is a phone company : it has few very large customers and the majority are single - time walk - in clients . we can then expect that the suppliers with a key - customer will be subject to transmission of financial signals from their key customer , either directly by peer - to - peer interaction or indirectly by responding quickly to the factors that influence the key - customer . in this small subgroup of the firms , as the ratio of suppliers with a key customer to those without is 2:1 , we should expect to see a contagion effect that spans the firms with a key customer at the very least . we checked the contagion of the growth between the customers and the suppliers on the selected sample of suppliers with the matching 80 - 120% . we compared the actual measured growth in net - sales ( cash and credit ) from 2007 to 2008 of each supplier @xmath2 : @xmath26 against an estimated aggregated growth of purchases from all its customers , we assume that the change in purchases of a buyer @xmath3 between one year to the next is uniform across all his suppliers . so the estimated purchases of buyer @xmath3 from supplier @xmath2 in 2008 is the payment received in 2007 , @xmath27 , weighed by the trend in purchases of @xmath3 : @xmath28 summing over all supplier @xmath2 s customers gives an estimate of the growth in his sales from 2007 to 2008 : @xmath29 if we place ( [ eq : loggrowth ] ) on the y - axis and ( [ eq : logpredictgrowth ] ) on the x - axis we obtain a 2-dimensional scatter in which a dot will travel in the upward direction away from the origin to mark a firm with growing sales , and will travel to the right of the origin to designate growing _ predicted _ growth in sales . this scatter plot is shown in figure [ subfig:3bgrowth ] . assuming that the links in the 2007 network are constant across the three - year frame ( 2006 , 2007 , 2008 ) and that customer s purchases correspond to supplier s sales , the resulting pattern is expected to form a straight line through the origin with a slope of @xmath30 . 0.5 , recorded in the financial statements , versus our estimation of growth of the collective / aggregated orders from all his customers . the side / top bar plots in both subfigures are used to estimate the centroid of the clouds of points . figure [ subfig:3agrowth ] has a centroid in the first quadrant , defined by a positive value of the mean growth rate of sales , and a positive value of the mean growth rate of estimated demand . figure [ subfig:3agrowth ] has a centroid in the third quadrant , defined by a negative value of the mean growth rate of sales , and a negative value of the mean growth rate of estimated demand.,title="fig : " ] 0.5 , recorded in the financial statements , versus our estimation of growth of the collective / aggregated orders from all his customers . the side / top bar plots in both subfigures are used to estimate the centroid of the clouds of points . figure [ subfig:3agrowth ] has a centroid in the first quadrant , defined by a positive value of the mean growth rate of sales , and a positive value of the mean growth rate of estimated demand . figure [ subfig:3agrowth ] has a centroid in the third quadrant , defined by a negative value of the mean growth rate of sales , and a negative value of the mean growth rate of estimated demand.,title="fig : " ] we applied the same aggregation and estimation procedure to the prior period . applying similar reasoning , we write the growth in sales ( 2006 - 2007 ) as @xmath31 again , assuming that any change in the customer pool of a supplier between 2006 and 2007 is negligible , we write the aggregated growth in orders for that period as @xmath32 figure [ subfig:3agrowth ] displays this scatter plot . it is important to note the magnitude of the growth rates : for the majority of the suppliers , a small increase in orders to the supplier correspond with a small increase in the annual net - sales . this is the reason that the bulk of the points are close to the origin of the axes . i.e. the growth rate distribution is extremely narrow and deviations are dominated by the rare events . we should still expect that by the scaling nature of the growth rates @xcite , the rare events will render the same occurrence as the frequent ones . in the subplots of figure [ fig:3growth ] we added ( top and right side of each subplot ) a box - and - whisker plot on the sides to mark the univariate growth rate and predicted growth distributions . in these sidebars we can note two features : * ( 1 ) * that the estimated growth rate distribution is also narrow , i.e. shows similarity to the tent - shaped actual growth rate distributions , and that * ( 2 ) * the positions of the median growth rate values indicate that the centroid of the pattern in the period 2007/2006 ( panel [ subfig:3agrowth ] ) , the median value of both , the estimated and the real growth of all suppliers is positive . the growth in the estimated sales is larger then the measured one . in contaxt to that , in the next period , 2008/2007 ( panel [ subfig : gsyp1yvsyi ] ) , the centroid sits in the third quadrant , i.e. the meadian values of the box plots in [ subfig : gsyp1yvsyi ] show negative growth of both the estimated and the real sales of the suppliers.this is an evidence of the transition between an economic boom ( a positive growth in sales and demand ) in the first period and a bust ( negative growth in demand and sales ) in the second period . the results show that a decline of growth in sales and purchases in the second period ( an incline in the first ) occurred concurrently for suppliers and buyers . this simultaneous switch in the typical behavior is the effect of the credit crunch : the purchasing power of the customers has shrunk as most of the suppliers were not able to allow credit to their customers . however , figure [ fig:3growth ] fails to prove the correlation between the estimated and the real sales of customers . even by visual inspection , it is evident that the slope of @xmath30 is not present . the conclusion we draw from the plot is that in the selected sample little or no correlation exists between the aggregated changes in purchases by a supplier s customers and the growth rate in that supplier s sales . some companies , however , do not ` follow the crowd ' by going negative for several reasons apply , among which are regulatory actions and the sectoral behavior which we will analyze in the sequel . the sub - sample of the network ( @xmath33 ) consists of a heterogeneous set : there is a large variability in the connectivity pattern ( in - degree ) and in the net - sales . but the most relevant diversification factor ( and possibly related to the previous ones ) is that the companies come from different industries . about a half of the selected suppliers are in the manufacturing sector ( industrial classification numbers 15xx-37xx ) . the other half of the sample are in other industries , primarily in construction , wholesale , and transport . the degree distributions and the composition of the sample satisfying the 80 - 120% matching criterion are given in figure [ fig : indegreehist9panel ] . there is a striking difference in the connectivity between the sub - samples from different industrial sectors . the manufacturing sector ( d ) is the most populated , and the in - degree histogram of the companies within this sample shows that the companies follow the general in - degree distribution of figure [ fig : indegree ] . the second largest sector is wholesale trade and retail ( g ) . we note that the number of companies with large in - degree in sector g is exaggerated . comparing with the histogram of sector d , in g , the number of firms with @xmath34 is as large as that in d although the total number of firms is 4 times smaller . this is due to the different nature of their businesses as will be explained further in the text . the third largest sector is k. in our data set , most of the companies in this sector are it ( software ) companies . again , the accounting procedures in it are different from the ones in manufacturing . software companies will often provide services rather than goods . in order to understand the differences in growth rate correlation between the suppliers coming from different industrial sectors and their customers , the scatter plot in figure [ subfig : gsyp1yvsyi ] was split by sectoral subgroups . the subgroups are shown in figure [ fig : indegreehist9panel ] , and the split growth rate scatter plots are given in figure [ fig : firmsinview ] . in this figure , there are significant differences on the sectoral level . most remarkable is the characteristic behavior of the g sector ( the general name is ` wholesale and retail ' , but in our database it is mostly composed of wholesale firms ) . in this sector a significant number of firms has an estimated growth of orders close to one , but the measured growth in sales shows a notable variability away from no - growth . is given in the title of each panel . ] versus estimated growth of orders of all customers of a supplier @xmath2 , per industrial sector . the sectors are the same as in figure [ fig : indegreehist9panel ] ] another sector that shows uncommon behavior is sector f , construction . in this sector the growth response is opposite to the situation in sector g : there s a large variability in the expected growth of customer orders but the recorded growth of supplier sales shows very little deviation from a state of no - growth . the atypical in - degree distribution in the construction industry is discussed in miura et al . @xcite . in the japanese industrial business network , they ran a flow algorithm and observed a difference between in - degree and the ability to source or receive money . in construction , a large in - degree of a firm may not indicate large amounts of inflowing funds since at times they may relay funds via small single - customer subcontracting firms that dominate their payment distribution ( firms that exist only to operate in a single one project ) . in our network we also observe greater intra - industry trade in construction , more than would be expected by chance . however , as can be seen from figure [ fig : indegreehist9panel ] , sector f ( center panel ) is small in total number of firms . 0.5 eur . correlation coefficients for each rating / size category appear in the legend in the upper left corner of each plot.,title="fig : " ] 0.5 eur . correlation coefficients for each rating / size category appear in the legend in the upper left corner of each plot.,title="fig : " ] in order to factor out the possible cause of heterogeneity in behavior we chose suppliers that belong in a single industrial sector . being the largest in the sample , we chose a sub - sector of manufacturing , ` machinery and mechanical equipment ' . this sector is well represented in the italian trade network . we further tried to distill the effect of contagion by observing the geometric mean of growth rates in two consecutive periods 2007/6 and 2008/7 . this is commonly termed the compound annual growth rate ( cagr ) . the plots of cagr are given in figure [ fig : cagr2panel ] . there we placed the cagr of the supplier on the y - axis versus the predicted cagr from the sum of the purchases of his trade - credit customers on the x - axis . colors of the dots mark the rating class of the companies : abc = a in gray , b in red , and c in blue . the scatter plot in figure [ fig : cagr2panel ] may be interpreted in the following way : the companies in the _ first quadrant _ have managed to keep positive growth of both sales and orders in the two - year period . in both panels we observe that points in this quadrant are red . this means that the suppliers with rating class b ( 4 - 6 ) managed to maintain their sales and orders with the same partners over the two year period and through the financial crisis . although this result is in contrast with the hypotheses proposed in @xcite , it did not come as a surprise . the homophily measurement on the same data set that is discussed in kelman et al @xcite gives clear evidence that same rating - class firms have a greater probability to attach to each other , with a small tendency of the customer to create business with a higher rated supplier . in the _ second quadrant _ are locate suppliers that kept their growth in sales , opposing the downward trend in orders by their customers . negative estimated growth but positive real growth could happen for many reasons . the most obvious one draws from our assumption that the network is static . this is only an approximation and although in the manufacturing sector it is mostly a good one , in some cases it is possible that financial changes take place and trading partners change over time , especially the ones that are credit - constrained . according to delli gatti et al . @xcite , the origin of fluctuations is due to the ever - changing configuration of the network of heterogeneous firms and the entire dynamics is being shaped by financial variables . the number of companies in this quadrant is also very small ( only 4 firms with sales larger than 1 million eur out of only 13 firms in the sample ) . the suppliers in the _ third quadrant _ were affected by the crisis the most . both the orders from them and their own sales have decreased . in this range , surprisingly , we find the suppliers from the top rating class ` a ' . having a good credit - rating is related with the current ratio and liquidity of the firm as previously demonstrated by beaver @xcite and later by ohlson @xcite and others . also , having a good credit - rating corresponds with easy access to bank credit . i.e. the borrower is able to obtain a greater proportion of his collateral on loan , with a low interest . however , a good liquid position does not correspond with the state of the market in general , therefore these companies were not able to maintain their growth rate in the time frame of the crisis . last , in the _ fourth quadrant _ are the suppliers , which estimated aggregated orders grew while their sales decreased . there could be several reason for this , though this might also be a sign that the network was not stable as expected , so the customers have purchased from a new supplier . the most interesting outcome of these measurements is the correlation of the cagr with both rating and firm size : observing the pearson s correlation coefficients , given in the keys of the subfigures of figure [ fig : cagr2panel ] , we could see that the correlation between the compound annual growth rates in sales and orders are the highest ( @xmath35 ) in the case of sized companies . in the second place come the medium credit - rated companies , in the case of all firms , with the correlation coefficient of @xmath36 . the companies with rating class c did not exhibit correlations greater than @xmath37 either in the case of all firms or large firms . an additional support for the very good correlation between the expected growth of customers orders and the growth of sales in the companies with top ranking ( and therefore good access to credit ) , can be find in the literature . recently , several scholars @xcite , @xcite attempted to model and empirically confirm the effect of the 2007 - 2008 financial crisis on between - firm liquidity provision . searching for a causal effect of a credit - rationing by the banks , garcia - appendini & montoriol - garriga @xcite tested the hypothesis that firms with high liquidity levels before the crisis , increased the trade credit extended to other corporations and subsequently experienced better performance compared to ex - ante cash - poor firms . they conclude that trade credit taken by constrained firms increased during this period . therefore , they have intested in mainaining their customers and this might explain the high correlation which we measured . this work examines correlation of the estimated growth rates of customers orderd with the measured growth rate of suppliers sales in a many - to - one setting , where the customers of each supplier are responsible for at least 80% of the supplier s sales , measured at the beginning of the second time period . by establishing their trade connections in during one year period , a growth rate prediction is made from a combination of the trend in purchases by these customers , and the social structure in the supplier s neighborhood . this prediction is compared with the real growth in sales of the suppliers . results indicate the existence of growth rate contagion only inside the following restricted sub - selections of the manufacturing firms , and compound over two years : * ( 1 ) * the large firms with a - class credit - rating , and * ( 2 ) * suppliers of any size that have medium credit - rating ( b - class ) . this gives evidence that direct contagion of growth rates between customers and suppliers is sensitive to the following factors : * sectoral heterogeneity : in general , the industry of a supplier and a customer may or may not be the same . each industrial sector has a behavior that is characteristic to it due to the accounting procedures required in that industry . * in data coming from the bank , microscopic effects are still secondary compared to macroscopic effects , even during a state of crisis since the problematic customers may avoid approaching the bank . * missing data : while carefully considering numerical drift and retaining the overall statistics , the cleaned and filtered sample is still three orders of magnitude smaller than the number of firms in the full data set . we are also aware that performing the mesoscopic many - to - one approach in the rather limited ( especially in terms of network dynamics ) data has been based on the assumptiosn that may not be valid - especially in the case of the crisis . it is expected that the customers , in the case when they have to shrink their orders , would be selective in the choice of the suppliers and they would not shrink their orders uniformly as our model assumes . however in order to know how the customers make these decision we would need to have more data . as a final note , the macroscopic effect that was captured by the measurements in figure [ fig:3growth ] is realistic and its interpretation is supported by the official statistics on industrial production and business confidence in the given period : the statistics show that the italian industrial production peaked in 2007 and then declined , reaching a 10-year low in 2009 . the work presented in this article is partly supported by the project `` a large scale network analysis of firm trade credit '' , project grant ` in01100017 ` , institute for new economic thinking ( inet ) . bree ds , kelman g , ussher l , lamieri m , solomon s ( 2015 ) . too dynamic to fail - empirical support for an autocatalytic model of minskysfinancial instability hypothesis . to appear in the journal of economic interaction and coordination . tamura k , miura w , takayasu m , takayasu h , kitajima s , et al . ( 2012 ) estimation of flux between interacting nodes on huge inter - firm networks . world scientific , volume 16 , pp . miura w , takayasu h , takayasu m ( 2012 ) the origin of asymmetric behavior of money flow in the business firm network . european physical special topics 212 : 65 - 75 . watanabe h , takayasu h , takayasu m ( 2012 ) biased diffusion on japanese inter - firm trading network : estimation of sales from network structure . new j phys 14 . kiyotaki , nobuhiro , and john moore . credit cycles . . national bureau of economic research , 1995 . boissay f ( 2006 ) credit chains and the propagation of financial distress . technical report 573 , european central bank . url http://ideas.repec.org/p/ecb/ecbwps/20060573.html . petersen ma , rajan rg ( 1997 ) trade credit : theories and evidence . the review of financial studies 10 : 661691 . schumpeter ja ( 1976 ) capitalism , socialism and democracy . new york : allen & unwin . solomon s , richmond p ( 2002 ) stable power laws in variable economies ; lotka - volterra implies pareto - zipf . the european physical journal b - condensed matter and complex systems 27 : 257261 . challet d , solomon s , yaari g ( 2009 ) the universal shape of economic recession and recovery after a shock . economics : the open - access , open - assessment e - journal 3 . goldenberg j , libai b , solomon s , jan n , stauffer d ( 2000 ) marketing percolation . physica a : statistical mechanics and its applications 284 : 335347 . solomon s , golo n ( 2013 ) minsky financial instability , interscale feedback , percolation and marshall walras disequilibrium . accounting , economics and law 3 : 167260 . minsky hp ( 1982 ) the financial - instability hypothesis : capitalist processes and the behavior of the economy . in : kindleberger c , laffargue jp , editors , financial crises : theory , history and policy , cambridge university press . shnerb nm , louzoun y , bettelheim e , solomon s ( 2000 ) the importance of being discrete : life always wins on the surface . proceedings of the national academy of sciences 97 : 1032210324 . louzoun y , shnerb nm , solomon s ( 2007 ) microscopic noise , adaptation and survival in hostile environments . the european physical journal b - condensed matter and complex systems 56 : 141148 . klass os , biham o , levy m , malcai o , solomon s ( 2007 ) the forbes 400 , the pareto power - law and efficient markets . the european physical journal b 55 : 143 - 147 . choi y , douady r ( 2012 ) financial crisis dynamics : attempt to define a market instability indi- cator . quantitative finance 12 : 13511365 . huang zf , solomon s ( 2001 ) finite market size as a source of extreme wealth inequality and market instability . physica a : statistical mechanics and its applications 294 : 503513 . krugman p ( 1999 ) balance sheets , the transfer problem , and financial crises . international tax and public finance 6 : 459472 . bottazzi g , grazzi m , secchi a , tamagni f ( 2011 ) financial and economic determinants of firm default . journal of evolutionary economics 21 : 373406 . schwarzkopf y , axtell rl , farmer jd ( 2010 ) the cause of universality in growth fluctuations . arxiv nonlinear sciences e - prints . shenoy j , williams r ( 2011 ) customer - supplier relationships and liquidity management : the joint effects of trade credit and bank lines of credit . kelman g , bree d , manes e , lamieri m , golo n , et al . ( 2015 ) dissortative from the outside , assortative from the inside : social structure and behavior in the industrial trade network . in : proceedings of the 48th annual hawaii international conference on system sciences . ieee , computer society press , 2015 , p. 10 . delli gatti d , gaffeo e , gallegati m , giulioni g , palestrini a ( 2008 ) emergent macroeconomics : an agent - based approach to business fluctuations . beaver wh ( 1966 ) financial ratios as predictors of failure . journal of accounting research ( 4 ) empirical research in accounting : selected studies : 71 - 111 . ohlson ja ( 1980 ) financial ratios and the probabilistic prediction of bankruptcy . journal of accounting research 18 : 109 - 131 . garcia - appendini e , montoriol - garriga j ( 2012 ) firms as liquidity providers : evidence from the 2007 - 2008 financial crisis . available at ssrn 2023583 .	we propose a novel approach and an empirical procedure to test direct contagion of growth rate in a trade credit network of firms . our hypotheses are that the use of trade credit contributes to contagion ( from many customers to a single supplier - `` many to one '' contagion ) and amplification ( through their interaction with the macrocopic variables , such as interest rate ) of growth rate . in this paper we test the contagion hypothesis , measuring empirically the mesoscopic `` many - to - one '' effect . the effect of amplification has been dealt with in our paper @xcite . our empirical analysis is based on the delayed payments between trading partners across many different industrial sectors , intermediated by a large italian bank during the year 2007 . the data is used to create a weighted and directed trade credit network . assuming that the linkages are static , we look at the dynamics of the nodes / firms . on the ratio of the 2007 trade credit in sales and purchases items on the profit and loss statements , we estimate the trade credit in 2006 and 2008 . applying the `` many to one '' approach we compare such predicted growth of trade ( demand ) aggregated per supplier , and compare it with the real growth of sales of the supplier . we analyze the correlation of these two growth rates over two yearly periods , 2007/2006 and 2008/2007 , and in this way we test our contagion hypotheses . we could not find strong correlations between the predicted and the actual growth rates . we provide an evidence of contagion only in restricted sub - groups of our network , and not in the whole network . we do find a strong macroscopic effect of the crisis , indicated by a coincident negative drift in the growth of sales of nearly all the firms in our sample .
in this paper , we introduce new finite difference schemes for the hamilton - jacobi equation @xmath4 and prove rates of convergence . the hamilton - jacobi equation appeared recently as the continuum limit of nondominated sorting , which is widely used in scientific and engineering contexts @xcite . let us briefly describe the connection . let @xmath5 be _ i.i.d . _ random variables on @xmath6 with continuous density @xmath7 . let @xmath8 denote the elements in @xmath9 that are coordinatewise minimal . the set @xmath8 is called the _ first pareto front _ of @xmath10 , and the elements of @xmath8 are called _ pareto optimal _ or _ nondominated_. the _ second pareto front _ , denoted @xmath11 , is the set of minimal elements from @xmath12 , and the @xmath13 _ pareto front _ is defined as @xmath14 the process of sorting the set @xmath10 into pareto fronts , or nondominated layers , is called _ nondominated sorting _ , and is widely used in multi - objective optimization ( see @xcite and references therein ) , with recent applications to machine learning @xcite . it turns out that nondominated sorting is equivalent to finding the longest chain in a partially ordered set , which has a long history in probability and combinatorics @xcite . it was shown in @xcite that the pareto fronts converge almost surely in the limit as @xmath15 to the level sets of the unique nondecreasing is _ nondecreasing _ if @xmath16 is nondecreasing for all @xmath17 . ] viscosity solution of . figure [ fig : demo ] gives an illustration of this continuum limit . in @xcite , a fast algorithm called _ pde - based ranking _ was proposed for approximate nondominated sorting of large datasets . the basic idea is to estimate the density @xmath7 from a ( relatively small ) subset of @xmath5 , and then use the numerical solution of the hamilton - jacobi equation as an approximation of nondominated sorting . it was shown in @xcite that pde - based ranking is substantially faster than nondominated sorting in relatively low dimensions , while maintaining high levels of sorting accuracy . the numerical scheme for used in @xcite is based on backward finite differences and can be solved very efficiently in a single pass . due to the fact that information flows along coordinate axes in the definition of nondominated sorting , this scheme is upwind ( or monotone ) , and convergence of the scheme was established in @xcite . although the scheme used in @xcite is simple and efficient , it suffers from poor accuracy , with formal rates of convergence on the order of @xmath1 for a grid with spacing @xmath18 in dimension @xmath0 . in this paper , we propose two new and highly efficient finite difference schemes for solving . both schemes have a formal accuracy of @xmath2 when the solution is smooth , and we prove the usual @xmath3 rates in the context of non - smooth viscosity solutions . these schemes can be used to increase the accuracy of pde - based ranking @xcite _ without _ increasing computational complexity . alternatively , with these highly accurate schemes we can afford to use a coarser grid resolution , and thus we can reduce the computational complexity of pde - based ranking while maintaining high accuracy . this is particularly important in applications of nondominated sorting @xcite , which will benefit from highly accurate and efficient algorithms for sorting massive datasets . we detail the new schemes and our main results in the next section . we pose the hamilton - jacobi equation on a compact domain as follows : @xmath19^n\\ u & = 0 & & \text{on } \gamma , \end{aligned}\right\}\ ] ] where @xmath20^n \setminus ( 0,1]^n$ ] and @xmath21 . notice we have modified by taking the positive parts of the partial derivatives . this is necessary to obtain existence of a viscosity solution of , and is a well - known issue with viscosity solutions on boundaries of domains ( see @xcite ) . we elaborate on this briefly in section [ sec : comparison ] . we should mention that there is no loss of generality in considering the domain @xmath22^n$ ] in . indeed , we can make a simple scaling argument to transform the domain of into @xmath23 $ ] for any @xmath24 . let @xmath18 , @xmath25 , and @xmath26 . for @xmath27 set @xmath28 . we recall the numerical scheme for from @xcite : @xmath29^n_h\\ u_h(x ) & = 0 & & \text{if } x \in \gamma_h , \end{aligned}\right\}\ ] ] where @xmath30^n_h \to { \mathbb{r}}$ ] is the numerical solution and @xmath31 the solution @xmath32 of can be solved efficiently in a single pass , which is reminiscent of the fast marching @xcite and fast sweeping @xcite algorithms . in dimension @xmath33 , the scheme is quadratic and can be solved in closed form @xmath34 in dimensions @xmath35 , the scheme can be solved via any iteration method , such as a bisection search . convergence of to the viscosity solution of was established in @xcite . while is optimal in terms of computational complexity on a fixed grid , its accuracy is at best @xmath1 . to see this , consider the special case of @xmath36 and @xmath37 . the solution @xmath38 is smooth on @xmath39^n$ ] , and has a gradient singularity on the boundary @xmath40 . we can use the comparison principle @xcite for to show that in general @xmath41^n } f^\frac{1}{n } \leq u(x ) \leq n(x_1\cdots x_n)^\frac{1}{n } \sup_{[0,1]^n } f^\frac{1}{n}.\ ] ] therefore , the gradient singularity on @xmath40 is typical for solutions of whenever @xmath42^n } f > 0 $ ] . let @xmath43 . by the concavity of @xmath44 @xmath45 on the other hand , if @xmath46 then @xmath47 therefore , for any @xmath48^n_h$ ] such that @xmath46 for some @xmath17 we have @xmath49 setting @xmath50 and invoking the comparison principle for yields @xmath51 letting @xmath52 we find that @xmath53 therefore , the scheme makes an error on the order of @xmath1 in the immediate vicinity of the boundary @xmath40 . since @xmath38 is generally not smooth , the best theoretical rate that one can prove in the context of viscosity solutions is typically strictly worse than the formal rate ( see @xcite ) . in section [ sec : numerics ] , we show numerical results indicating that the @xmath54 convergence rate of @xmath1 is typically observed in practice . since the slow convergence rate is caused by a singularity in the gradient of @xmath38 on @xmath40 , it is natural to look for a transformation of @xmath38 that removes this singularity . with this in mind , we set @xmath55 where @xmath38 is the nondecreasing viscosity solution of . when @xmath56 we have @xmath57 , which is lipschitz continuous ( in fact smooth ) on @xmath22^n$ ] . in general , we prove in lemma [ lem : vexun ] that @xmath58^n)$ ] whenever @xmath59^n)$ ] . furthermore , @xmath60 is a viscosity solution of @xmath61^n\\ v & = 0 & & \text{on } \gamma . \end{aligned}\right\}\ ] ] since @xmath60 is lipschitz continuous , it is reasonable to suspect that a numerical scheme for would have a better convergence rate than . we therefore propose the following finite difference scheme for : @xmath62^n_h\\ v_h(x ) & = 0 & & \text{if } x \in \gamma_h , \end{aligned}\right\}\ ] ] where @xmath63^n_h \to { \mathbb{r}}$ ] . we take @xmath64 to be the largest solution of at each @xmath48^n_h$ ] . it is interesting to note that when @xmath7 is constant , the solution of the scheme is @xmath65 , which is the _ exact _ solution of . the scheme can be solved efficiently in a single pass , similar to , and in dimension @xmath33 we have the closed form expression @xmath66 where @xmath67 in theorem [ thm : convpv ] , we prove convergence of when @xmath7 is continuous and nonnegative . our main result is the following convergence rate . [ thm : ratev ] suppose @xmath68^n)$ ] and @xmath69 . let @xmath70 be the solution of and let @xmath38 be the nondecreasing viscosity solution of . then @xmath71 and @xmath72_{1;[0,1]^n},\inf_{[0,1]^n}f\big)$ ] . notice that has a zeroth order term with the wrong sign for comparison principle arguments to hold . we can see this by observing that the method of vanishing viscosity takes the form @xmath73 since the standard proof of convergence rates for numerical approximations to viscosity solutions is based on the proof of the comparison principle @xcite , we can not directly apply these techniques to . our proof of theorem [ thm : ratev ] passes through an auxiliary problem , which actually suggests another numerical scheme for solving . based on our observation that @xmath37 is the viscosity solution of corresponding to @xmath36 , it is natural in more general settings to make the ansatz @xmath74 for some function @xmath75^n \to [ 0,\infty)$ ] . if this ansatz is correct , then @xmath76 would be a viscosity solution of @xmath77^n.\ ] ] it turns out that is well - posed within the class of bounded viscosity solutions _ without _ imposing a boundary condition . the boundary condition is actually encoded into the pde due to the fact that the term @xmath78 vanishes on @xmath79 . this suggests , for example , that we should expect @xmath80 . due to the degeneracy of the terms @xmath81 , there are in general infinitely many unbounded viscosity solutions of . we characterize @xmath76 from as the maximal bounded viscosity solution of , and we show in lemma [ lem : wlip ] that @xmath82^n)$ ] whenever @xmath59^n)$ ] . we propose the following numerical scheme for : @xmath83^n_h,\ ] ] where @xmath84^n_h \to [ 0,\infty)$ ] . here , we take @xmath85 to be the largest solution of at each @xmath86^n_h$ ] . we note that for @xmath86^n_h$ ] such that @xmath87 , the quantity @xmath88 is undefined . since this term appears in the form @xmath89 , its value is not used in the scheme . it is interesting to note that when @xmath7 is constant , the scheme gives the _ exact _ solution of . this scheme can be solved efficiently in a single pass , and in dimension @xmath33 the scheme can be solved in closed form @xmath90 where @xmath91 and @xmath92 since the zeroth order term in has the correct sign , we can use a modification of the standard convergence proof to establish the following convergence rate . [ thm : rate1 ] suppose that @xmath68^n)$ ] and @xmath93 . let @xmath94 be the solution of , and let @xmath76 be the maximal bounded viscosity solution of . then latexmath:[\[\label{eq : rate1 } where @xmath72_{1;[0,1]^n } , \inf_{[0,1]^n}f\big)$ ] . our proof of theorem [ thm : ratev ] proceeds by first showing ( in lemma [ lem : changeofvar ] ) that @xmath96 and then invoking theorem [ thm : rate1 ] . although both and have the same provable convergence rates , our numerical results presented in section [ sec : numerics ] suggest that in general has a better experimental convergence rate than . this can be explained by observing that @xmath38 can only have gradient singularities when transitioning from zero to a positive value . the transformation @xmath97 regularizes these gradient singularities _ anywhere _ in the domain @xmath22^n$ ] , and not just on the boundary @xmath40 . on the other hand , the transformation @xmath98 is designed only to capture singularities on the boundary @xmath40 . in section [ sec : numerics ] , we give an example of a discontinuous function @xmath7 for which exhibits a better convergence rate than for the reason outlined above . this paper is organized as follows . in section [ sec : comparison ] , we prove comparison principles for viscosity solutions of and . in section [ sec : conv ] , we use the barles - souganidis framework @xcite to prove convergence of the schemes and under the assumption that @xmath7 is continuous and nonnegative . in section [ sec : rates ] we prove theorems [ thm : ratev ] and [ thm : rate1 ] establishing rates of convergence for and when @xmath7 is positive and lipschitz . the proofs of the convergence rates require lipschitz estimates for the viscosity solutions of and . these are obtained in section [ sec : s3lip ] . in section [ sec : numerics ] , we show the results of numerical simulations comparing all three schemes . we first prove comparison principles for that will be utilized later in the convergence proofs . let us first briefly comment on the differences between and . as we mentioned in section [ sec : main ] , it is necessary to modify by taking the positive parts of @xmath99 when posing the pde on compact domains . to see why this is necessary , consider with @xmath36 in dimension @xmath33 without this modification : @xmath100 ^ 2\\ u & = 0 & & \text{on } \gamma . \end{aligned}\right\}\ ] ] this hamilton - jacobi equation of course has a classical solution @xmath101 that is smooth on @xmath39 ^ 2 $ ] . however , @xmath38 is not a viscosity solution of . to see this , let @xmath102 . then @xmath103 has a local maximum at @xmath104 relative to @xmath39 ^ 2 $ ] for every @xmath105 . since @xmath106 for @xmath107 , the viscosity subsolution property fails to hold at @xmath104 this is a well - known issue with viscosity solutions on boundaries of domains ( see @xcite ) . notice , however , that @xmath108 taking the positive parts of @xmath99 in gives the pde a useful monotonicity property that we will exploit in this paper . since it is useful to abstract this property , we make the following definition . [ def : directed ] let @xmath109 . we say @xmath110^n\times { \mathbb{r}}\times { \mathbb{r}}^n \to { \mathbb{r}}$ ] is _ directed _ if for all @xmath111^n\times{\mathbb{r}}$ ] @xmath112 if @xmath113 is directed , it is simple to construct a monotone ( or upwind ) numerical scheme using backward difference quotients . indeed , let us consider the scheme @xmath114^n_h,\ ] ] where @xmath115^n_h \to { \mathbb{r}}$ ] and @xmath116 to see that is monotone , suppose that @xmath117 and @xmath118 for @xmath119^n_h$ ] . then @xmath120 for all @xmath121 , and since @xmath113 is directed @xmath122 the following theorem is a direct consequence of this monotonicity and ( * ? ? ? * theorem 2.1 ) . [ thm : bs - conv ] suppose that @xmath113 is continuous and directed . let @xmath123 be solutions of satisfying @xmath124^n_h } } \|u_h(x)\| < \infty.\ ] ] then @xmath125 is a viscosity subsolution ( resp . supersolution ) of @xmath126 in @xmath127^n$ ] . when @xmath128 , there is a unique nondecreasing viscosity solution @xmath38 of @xcite . we now show that the restriction of @xmath38 to @xmath22^n$ ] is the unique nondecreasing viscosity solution of . this establishes the equivalence of and . [ lem : comp ] suppose @xmath129^n)$ ] is nonnegative . let @xmath130^n)$ ] be a viscosity subsolution of , and let @xmath131^n)$ ] be a nondecreasing viscosity supersolution of . then @xmath132 on @xmath22^n$ ] . let @xmath133 and set @xmath134 . since @xmath60 is nondecreasing , @xmath135 is a viscosity solution of @xmath136^n.\ ] ] the standard comparison argument based on doubling the variables ( see @xcite ) shows that @xmath137 on @xmath22^n$ ] . sending @xmath138 completes the proof . [ lem : equivalence ] let @xmath128 be nonnegative and let @xmath38 be the nondecreasing viscosity solution of . then the restriction of @xmath38 to @xmath22^n$ ] is the nondecreasing viscosity solution of . by lemma ( * ? * lemma 3.3 ) , the numerical solutions @xmath32 of satisfy the estimate @xmath139 for all @xmath140^n_h$ ] , where @xmath141^n}f)$ ] . combined with theorem [ thm : bs - conv ] , this shows that @xmath32 converges uniformly on @xmath22^n$ ] to the unique nondecreasing viscosity solution of . by ( * theorem 3.4 ) , we also have @xmath142 uniformly on @xmath22^n$ ] , which completes the proof . in this section , we establish a comparison principle for . when @xmath129^n)$ ] , the function @xmath55 is a nondecreasing viscosity solution of ( see lemma [ lem : nonuniq ] ) . we will call @xmath97 the _ maximal _ viscosity solution of ( see lemma [ lem : vmaximal ] ) . since has a zeroth order term of the wrong sign for comparison to hold directly , we find that actually has infinitely many nondecreasing viscosity solutions . [ lem : nonuniq ] let @xmath129^n)$ ] be nonnegative , let @xmath143^n$ ] , and let @xmath38 be the nondecreasing viscosity solution of @xmath144\\ u & = 0 & & \text{on } \prod_{i=1}^n [ y_i,1]\setminus \prod_{i=1}^n ( y_i,1 ] , \end{aligned}\right\}\ ] ] and extend @xmath38 to @xmath22^n$ ] by setting @xmath145 for @xmath146^n \setminus\prod_{i=1}^n [ y_i,1]$ ] . then @xmath55 is a nondecreasing viscosity solution of . let @xmath147^n$ ] and let @xmath148^n)$ ] such that @xmath149 has a local maximum at @xmath150 . we also assume that @xmath151 . if @xmath152 then since @xmath38 is nondecreasing , we see that @xmath153 for all @xmath154 that are coordinatewise less than @xmath150 . it follows that @xmath155 for all @xmath17 , and the subsolution property is trivially satisfied . if @xmath156 , then @xmath157 and therefore @xmath158 $ ] . setting @xmath159 we find that @xmath160 has a local maximum at @xmath150 . therefore @xmath161 since @xmath162 , this becomes @xmath163 which verifies the subsolution property . the proof of the supersolution property is similar . the lack of uniqueness of nondecreasing viscosity solutions of indicates that we can not expect a comparison principle to hold for arbitrary sub- and supersolutions of . however , we show in the following lemma that every subsolution is bounded above by @xmath97 . this turns out to be sufficient to prove convergence of . [ lem : vmaximal ] let @xmath129^n)$ ] be nonnegative and let @xmath164^n)$ ] be a nonnegative viscosity subsolution of . then @xmath165 on @xmath22^n$ ] , where @xmath38 is the nondecreasing viscosity solution of . we define @xmath166 . let @xmath147^n$ ] and @xmath148^n)$ ] such that @xmath167 has a local maximum at @xmath150 and @xmath168 . letting @xmath169 , we see that @xmath170 has a local maximum at @xmath150 and therefore @xmath171 since @xmath172 , we have @xmath173 if @xmath156 then @xmath174 if @xmath152 , then since @xmath60 is nonnegative , @xmath175 for all @xmath17 , which verifies . by lemma [ lem : comp ] , @xmath176 , where @xmath38 is the unique nondecreasing viscosity solution of . before establishing a comparison principle for , let us comment on the properties of solutions of . let @xmath129^n)$ ] be nonnegative and let @xmath38 be the nondecreasing viscosity solution of . by lemma [ lem : comp ] we have @xmath177^n}f^\frac{1}{n } \ \ \text { for all } x \in ( 0,1]^n.\ ] ] setting @xmath178^n)$ ] we have @xmath179^n}f^\frac{1}{n } \ \ \text { for all } x \in ( 0,1]^n.\ ] ] we also have that @xmath76 is a viscosity solution of that satisfies @xmath180^n\ ] ] in the viscosity sense for all @xmath17 . to see this , let @xmath147^n$ ] and @xmath181^n)$ ] such that @xmath182 has a local minimum at @xmath150 . we can also assume @xmath183 , so that @xmath184 in a neighborhood of @xmath150 . setting @xmath185 it follows that @xmath186 in a neighborhood of @xmath150 and @xmath187 . therefore @xmath188 has a local minimum at @xmath150 . since @xmath38 is nondecreasing , @xmath189 for all @xmath17 and @xmath190 a simple computation shows that @xmath191 and @xmath192 the subsolution property is verified similarly . we will call @xmath76 the _ maximal bounded _ viscosity solution of . notice the boundary condition @xmath193 on @xmath40 is only used to show that @xmath76 is bounded . indeed , it is clear that the argument above holds when @xmath38 is _ any _ viscosity solution of @xmath194^n.\ ] ] this yields an infinite number of unbounded viscosity solutions of . for instance , when @xmath36 the function @xmath195 for any @xmath196 is a viscosity solution of . taking @xmath197 gives the bounded viscosity solution of interest from . the following theorem characterizes this solution as the unique bounded viscosity solution of satisfying . [ thm : comparison - w ] assume that @xmath129^n)$ ] is nonnegative . let @xmath198^n)$ ] and @xmath199^n)$ ] be bounded viscosity sub- and supersolutions of , respectively , and suppose that @xmath200 satisfies in the viscosity sense for all @xmath17 . then @xmath201 on @xmath39^n$ ] . for @xmath202 , we define @xmath203^n\\ 0 , & \text{if } x \in \gamma.\end{cases}\ ] ] since @xmath198^n)$ ] is bounded , @xmath204^n)$ ] . similarly , @xmath205^n)$ ] . we first show that @xmath206 is a nondecreasing viscosity supersolution of . let @xmath207^n$ ] and @xmath181^n)$ ] such that @xmath208 has a local minimum at @xmath209 and @xmath210 . we define @xmath211^n.\ ] ] then it follows that @xmath212 has a local minimum at @xmath209 . since @xmath200 is a viscosity supersolution of we have @xmath213 since @xmath214 and @xmath200 satisfies we have @xmath215 therefore @xmath216 for all @xmath17 and @xmath217 this establishes that @xmath206 is a nondecreasing ( e.g. , see ( * ? ? ? * lemma 5.17 ) ) viscosity supersolution of . we can similarly show that @xmath218 is a viscosity subsolution of . the proof is completed by invoking lemma [ lem : comp ] . in this section we prove convergence of the schemes and under the assumption that @xmath7 is continuous and nonnegative . let @xmath18 and @xmath48^n_h$ ] . given values for @xmath219 , we define @xmath64 to be the largest solution of . if we let @xmath220 for @xmath221 and @xmath222 , then this is equivalent to finding the largest solution @xmath223 of @xmath224 we define @xmath225 since @xmath226 is always a solution of , it is easy to see that @xmath227 is a nonnegative real number for all @xmath228 . with these definitions , the solution @xmath70 of satisfies @xmath229 for @xmath230 and @xmath231^n_h.\ ] ] we shall refer to @xmath70 as the _ maximal _ solution of . we now establish some important properties of @xmath10 . [ lem : scheme ] let @xmath228 . then 1 . @xmath232 and @xmath233 if and only if @xmath234 , 2 . @xmath235 for all @xmath236 , 3 . if @xmath237 then @xmath238 whenever @xmath239 , and 4 . @xmath240 is nondecreasing in all variables . by symmetry , we may assume that @xmath241 . for ( i ) , we simply note that @xmath242 and @xmath243 . therefore there exists @xmath244 such that @xmath245 . it follows that @xmath246 . if @xmath247 , then clearly @xmath248 . conversely , suppose that @xmath237 . if @xmath249 then @xmath250 . if @xmath251 , then @xmath252 and hence @xmath253 . for ( ii ) if @xmath234 , then @xmath248 , and it is clear that @xmath254 for all @xmath255 . if @xmath237 , then by ( i ) , @xmath253 and @xmath256 . for any @xmath257 such that @xmath245 we have @xmath258 it follows that @xmath235 for all @xmath255 . this establishes ( ii ) . for ( iii ) , suppose first that @xmath249 . then @xmath259 and @xmath260 , from which ( iii ) immediately follows . if @xmath261 , then since @xmath237 we have @xmath262 for all @xmath263 . therefore there exists @xmath264 such that @xmath262 for @xmath265 . define @xmath266 clearly @xmath267 and @xmath268 . for any @xmath269 satisfying @xmath245 , we have by that @xmath270 . it follows that @xmath271 , which establishes ( iii ) . for ( iv ) we set @xmath272 by ( i ) , @xmath253 for every @xmath273 . therefore @xmath274 for all @xmath273 , and it follows from the implicit function theorem that the restriction of @xmath10 to @xmath275 is smooth . since @xmath256 we have @xmath276 differentiating in @xmath277 for @xmath278 we obtain @xmath279 since @xmath280 for all @xmath281 , we have @xmath282 it follows that @xmath283 . differentiating in @xmath284 we have @xmath285 as before , it follows that @xmath286 . therefore @xmath10 is strictly increasing on @xmath275 . if @xmath234 , then @xmath233 is nondecreasing . if @xmath287 , then @xmath260 is again nondecreasing . the continuity of @xmath10 establishes ( iv ) . [ rem : lips ] in the proof of lemma [ lem : scheme ] ( iv ) , we can use the inequality @xmath288 to find that @xmath289 for all @xmath273 and @xmath278 . since @xmath290 , we have the bound @xmath291 using the properties of @xmath10 from lemma [ lem : scheme ] we can establish a comparison principle for the scheme . [ thm : svcomp ] let @xmath18 and suppose @xmath292 . let @xmath293 be a subsolution of and let @xmath294 be a supersolution of satisfying @xmath295^n_h.\ ] ] then @xmath296 on @xmath22^n_h$ ] . we will prove the result by induction . we have @xmath297 for @xmath298 by definition . now let @xmath48^n_h$ ] and suppose that @xmath299 since @xmath293 is a subsolution of we have @xmath300 it follows from lemma [ lem : scheme ] ( ii ) that @xmath301 recalling and lemma [ lem : scheme ] ( iv ) we have @xmath302 the proof is completed by induction . [ rem : poscomp ] if @xmath7 is positive and @xmath60 is a supersolution of that is positive on @xmath39^n_h$ ] , then it follows from lemma [ lem : scheme ] ( iii ) that @xmath303^n_h.\ ] ] [ rem : s2exact ] notice that @xmath304^n}f \ \ \text { and } \ \ \underline{v}(x ) = ( x_1\cdots x_n ) \inf_{[0,1]^n}f\ ] ] are super- and subsolutions of , respectively . by theorem [ thm : svcomp ] and remark [ rem : poscomp ] @xmath305^n } f \leq v_h(x ) \leq ( x_1\cdots x_n ) \sup_{[0,1]^n}f \ \ \text { for all } x\in [ 0,1]_h^n.\ ] ] as a consequence , if @xmath306 then @xmath307 , which is _ exactly _ equal to the maximal viscosity solution of . we can now establish convergence of the scheme . [ thm : convpv ] let @xmath129^n)$ ] be nonnegative , and for each @xmath18 let @xmath70 denote the maximal solution of . then @xmath308 uniformly on @xmath22^n$ ] as @xmath309 , where @xmath60 is the maximal viscosity solution of . let @xmath310 by theorem [ thm : bs - conv ] and remark [ rem : s2exact ] , @xmath311^n)$ ] is a viscosity subsolution of . by lemma [ lem : vmaximal ] , @xmath312 . let @xmath32 be the solution of , and let @xmath313 . since @xmath314 is convex for @xmath105 and @xmath32 is nondecreasing , we have @xmath315 therefore @xmath316^n_h.\ ] ] by theorem [ thm : svcomp ] , @xmath317 on @xmath22^n_h$ ] . as in the proof of lemma [ lem : equivalence ] , we have that @xmath318 uniformly on @xmath22^n$ ] as @xmath319 . it follows that @xmath320 , which completes the proof . [ rem : onesided ] notice in the proof of theorem [ thm : convpv ] we showed that @xmath321 , where @xmath32 is the numerical solution of , and @xmath70 is the solution of . under the assumptions of theorem [ thm : ratev ] , this gives a one - sided convergence rate for of the form @xmath322 where @xmath38 is the nondecreasing viscosity solution of . when @xmath38 is concave , we can actually prove that @xmath323 ( see ( * ? ? ? * lemma 3.5 ) ) . recall that we defined the solution @xmath84^n_h\to { \mathbb{r}}$ ] of inductively by taking @xmath85 to be the largest solution of the polynomial equation defining at each @xmath154 . it is easy to see that @xmath324^n_h \ \text { and } \ i \in \{1,\dots , n\}.\ ] ] the following lemma shows that admits a comparison principle whenever the supersolution satisfies . [ lem : s3comp ] suppose @xmath7 is nonnegative . let @xmath325 and @xmath200 be sub- and supersolutions of , respectively , and suppose @xmath200 satisfies . then @xmath326 on @xmath22^n_h$ ] . let @xmath86^n_h$ ] be a maximum of @xmath327 , and suppose to the contrary that @xmath328 . then we have @xmath329 recalling we have @xmath330 which is a contradiction . we now prove convergence of . [ thm : convs3 ] let @xmath129^n)$ ] be nonnegative , and for @xmath18 let @xmath94 be the maximal solution of . then @xmath331 uniformly on @xmath39^n$ ] as @xmath309 , where @xmath76 is the maximal bounded viscosity solution of . let @xmath332 we can use the comparison principle from lemma [ lem : s3comp ] to show that @xmath333^n}f^\frac{1}{n } \leq w_h \leq \sup_{[0,1]^n}f^\frac{1}{n}.\ ] ] therefore @xmath333^n}f^\frac{1}{n } \leq \underline{w } \leq { { \overline{w } } } \leq \sup_{[0,1]^n}f^\frac{1}{n}.\ ] ] by theorem [ thm : bs - conv ] , @xmath334^n)$ ] is a bounded viscosity subsolution of and @xmath335^n)$ ] is a bounded viscosity supersolution of . furthermore , since @xmath94 satisfies another application of theorem [ thm : bs - conv ] shows that @xmath336 is a viscosity solution of for all @xmath17 . by theorem [ thm : comparison - w ] , @xmath337 , where @xmath76 is the maximal bounded viscosity solution of in order to obtain a rate of convergence , it is necessary to show that the either the numerical solution or the viscosity solution of the continuum equation is lipschitz continuous up to the boundary @xmath40 . since none of the hamiltonians considered in this paper are coercive , the standard textbook estimates @xcite do not apply . in dimension @xmath33 , we can prove the required lipschitz estimate for @xmath60 directly by differentiating and using a comparison principle . let us briefly sketch the argument , which is made rigorous at the level of the numerical scheme in theorem [ thm : vlipd2 ] . formally differentiating in the variable @xmath338 we have @xmath339 setting @xmath340 and noting that @xmath341 ^ 2 } f$ ] we have @xmath342_{1;[0,1]^2}\sup_{[0,1]^2 } f.\ ] ] consider using a comparison function of the form @xmath343 . since @xmath344 , the first term in can be ignored . furthermore , for @xmath345 ^ 2}f$ ] , the sign of the zeroth order term in is positive , which suggests that a comparison principle should hold , allowing us to show that @xmath346 for a possibly larger constant @xmath347 . of course , @xmath60 is not in general smooth enough to use this argument directly , so we need to regularize . one obvious approach would be the method of vanishing viscosity . however , there are some technical challenges with this approach , due to the sign of the zeroth order term in and the ( non - obvious ) fact that for positive viscosity , @xmath348 can fail to be nondecreasing . our approach is to apply the argument above at the level of the numerical scheme . since we showed convergence of in theorem [ thm : convpv ] , a lipschitz estimate on @xmath70 directly carries over to the maximal viscosity solution @xmath60 of . heading in this direction , we now recall some basic properties of finite differences . [ prop : finite - diff - properties ] for all @xmath349 we have 1 . @xmath350 and @xmath351 , 2 . @xmath352 , 3 . @xmath353 , we now give a preliminary regularity result for @xmath70 . [ thm : vlipd2 ] let @xmath33 , let @xmath68 ^ 2)$ ] be nonnegative , and let @xmath70 be any nondecreasing solution of . then for @xmath354 we have @xmath355_{1;[0,1]^2 } + 3\sup_{[0,1]^2}f \ \ \text { for all } x \in ( 0,1]^2_h.\ ] ] let us prove the result for @xmath356 ; the case of @xmath357 follows by symmetry . we extend @xmath70 to a function on @xmath358 ^ 2_h$ ] by setting @xmath229 for @xmath359 ^ 2_h\setminus [ 0,1]^2_h$ ] . with these definitions , we have @xmath360 for @xmath298 . since @xmath70 is nondecreasing , we therefore have @xmath361 ^ 2_h.\ ] ] we now apply the operator @xmath362 to the equation above and use proposition [ prop : finite - diff - properties ] ( ii ) to obtain @xmath363 for all @xmath48 ^ 2_h$ ] . setting @xmath364 and using proposition [ prop : finite - diff - properties ] ( i ) we have @xmath365 applying theorem [ thm : svcomp ] and remark [ rem : s2exact ] @xmath366_{1;[0,1]^2}\sup_{[0,1]^2}f=:m\ ] ] for all @xmath48 ^ 2_h$ ] , and @xmath367 for @xmath368 . we can assume @xmath369 , otherwise @xmath370 and is trivial . let @xmath371 ^ 2}f\right)(1 + x_2)$ ] and note that since @xmath70 is nondecreasing we have @xmath372 ^ 2}f.\ ] ] we claim that @xmath373 . assume by way of contradiction that @xmath374 ^ 2_h}(\psi - { { \overline{\psi } } } ) > 0 $ ] . since @xmath375 on @xmath376 , @xmath377 must attain its positive maximum at some @xmath48 ^ 2_h$ ] . it follows that @xmath378 for @xmath202 and @xmath379 . since @xmath380 and @xmath70 is nondecreasing , we can combine and to find that @xmath381 ^ 2}f \leq m,\ ] ] which is a contradiction to the positivity of @xmath382 . by cauchy s inequality @xmath383 ^ 2}f\right)(1 + x_2 ) \leq [ f]_{1;[0,1]^2 } + 3\sup_{[0,1]^2}f,\ ] ] for all @xmath48 ^ 2_h$ ] . since @xmath308 uniformly ( theorem [ thm : convpv ] ) , we can extend the lipschitz estimate in theorem [ thm : vlipd2 ] to . [ cor : vlipd2 ] let @xmath33 , let @xmath68 ^ 2)$ ] be nonnegative , and let @xmath60 be the maximal viscosity solution of . then there exists @xmath347 such that @xmath384_{1;[0,1]^2 } \leq c \\|f\\|_{c^{0,1}([0,1]^n)}.\ ] ] the proof of theorem [ thm : vlipd2 ] does not work in dimensions @xmath35 due to the additional nonlinearity of , which increases with @xmath0 . in section [ sec : second ] we study the auxiliary problem , which allows us to prove lipschitz regularity of the maximal viscosity solution of in arbitrary dimension ( see lemma [ lem : vexun ] ) , provided the strictly stronger condition @xmath59^n)$ ] holds . we expect theorem [ thm : vlipd2 ] to hold in dimensions @xmath35 under the natural condition @xmath68^n)$ ] , but we currently do not know how to prove this . in this section we prove lipschitz regularity and rates of convergence for the scheme . this establishes one of our main results ( theorem [ thm : rate1 ] ) . these results are used in section [ sec : ratess2 ] to prove similar rates of convergence for . throughout this section we set @xmath385 for @xmath386 . let us first give a formal argument suggesting a lipschitz estimate on the maximal bounded viscosity solution @xmath76 of . assuming @xmath76 is smooth and @xmath93 , we can take the logarithm of to obtain @xmath387 differentiating both sides in the variable @xmath277 and setting @xmath388 we have @xmath389 where @xmath390 if @xmath391 and @xmath392 if @xmath393 . notice that at a positive maximum of @xmath44 we have @xmath394 for all @xmath395 and hence @xmath396 where we employed the inequality of arithmetic and geometric means in the first inequality . this suggests that @xmath397^n } \frac{1}{n } f^{1-\frac{1}{n}}f_{x_i } = \sup_{[0,1]^n } \partial_{x_i } ( f^\frac{1}{n}).\ ] ] in theorem [ thm : lip ] and lemma [ lem : wlip ] , we make this argument rigorous by applying it to the numerical scheme . one obvious gap in the argument above is the existence of a positive maximum for @xmath398 . indeed , when @xmath76 is any of the infinitely many unbounded viscosity solutions of , @xmath398 either has no maximum value , or no minimum value , on @xmath39^n$ ] . this is a reflection of the fact that has no boundary condition . this issue is resolved by showing ( in proposition [ prop : extension ] ) that the numerical solution @xmath94 can be extended by projection to a solution of on the domain @xmath399^n_h$ ] . for @xmath400 , let us set @xmath401 [ prop : extension ] suppose @xmath7 is nonnegative , and let @xmath76 be a sub- ( resp . super- ) solution of . then @xmath402 is a sub- ( resp . super- ) solution of @xmath403^n_h,\ ] ] where @xmath404 . we claim that @xmath405^n_h , \ , i \in \{1,\dots , n\}.\ ] ] if @xmath406 the result is trivial . hence we may assume that @xmath407 , so that @xmath408 . then @xmath409 which establishes the claim . for @xmath410^n_h\setminus [ 0,1]^n_h$ ] , we have @xmath411^n_h$ ] and @xmath412 since @xmath413 , the result immediately follows . we also note that we can extend the product rule for finite differences in proposition [ prop : finite - diff - properties ] ( iii ) by induction as follows : [ lem : prod - rule ] for all @xmath414 and @xmath415 we have @xmath416 where the final summation is over all @xmath417 with @xmath418 . the proof of lemma [ lem : prod - rule ] is deferred to the appendix . we now establish a lipschitz estimate on the solution @xmath94 of . [ thm : lip ] suppose that @xmath68^n)$ ] is positive on @xmath22^n$ ] , and let @xmath94 be the maximal solution of . then @xmath419 for all @xmath420^n_h \ , : \ , x_k \leq 1-h\}$ ] . in particular @xmath421_{1;[0,1]^n}\sup_{[0,1]^n}f^{\frac{1}{n}-1}.\ ] ] for notational convenience , we will write @xmath76 in place of @xmath94 throughout the proof . by proposition , we can extend @xmath76 and @xmath7 to functions on @xmath399^n_h$ ] by setting @xmath422 and @xmath423 , and we have @xmath424^n_h.\ ] ] let @xmath425 for all @xmath410^n_h$ ] with @xmath426 . the proof is split into two steps . we first show that @xmath427 we may assume @xmath44 is positive somewhere , otherwise trivially holds . for any @xmath154 such that @xmath428 , we have @xmath429 . this implies that @xmath430 . likewise , whenever @xmath431 , @xmath432 and we have @xmath433 . it follows that @xmath44 attains its positive maximum value at some @xmath434 . therefore @xmath435 since @xmath436 , @xmath437 where @xmath438 by lemma [ lem : prod - rule ] @xmath439 using proposition [ prop : finite - diff - properties ] ( ii ) we have @xmath440 for @xmath86^n_h$ ] . invoking proposition [ prop : finite - diff - properties ] ( i ) and noting that @xmath441 , we deduce @xmath442 inserting into yields @xmath443 for all @xmath395 . combining this with and recalling that @xmath444 gives @xmath445 using the inequality of geometric and arithmetic means yields @xmath446 combining and establishes . we now show that @xmath447 we may assume that @xmath44 assumes negative values , otherwise is trivial . as in the first part of the proof , @xmath44 attains its negative minimum at some @xmath448 . therefore @xmath449 by , @xmath450 for all @xmath395 . combining this with lemma [ lem : prod - rule ] yields @xmath451 this completes the proof . [ rem : wlip ] since @xmath331 uniformly on @xmath22^n$ ] , it is an easy consequence of theorem [ thm : lip ] that @xmath82^n)$ ] and @xmath452_{1;[0,1]^n } \leq \frac{1}{\sqrt{n } } [ f]_{1;[0,1]^n}\sup_{[0,1]^n } f^{\frac{1-n}{n}},\ ] ] whenever @xmath68^n)$ ] is positive . a sharper result is contained in the following lemma . [ lem : wlip ] let @xmath7 be a nonnegative function for which @xmath59^n)$ ] , and let @xmath76 be the maximal bounded viscosity solution of . then @xmath82^n)$ ] and @xmath453_{1;[0,1]^n } \leq \sqrt{n}[f^\frac{1}{n}]_{1;[0,1]^n}.\ ] ] we first assume that @xmath454^n)$ ] and @xmath93 . for @xmath18 , let @xmath94 be the maximal solution of . since @xmath454^n)$ ] , there exists a constant @xmath347 such that @xmath455 for all @xmath17 and all @xmath456^n_h \ , : \ , x_i \leq 1-h\}$ ] . by theorem [ thm : lip ] @xmath457_{1;[0,1]^n } + ch.\end{aligned}\ ] ] the opposite inequality is obtained similarly , and we find that @xmath458_{1;[0,1]^n } + ch,\ ] ] for all @xmath17 and all @xmath459 . the estimate follows from the uniform convergence @xmath331 as @xmath319 . we now suppose that @xmath7 is nonnegative and @xmath59^n)$ ] . extend @xmath7 to a function on @xmath460 by setting @xmath461 for @xmath462^n$ ] , where @xmath463^n$ ] is the closest point projection . since @xmath464 is 1-lipschitz , @xmath465_{1,{\mathbb{r}}^n } = [ f^\frac{1}{n}]_{1,[0,1]^n}$ ] let @xmath466 denote the standard mollifier , and define @xmath467 it is easy to verify that @xmath468^n)$ ] , @xmath469 uniformly on @xmath22^n$ ] as @xmath470 , and @xmath471_{1;[0,1]^n } \leq [ f^\frac{1}{n}]_{1;[0,1]^n}.\ ] ] let @xmath472 denote the maximal bounded viscosity solution of corresponding to @xmath473 . since @xmath473 is smooth and positive , the argument above yields @xmath474_{1;[0,1]^n } \leq \sqrt{n}[(f^{\varepsilon})^\frac{1}{n}]_{1;[0,1]^n } \stackrel{\eqref{eq : moll}}{\leq}\sqrt{n } [ f^\frac{1}{n}]_{1;[0,1]^n}.\ ] ] by the comparison principle for ( theorem [ thm : comparison - w ] ) and standard results on viscosity solutions @xcite , we have @xmath475 uniformly on @xmath22^n$ ] as @xmath470 . therefore @xmath476_{1;[0,1]^n } \leq \sqrt{n } [ f^\frac{1}{n}]_{1;[0,1]^n},\ ] ] which completes the proof . [ rem : tight ] the regularity result in lemma [ lem : wlip ] is tight , and can not be significantly generalized . for intance , if @xmath477 , then @xmath7 is smooth , but @xmath478 is not lipschitz on @xmath22^n$ ] . the corresponding solution of is @xmath479 , which also fails to be lipschitz on @xmath22^n$ ] . note that the solution of is @xmath480 , which is smooth on @xmath22^n$ ] . we can now extend lipschitz regularity to viscosity solutions of . [ lem : vexun ] suppose @xmath7 is a nonnegative function for which @xmath59^n)$ ] , and let @xmath60 be the maximal viscosity solution of . then @xmath58^n)$ ] and @xmath481_{1;[0,1]^n } \leq c \\|f^\frac{n-1}{n}\\|_{l^\infty([0,1]^n)}\\|f^\frac{1}{n}\\|_{c^{0,1}([0,1]^n)}.\ ] ] notice that we can write @xmath482^n,\ ] ] where @xmath76 is the maximal bounded viscosity solution of . therefore @xmath483_{1;[0,1]^n } \leq [ w^n]_{1;[0,1]^n } + [ x_1\cdots x_n]_{1;[0,1]^n}\sup_{[0,1]^n } f.\ ] ] the estimate follows from lemma [ lem : wlip ] and the inequality @xmath484_{1;[0,1]^n } \leq n\\|w\\|^{n-1}_{l^\infty([0,1]^n)}[w]_{1;[0,1]^n}.\qedhere\ ] ] we expect lemma [ lem : vexun ] to hold under the weaker condition that @xmath68^n)$ ] , but we do not currently know how to prove this . since we know the scheme converges , we can prove a result analogous to proposition [ prop : extension ] regarding extensions of viscosity solutions of . [ prop : extension2 ] suppose @xmath129^n)$ ] is nonnegative and let @xmath76 be the maximal bounded viscosity solution of . then @xmath485 is a viscosity solution of @xmath486^n,\ ] ] where @xmath413 . let @xmath84^n_h \to{\mathbb{r}}_+$ ] denote the solution of for @xmath18 . by theorem [ thm : convs3 ] , @xmath331 uniformly on @xmath22^n$ ] . set @xmath487 for @xmath410^n_h$ ] . by proposition [ prop : extension ] , @xmath488 is a solution of the discrete scheme @xmath489^n_h.\ ] ] since @xmath490 uniformly on @xmath399^n$ ] , theorem [ thm : bs - conv ] shows that @xmath491 is a viscosity solution of . before proving a convergence rate for , we need another preliminary proposition . [ prop : h1 ] for @xmath492 such that @xmath493 @xmath494 since @xmath493 we can write @xmath495 in a sufficiently small neighborhood of @xmath496 . in this neighborhood @xmath497 where we used the inequality of arithmetic and geometric means . we now give the proof of theorem [ thm : rate1 ] . we extend @xmath7 to a function on @xmath399^n$ ] by setting @xmath423 . by proposition [ prop : extension ] we can extend @xmath94 to a function on @xmath399^n_h$ ] satisfying @xmath498^n_h,\ ] ] by setting @xmath499 . similarly , by proposition [ prop : extension2 ] we can extend @xmath76 to a function on @xmath399^n$ ] by setting @xmath422 , and @xmath76 is a viscosity solution of @xmath500^n.\ ] ] we will show that @xmath501 the opposite inequality is proved similarly . we can assume that @xmath502^n_h } ( w_h - w)>0 $ ] . for @xmath503 , @xmath410^n_h$ ] and @xmath504^n$ ] , set @xmath505 since @xmath499 , @xmath506 and @xmath507 , there exists @xmath508^n_h$ ] and @xmath509^n$ ] such that @xmath510^n_h \times ( -\infty,1]^n } \phi.\ ] ] let @xmath511 . then @xmath512 attains its maximum at @xmath513 with respect to the grid @xmath399^n_h$ ] . therefore @xmath514 for all @xmath17 . since @xmath508^n_h$ ] and @xmath113 is directed , @xmath515 since @xmath516 , we see that @xmath517 where we invoked remark [ rem : wlip ] in the last inequality . since @xmath518 we have @xmath519 where @xmath520 . by , @xmath521 . therefore , we can invoke the local lipschitzness of @xmath113 in all variables to obtain @xmath522 combining this with yields @xmath523 by we also have @xmath524 subtracting these equations yields @xmath525 since @xmath526^n_h } ( w_h - w)>0,\ ] ] we have @xmath527 . since @xmath528 , we can invoke proposition [ prop : h1 ] to find that @xmath529 choosing @xmath530 we have @xmath531^n_h } ( w_h - w ) \leq w_h(x_\alpha ) - w(y_\alpha ) \leq c\sqrt{h}. \qedhere\ ] ] we now prove a convergence rate for . this follows directly from theorem [ thm : rate1 ] and the following result . [ lem : changeofvar ] suppose @xmath68^n)$ ] is positive . let @xmath94 be the solution of , and let @xmath70 be the maximal solution of . then @xmath532_{1;[0,1]^n } , \inf_{[0,1]^n}f\big).\ ] ] set @xmath533 for @xmath86^n_h$ ] . by lemma [ lem : prod - rule ] , remark [ rem : wlip ] , and the inequality @xmath534 , we have @xmath535 here , we use the notation @xmath2 to denote function of @xmath154 that is bounded uniformly by @xmath536 with @xmath537_{1;[0,1]^n},\inf_{[0,1]^n}f\big).\ ] ] since @xmath94 satisfies we have @xmath538 it follows that @xmath539 therefore , there exists a constant @xmath540 such that @xmath541 let @xmath542 and set @xmath543 . then @xmath544 since @xmath545 and @xmath93 , we can use theorem [ thm : svcomp ] and remark [ rem : poscomp ] to conclude that @xmath546 on @xmath22^n_h$ ] . likewise , there exists a constant @xmath540 such that @xmath547 let @xmath548 , and take @xmath18 small enough so that @xmath549 . define @xmath550 . then @xmath551 since @xmath552 on @xmath39^n_h$ ] we have @xmath553 on @xmath22^n_h$ ] . we now have the proof of theorem [ thm : ratev ] . let @xmath94 be the solution of and let @xmath76 be the maximal bounded viscosity solution of . by theorem [ thm : rate1 ] , @xmath554^n_h.\ ] ] by lemma [ lem : changeofvar ] we have @xmath555 for all @xmath86^n_h$ ] and @xmath18 sufficiently small . the inequality follows from the lower bounds @xmath556^n } f \ \text { for all } x \in [ \delta,1]^n_h.\qedhere\ ] ] in this section , we discuss the bisection search method for solving , and we show the results of numerical simulations comparing the three schemes . in dimension @xmath33 , all three schemes have a closed form solution ( see , , and ) , and can therefore be efficiently solved up to machine precision . in dimensions @xmath35 , it is necessary to use an iterative method to approximate the solution of the scheme . due to the monotonicity properties of all three schemes , it is natural and efficient to use the bisection method , or slight variations thereof . the bisection method requires an initial interval to start the search , and a specified tolerance for terminating the bisections . the main idea is to set the tolerance to match the truncation error of each scheme , so that the error from the bisection method is no greater than the error incurred by using finite differences . we describe the details for each scheme below . for the scheme ( s1 ) , we use the bisection search method at each @xmath48^n_h$ ] to construct an approximation @xmath557 of the solution @xmath32 of . for the initial interval , we use @xmath558,\ ] ] and with a slight modification of the bisection method , we can ensure that the solution @xmath557 satisfies @xmath559^n_h,\ ] ] and @xmath560 for @xmath298 . by comparison for we have @xmath561^n_h,\ ] ] where @xmath32 is the exact solution of . since this error is no worse than the truncation error from using first order finite differences , it is unnecessary to continue the bisection search once is satisfied . for the scheme ( s2 ) , we use the same bisection search technique to find @xmath562 satisfying @xmath563^n_h,\ ] ] with @xmath564 for @xmath298 . by remark [ rem : lips ] , we may take the initial interval to be @xmath565.\ ] ] letting @xmath70 denote the exact solution of we have by theorem [ thm : svcomp ] and remark [ rem : s2exact ] that @xmath566 as before , this error is smaller than the truncation error introduced by discretizing , which justifies our choice of stopping condition for the bisection search . finally , for the scheme , we use the bisection method to obtain @xmath567 satisfying @xmath568^n_h.\ ] ] we take the initial interval for the search to be @xmath569,\ ] ] where @xmath570 by lemma [ lem : s3comp ] we have @xmath571 where @xmath94 is the exact solution of . we show here the results of some numerical simulations comparing the three schemes , , and . we first consider @xmath572 the solution of is given by @xmath573 the level sets of @xmath218 are illustrated in figure [ fig : demo1 ] . tables [ f1n2 ] , [ f1n3 ] and [ f1n4 ] show the @xmath54 errors and orders of convergence for @xmath574 in dimensions @xmath33 , @xmath575 , and @xmath576 , respectively . for each scheme , the @xmath54 errors are measured by how well the schemes approximate the viscosity solution @xmath218 of . .rates of convergence for @xmath574 in dimension @xmath33 . [ cols="^,^,^,^,^,^,^ " , ] the author is grateful to selim esedoglu and lawrence c. evans for valuable discussions about this work . we give the proof of lemma [ lem : prod - rule ] here . we proceed by induction . the base case of @xmath577 is exactly proposition [ prop : finite - diff - properties ] ( ii ) . assume holds for some @xmath415 . then by the inductive hypothesis and base case @xmath578 notice that @xmath579 where in the final summation , @xmath580 . we also have @xmath581 combining this with and we have @xmath582 this verifies for @xmath583 . the proof is completed by mathematical induction .	non - dominated sorting arranges a set of points in @xmath0-dimensional euclidean space into layers by repeatedly removing the coordinatewise minimal elements . it was recently shown that nondominated sorting of random points has a hamilton - jacobi equation continuum limit . the obvious numerical scheme for this pde has a slow convergence rate of @xmath1 . in this paper , we introduce two new numerical schemes that have formal rates of @xmath2 and we prove the usual @xmath3 theoretical rates . we also present the results of numerical simulations illustrating the difference between the formal and theoretical rates .
the shape of the fermi surface determines the kinetic and equilibrium properties of the electron gas in the crystal matter @xcite as well as the dynamics of a single electron in crystal @xcite . the shape of this surface can be reconstructed drastically by a high magnetic field @xcite . however , a uniform magnetic field changes totally the translation properties of an electron in the crystal lattice : the appearance of a new length scale ( the magnetic length ) leads to the famous phenomena related to the `` commensurability incommensurability '' transitions . in particular , a fractal structure arises in the diagram describing the dependence of the two - dimansional electron spectrum on the magnetic flux ( azbelhofstadter butterfly @xcite ) it was shown recently that this fractal structure is an inherent characteristic not only of two - dimensional spectral diagrams but also of the `` flux energy '' and `` angle energy '' diagrams for a three - dimensional magneto - bloch electron @xcite . the translation symmetry of the bloch electron in a uniform magnetic field is defined by the magnetic translation group @xcite , which has more complicated structure in comparison with the translation group without the field . in particular , the ordinary quasi - momentum is no longer a conserved quantity , therefore , at high magnetic fields a modification of the definition of the fermi surface is required . such a modification is given by v. ya . demikhovskii with coworkers @xcite . in papers @xcite , the construction and investigation of the fermi surfaces in the magnetic brillouin zone have been performed by means of the tight - binding approximation . in particular , it was shown that the change of the topology of the fermi surface under the influence of the magnetic field can cause the metal semiconductor transition . the procedure of obtaining the tight - binding schrdinger operator ( also called the harper operator ) leads to a rather rough approximation of the initial hamiltonian ; for example , in the 2d case it is known that the `` flux energy '' diagram for the original periodic landau operator differs essentially from the classical hofstadter butterfly @xcite . in this connection , it is interesting to build and investigate the fermi surfaces in the magnetic brillouin zone for the periodic landau operator without additional approximations . here we consider the case of the landau operator perturbed by a lattice @xmath0 of short - range scatterers . to get an explicitly solvable model we deal here with a limiting case of the short - range potential , namely , with the zero - range potential @xcite , also called `` point potential '' . we stress , that the method of zero - range perturbations for obtaining the fermi surfaces in 3d crystals without magnetic fields was successfully used earlier @xcite . on the other hand , the spectral properties of the three - dimensional landau operator perturbed by a periodic point potential was studied in papers @xcite . in particular , in the paper @xcite , a version of the magneto - bloch analysis was proposed which makes it possible to get an explicit form of the dispersion relation for the problem we consider here . we start with the quantum - mechanical hamiltonian @xmath1 of a single electron subject to a uniform magnetic field @xmath2 @xmath3 where @xmath4 and @xmath5 are the charge and mass of the electron , respectively , and @xmath6 is the vector potential of the field @xmath7 . we choose the symmetric gauge : @xmath8 ; in this case the form of @xmath1 in any cartesian coordinate system with the @xmath9-axis directed along the field @xmath7 is independent of the choice of the other coordinate axes . we use the standard notations @xmath10 ( the cyclotron frequency ) , @xmath11 ( the magnetic length ) , and @xmath12 ( the magnetic flux quantum ) . throughout the paper we denote by @xmath13 , @xmath14 , the vector of the magnetic flux density : if @xmath15 and @xmath16 are nonzero vectors , then @xmath17 is the total number of the magnetic flux quanta through the parallelogram spanned by @xmath15 and @xmath18 . note that @xmath19 . all important physical information concerning the hamiltonian @xmath1 is contained in the green function @xmath20 at energy @xmath21 defined as @xmath22 . it is known ( see , e.g. , ref . @xcite ) that this function can be represented in the following forms : @xmath23 where @xmath24\,,\ ] ] @xmath25 } { \sqrt{l+1/2-e/\hbar\omega}}\,l_l({\bf r}_\bot^2/2l_m^2)\,,\end{aligned}\ ] ] @xmath26 } { \left(1-e^{-t}\right)\exp\left[(1/2-e/\hbar\omega)t\right ] } \frac{dt}{\sqrt{\pi t}}\,.\ ] ] here @xmath27 is the projection of @xmath28 on the direction of the field @xmath29 and @xmath30 ; the function @xmath31 is the @xmath32-th laguerre polynomial . it is convenient to decompose @xmath20 into a sum of singular and regular ( with respect to the limit @xmath33 ) parts : @xmath34 } { \|{\bf r}-{\bf r}'\|}+ g_0^{\rm reg}({\bf r},{\bf r}';e)\,.\ ] ] note that @xmath35 is well defined at @xmath36 and @xmath37 were @xmath38 is the generalized riemann zeta - function ( in other words , the hurwitz zeta - function ) @xcite . now we consider a crystalline lattice @xmath39 with nodes @xmath40 ; the bravais lattice of the crystal we denote by @xmath0 . choose a basis @xmath15 , @xmath18 , @xmath41 of @xmath0 , then a basic cell @xmath42 of @xmath39 ( the wigner seitz cell ) is fixed : @xmath43 . if @xmath44 is the set of nodes of @xmath39 containing in @xmath42 , then @xmath45 ; this means that each vector @xmath46 can be represented in the form @xmath47 , where @xmath48 , @xmath49 ( note that such a representation is unique ) . without loss of generality we will suppose that @xmath50 . we choose the potential @xmath51 of the crystalline lattice @xmath39 in the form @xmath52 where @xmath53 is the confinement potential near the node @xmath54 . to obtain an explicitly solvable model , we pass to the zero - range limit for @xmath55 . this means that we consider @xmath55 as the limit of potentials of the form @xmath56 , where @xmath57 outside a small sphere of radius @xmath58 centered at zero , the coupling constant @xmath59 is of order @xmath58 , and @xmath60 . at the limit @xmath61 the potential @xmath55 is characterized by one parameter only , namely , by the scattering length @xmath62 , which is related to the binding energy of the ground state for @xmath55 by @xmath63 ( see ref . what is important , at the zero - range limit the green function @xmath64 of @xmath65 has the following explicit expression in terms of the green function @xmath66 of @xmath1 ( see ref . @xcite ) : @xmath67 where @xmath68 is the matrix inverse to the infinite matrix @xmath69 with elements @xmath70\delta_{\boldsymbol{\gamma},\boldsymbol{\gamma}\,{}'}+ ( 1-\delta_{\boldsymbol{\gamma},\boldsymbol{\gamma}\,{}'})g_0(\boldsymbol{\gamma},\boldsymbol{\gamma}\,{}';e)\,.\end{aligned}\ ] ] the hamiltonian @xmath71 in some sense is a three - dimensional analogue of the one - dimensional kronig penney hamiltonian . the spectral properties of @xmath65 can be extracted from the properties of invertibility of the infinite matrix @xmath69 . namely , at least for @xmath72 the energy @xmath21 belongs to the spectrum of @xmath65 if the matrix @xmath69 is not invertible . under so - called `` rationality condition '' @xcite , we can reduce the problem to invert the infinite matrix @xmath69 to a problem of finite - dimensional algebra . according to @xcite , we call the field @xmath7 _ rational _ with respect to @xmath0 if for some basis @xmath15 , @xmath16 , @xmath41 of @xmath0 , all the numbers @xmath73 are rational . it is clear that every @xmath7 obeys this condition after an appropriate infinitely small change of direction , therefore , the rationality condition imposes no essential restrictions on @xmath74 . if @xmath7 is rational with respect to @xmath0 , then a basis @xmath75 of @xmath0 can be chosen in such a way that @xmath76 and @xmath77 . denote @xmath78 , where @xmath79 and @xmath80 are coprime positive integers ; in our approach the crucial role is played by the following @xmath81 matrix @xmath82 , where @xmath83 is the number of nodes in @xmath44 : @xmath84\times\\\ ] ] @xmath85 @xmath86\,.\end{aligned}\ ] ] here @xmath87 denotes the pair @xmath88 with @xmath48 , @xmath89 and @xmath90 is the magnetic quasimomentum . the magnetic brillouin zone is defined by the inequalities @xmath91 , @xmath92 , @xmath93 and , in this zone , the dispersion relation for @xmath65 reads @xmath94 for a fixed @xmath95 , equation ( [ kp ] ) has infinitely many solutions @xmath96 , @xmath97 ; each @xmath96 is an @xmath80-fold degenerate eigenvalue of @xmath65 . dispersion laws @xmath98 are continuous functions of @xmath95 ; the set of all values of @xmath96 is an interval @xmath99 . in a natural way , the intervals @xmath99 form clusters called _ magnetic bands _ or _ landau bands _ , each magnetic band contain @xmath100 adjacent intervals @xmath99 ( _ subbands _ ) . for the fermi energy @xmath101 , the equation @xmath102 determines a periodic surface in the quasimomentum space ( the _ fermi surface _ ) . using eqs . ( [ kpr ] ) ( [ fe ] ) we analyze numerically the shape of the fermi surfaces in a crystal with the simple - cubic lattice for various directions of the field @xmath29 . the lattice constant @xmath103 nm is chosen relevant to the geometric parameters for the 3d regimented quantum dot superlattice considered recently in ref . @xcite . as to the scattering length , we put @xmath104 nm , this corresponds to the binding energy @xmath105 mev . let the magnetic field @xmath7 is directed along the edge @xmath41 of the cubic elementary cell and @xmath106 ; this value of @xmath107 corresponds to the field strength @xmath108 t ( note in this connection that it was reported on the possibility to achieve the magnetic field strength 28 mg @xcite ) . in the considered case , the lowest magnetic band of the spectrum has the form @xmath109 $ ] , where @xmath110 . the fermi surfaces for various energy values from this band are depicted in fig . 1 . the warped ellipsoids in the panels ( a ) and ( d ) are the electron and hole surfaces respectively ; in this case _ there exist only closed trajectories of the charged particle in the @xmath111-space . _ other topologies of the fermi surface are shown in panels ( b ) and ( c ) ; according to the definition from @xcite these surfaces have the topology of range 2 and 3 respectively . surprisingly , an inclination of @xmath7 to the edge of the cubic cell causes a profound reconstruction of the fermi surfaces . as an example we consider @xmath112 . in the new basis @xmath113 , @xmath114 , @xmath115 , we have @xmath116 and @xmath117 . therefore , the magnetic brilluoin zone is defined by the inequalities @xmath118 , @xmath119 , and the lowest magnetic band splits into 5 subbands . 2 shows variation of the fermi surface as the fermi energy @xmath101 varies in the first subband @xmath120 $ ] . the electron and hole fermi surfaces have the shapes of warped cylinders ( panels ( a ) and ( d ) respectively ) or highly twisted sheets ( panels ( b ) and ( c ) ) . it is particularly remarkable that in this case the surfaces _ are open in the @xmath121-direction_. in going from the first subband of the lowest magnetic band to the higher subbands , the shape of the fermi surface changes ; examples of this change are shown on fig . 3 , where the fermi energy @xmath101 lies in the second and fourth subbands respectively . in the examples above , the filed @xmath7 is parallel to a face of the cubic cell . we have built the fermi surface for the field of the form @xmath122 , where @xmath123 but by the technical reasons do not give the corresponding figure here . in this case , the fermi surface is divided into great numbers of connected pieces having the shape of winding sheets . such a complex structure of the fermi surface reflects the fractal structure of the spectrum of the periodic landau operator @xcite . we are thankful to s. albeverio , m. v. budantsev , v. ya . demikhovskii , and p. exner for valuable discussions and useful remarks . this work was supported by grants of intas 00 - 257 , dfg 436 rus 113/572/0 - 2 , and rfbr 02 - 01 - 00804 . 0 a. p. cracknell and k. c. wong , _ the femi surface _ ( clarendon press , oxford , 1973 ) . s. p. novikov , a. ya . _ uspekhi fiz . nauk _ * 168 * , 249 ( 1998 ) . j. caulfield , j. singleton , p. t. j. hendriks , j. a. a. j. perenboom , f. l. pratt , m. doporto , w. hayes , m. kurmoo , and p. day , _ j. phys . : condens . matter _ , * 6 * , l155 ( 1994 ) . a. a. house , n. harrison , s. j. blundell , i. deckers , j. singleton , f. herlach , w. hayes , j. a. a. j. perenboom , m. kurmoo , and p. day , _ phys . b _ , * 53 * , 9127 ( 1996 ) . n. harrison , d. w. hall , r. g. goodrich , j. j. vuillemin and z. fisk , _ phys . rev . lett_. * 81 * , 870 ( 1998 ) . m. za . azbel , _ zhurn . _ , * 46 * , 929 ( 1964 ) [ in russian ] . d. r. hofstadter , _ phys . b _ * 14 * , 2239 ( 1976 ) . m. koshino , h. aoki , k. kuroki , s. kagoshima , and t. osada , _ phys . lett . _ * 86 * , 1062 ( 2001 ) ; m. koshino , h. aoki , t. osada , k. kuroki , and s. kagoshima , _ phys . rev . b _ * 65 * , 045310 ( 2002 ) ; m. koshino and h. aoki , _ phys . rev . b _ * 67 * , 195336 ( 2003 ) . j. brning , v. v. demidov , and v. a. geyler , _ spectral diagrams of hofstadter type for the bloch electrons in three dimensions . _ ` cond - mat/0306391 ` . j. zak , _ phys . rev . a _ * 134 * , 1602 ( 1964 ) . v. ya . demikhovskii , a. a. perov , and d. v. khomitsky , _ phys . a _ , * 267 * , 408 ( 2000 ) . v. ya . demikhovskii and d. v. khomitsky , _ zhurn . _ , * 120 * , 191 ( 2001 ) [ in russian ] . g. petschel and t.geisel , _ phys . rev _ , * 71 * , 239 ( 1993 ) ; o. khn , v. fessatidis , h. l. cui , p. e. selbmann , and n. j. m. horing , _ phys . b _ * 47 * , 13019 ( 1993 ) . s. a. gredeskul , m. zusman , y. avishai , and m. ya . azbel , _ phys . _ , * 288 * ( 1997 ) , 223 ; v. a. geyler , i. yu . popov , a. a. ovechkina , and a. v. popov , _ chaos , solitons , and fractals _ * 11 * , 281 ( 2000 ) . a. i. baz , ya . b. zeldovich , and a. m. perelomov , _ scattering , reactions and decay in nonrelativistic quantum mechanics _ ( moscow , nauka , 1971 ) . yu . n. demkov and v. n. ostrovskiy , _ zero - range potentials and their applications in atomic physics _ ( plenum press , new york , 1988 ) . s. albeverio , f. gesztesy , r. hegh - kron and h. holden , _ solvable models in quantum mechanics _ ( springer - verlag , berlin , 1988 ) . r. hegh - kron , h. holden , s. johannesen and t. wentzel - larsen , _ j. math . * 27 * , 385 ( 1986 ) . y. avishai m. ya . azbel , s. a. gredeskul , _ phys . b _ , * 48 * , 17280 ( 1993 ) . v. a. geyler and v. v. demidov , _ theor . math . phys . _ * 103 * , 561 ( 1995 ) . v. a. geyler and v. a. margulis , _ theor . phys . _ * 70 * , 133 ( 1987 ) . h. bateman and a. erdlyi , _ higher transcendental functions , v. i _ ( mcgraw - hill , new york , 1953 ) . o. l. lazarenkova and a. a. balandin , _ phys . b _ * 66 * , 254319 ( 2002 ) . b. a. boyko , a. i. boyko , m. i. dolotenko et al . , book of abstracts , _ conference on megagauss magnetic field and related topics , _ ( tallahassee , usa ) , p. 149 .	a method of building and investigation of the fermi surfaces for three - dimensional crystals subjected to a uniform magnetic field is presented . the hamiltonian of a charged particle in the crystal is treated in the framework of the zero - range potential theory . the dispersion relation for the hamiltonian is obtained in an explicit form . + _ keywords : _ fermi surface ; crystal in a high uniform magnetic field . * fermi surfaces of crystals in a high magnetic field * + j. brning _ mathematisch - naturwisenschaftliche fakultt ii der humboldt - universitt zu berlin , + rudower chaussee 25 , berlin 12489 , germany _ + v. v. demidov _ laboratory of mathematical physics , mordovian state university , + bolshevistskaya 68 , saransk , mordoviya 430000 , russia _ + v. a. geyler _ laboratory of mathematical physics , mordovian state university + [email protected]_ +
the topic of invisibility which is a very old one in optics and electromagnetics ( wolf & habashy 1993 ) has lately acquired a new lease of life with the advent of metamaterials ( fedotov , mladyonov , prosvirnin & zheludev 2005 ; al & engheta 2006 ) . in a similar vein , it has recently been proposed that the exotic electromagnetic possibilities offered by metamaterials may be harnessed to achieve cloaking , at least in principle ( leonhardt 2006 ; milton & nicorovici 2006 ; pendry , schurig & smith 2006 ) . the theoretical arguments underlying this proposed cloaking are based on the facilitation of coordinate transformations by nonhomogeneous metamaterials . a quite different approach to concealing a material is pursued in this paper . it is based on the perceived deflection of light by a material slab translating at constant velocity . no special material properties are required , but for illustrative purposes , we consider an isotropic , homogeneous , dielectric material . in two previous studies , we have demonstrated that the much heralded negative phase velocity phenomenon often associated with negatively refracting electromagnetic metamaterials can be realized by conventional materials through the process of uniform motion ( mackay & lakhtakia 2004 , 2006a ) . here , we demonstrate that a substantial degree of concealment may also be realized by uniform motion . as regards notational matters , 3 vectors are in boldface , with the @xmath2 symbol identifying unit vectors . double underlining signifies a 3@xmath33 dyadic and @xmath4 is the identity 3@xmath33 dyadic . the superscript @xmath5 denotes the transpose of a column vector . the permittivity and permeability of vacuum are @xmath6 and @xmath7 . the vacuum wavenumber is @xmath8 with @xmath9 being the angular frequency , and the vacuum wavelength is @xmath10 . pw_section as a preliminary to concealment of a moving slab ( [ beam_section ] ) , let us consider a uniformly moving half space . suppose that a plane wave is launched with wavevector @xmath11 from vacuum ( @xmath12 ) towards the half space @xmath13 occupied by an isotropic , nondissipative , dielectric material . this material moves at constant velocity @xmath14 , parallel to the interface and in the plane of incidence . in an inertial frame of reference that moves with the same velocity @xmath15 with respect to the laboratory frame of reference wherein @xmath16 is specified , the refracting material is characterized by relative permittivity @xmath17 . the minkowski constitutive relations of the moving half space in the laboratory frame of reference are ( chen 1983 ) @xmath18 where @xmath19 & \#m = m \hat{\#v}\ , , \quad \displaystyle { m= \frac{\beta \le \eps_r - 1 \ri}{1 - \eps_r \beta^2}}\ , , & \beta = v \sqrt{\epso \muo}\ , \end{array } \right\}\,.\ ] ] in order to exclude the possibility of evanescent plane waves , @xmath20 is assumed . the envisaged scenario is illustrated schematically in figure [ fig1 ] . the angle @xmath21 between the refracted wavevector @xmath22 , as observed from the laboratory frame of reference , and the unit vector @xmath23 normal to the interface is related to the angle of incidence @xmath24 by ( chen 1983 ) @xmath25 where @xmath26 is the wavenumber of the refracted wave and @xmath27 . since @xmath28 , refraction is positive @xmath29 ( mackay & lakhtakia 2006b ) . the time averaged poynting vector of the refracted plane wave is given by ( chen 1983 ) @xmath30 where @xmath31 and @xmath32 are constants , and the angle between @xmath23 and @xmath33 is @xmath34 as an illustrative example , the angle @xmath35 is plotted in figure [ fig2 ] against @xmath36 for @xmath37 , for the half space characterized by @xmath38 . the orientation of the refracted time averaged poynting vector rotates towards the direction of motion as @xmath39 increases from @xmath40 . the counterposition regime which occurs where @xmath41 for @xmath42 is discussed elsewhere ( lakhtakia & mccall 2004 ; mackay & lakhtakia 2006b ) . in connection with figure [ fig2 ] , it is of particular interest here that @xmath43 at ( i ) @xmath44 for @xmath45 , ( ii ) @xmath46 for @xmath47 , and ( iii ) @xmath48 for @xmath49 . that is , there exist angles of incidence at which the time averaged poynting vector is not deflected by the uniformly moving half space . this suggests that it may be possible for a light beam not to be confused with a plane wave to pass through a uniformly moving slab at an oblique angle without experiencing a lateral shift in position . that suggestion inspired the research presented in the next section . beam_section suppose that the uniformly moving half space considered in [ pw_section ] is now replaced by a slab of thickness @xmath50 moving at constant velocity @xmath51 parallel to its two surfaces , as schematically illustrated in figure [ fig3 ] . the slab which is characterized , as before , by @xmath20 in a co moving reference frame is immersed in vacuum . a 2d beam with electric field phasor ( haus 1984 ) @xmath52 is incident upon the slab at a mean angle @xmath53 relative to the slab normal direction @xmath54 . the beam is represented as an angular spectrum of plane waves , with @xmath55 being the wavevector of each planewave contributor . the angular spectral function @xmath56 is taken to have the gaussian form ( haus 1984 ) @xmath57 with @xmath58 being the width of the beam waist . two polarization states are considered : parallel to the plane of incidence , i.e. , @xmath59 and perpendicular to the plane of incidence , i.e. , @xmath60 the electric field phasor of the reflected beam is given as @xmath61 with @xmath62 and @xmath63 & \mbox{for } & \#e_i ( \vartheta ) = \#e_\parallel \vspace{4pt } \\ r_\perp \,\#e_\perp & \mbox{for } & \#e_i ( \vartheta ) = \#e_\perp \end{array } \right . .\ ] ] the electric field phasor of the transmitted beam is given as @xmath64 with @xmath65 and @xmath66 expressions for the reflection coefficients @xmath67 and transmission coefficients @xmath68 are provided in equations in the appendix . in view of figure [ fig2 ] , we fixed the mean angle of incidence of the beam at @xmath69 and explored the behaviour of the transmitted beam for @xmath70 , @xmath46 and @xmath71 . the energy density of the beam in both half spaces , as measured by @xmath72 is mapped for @xmath73 and @xmath74 in figure [ fig4 ] with the slab thickness @xmath75 . the restriction @xmath76 $ ] was imposed to exclude evanescence . a beam waist of @xmath77 was selected for all calculations . we considered @xmath78 for both @xmath79 and @xmath80 . regardless of the polarization state , the transmitted beam does not undergo a lateral shift ( relative to the incident beam ) when @xmath46 . however , the transmitted beam is laterally shifted in the direction of @xmath81 when @xmath70 and in the direction of @xmath82 when @xmath83 . the energy densities of the reflected and transmitted beams are sensitive to @xmath39 and the polarization state of the incident beam . a more quantitative representation of the transmitted beam is provided in figure [ fig5 ] , wherein @xmath84 is plotted against @xmath85 for @xmath86 at @xmath87 . for comparison , @xmath84 for the beam in the absence of the moving slab is also plotted . it is clear that the beam position for @xmath46 coincides with the beam position in the absence of the moving slab . at @xmath46 , the peak energy density of the transmitted beam for the case of parallel polarization is approximately 11@xmath88 less than it would be if the slab were absent ; the corresponding figure for perpendicular polarization is @xmath89 . the median shift of transmitted beam in relation to the incident beam is defined as @xmath90 for both parallel and perpendicular polarizations , @xmath91 is plotted against @xmath36 in figure [ fig6 ] . thus , regardless of the polarization state , the beam can be shifted laterally along @xmath92 by means of uniform motion . in particular , the zero beam shift at @xmath46 is further confirmed in figure [ fig6 ] . our numerical investigations show that a 2d beam can pass obliquely through a uniformly moving slab without undergoing a lateral shift in its position . at a fixed angle of beam incidence , this effect occurs only for a unique translational slab velocity . however , extrapolating from figure [ fig2 ] , for every angle of beam incidence a slab velocity can be found at which the beam undergoes no lateral shift . furthermore , for a fixed angle of incidence , a pulsed beam will undergo zero lateral deflection , provided that the constitutive parameters do not vary with angular frequency in the pulse spectrum . the degree of concealment achieved by uniform motion is not 100@xmath88 due to reflections but , in the particular case of the example considered in [ beam_section ] , almost 90@xmath88 of the peak energy density of the beam can be transmitted without deflection . * acknowledgement : * tgm is supported by a _ royal society of edinburgh / scottish executive support research fellowship_. the reflection coefficients @xmath67 and transmission coefficients @xmath68 are straightforwardly calculated by solving the reflection transmission problem as a boundary value problem . we outline the procedure here , further details being available elsewhere ( lakhtakia & messier 2005 ) . consider the plane wave with electric and magnetic field phasors @xmath93 propagating in the @xmath94 plane . as in [ beam_section ] , a moving slab described by the minkowski constitutive relations occupies the region between @xmath95 and @xmath96 ; elsewhere there is vacuum . we write @xmath97 substitution of equations and into the source free maxwell curl postulates @xmath98 delivers four differential equations and two algebraic equations . the latter two equations are easily solved for @xmath99 and @xmath100 . thereby , the four differential equations may be expressed in matrix form as @xmath101 where @xmath102 is a column vector and @xmath103 is a 4@xmath34 matrix with @xmath104 the solution to is conveniently expressed as @xmath105 in terms of the transfer matrix @xmath106 now we turn to the incident , reflected and transmitted plane waves . let the incident plane wave be represented in terms of linear polarization components as @xmath107 the corresponding reflected and transmitted plane waves are given as @xmath108 and @xmath109 respectively . by application of the boundary conditions at @xmath110 and @xmath96 to the solution , the reflection and transmission coefficients are found to be related by the matrix algebraic equation @xmath111 wherein @xmath112 thus , after some manipulation , the reflection and transmission coefficients emerge as @xmath113	the perceived lateral position of a transmitted beam , upon propagating through a slab made of homogeneous , isotropic , dielectric material at an oblique angle , can be controlled through varying the velocity of the slab . in particular , by judiciously selecting the slab velocity , the transmitted beam can emerge from the slab with no lateral shift in position . thereby , a degree of concealment can be achieved . this concealment is explored in numerical calculations based on a 2d gaussian beam . 0.2 cm * tom g. mackay@xmath0 and akhlesh lakhtakia@xmath1 * 0.2 cm @xmath0 school of mathematics + james clerk maxwell building + university of edinburgh + edinburgh eh9 3jz , uk + email : [email protected] 0.4 cm @xmath1 catmas computational & theoretical materials sciences group + department of engineering science & mechanics + 212 earth & engineering sciences building + pennsylvania state university + university park , pa , 168026812 , usa + email : [email protected] 0.4 cm 0.2 cm * keywords : * _ minkowski constitutive relations , moving slab , gaussian beam , counterposition _ 0.4 cm
confirmation of cp violation in @xmath1 meson decays through the measurement of the time - dependent decay rate asymmetry has demonstrated the power of the high luminosity @xmath1 factories , belle at kekb , kek and babar at pep - ii , slac . together with measurements of the angles and sides of the unitarity triangle , the picture of cp violation through the kobayashi - maskawa mechanism has been established to be at least the dominant source of all cp violating phenomena in the high energy physics known to date . this was possible only by building the `` @xmath1 factories '' with two orders of magnitude higher luminosity than existing facilities at that time , and with a new concept of the boosted center - of - mass frame with asymmetric beam energies for an @xmath4 collider . the @xmath1 factories have also provided a rich field for many other non - trivial tests of the standard model ( sm ) with their power based on a huge statistics of @xmath1 decays , @xmath2 decays and @xmath3 decays . however , in spite of its success , the sm is still considered to be a low energy approximation of a more fundamental physics beyond the sm ( bsm ) for many reasons . for example , cp violation in the sm can not explain the baryon number asymmetry in the universe ; another more theoretical argument is that bsm physics is expected to lie at the tev energy scale in order to solve the hierarchy problem . on the other hand , there is a mystery called the flavor problem : a new interaction around a tev energy scale would alter the flavor changing neutral current amplitudes and a large deviation due to bsm would be expected , unless the coupling constants are fine tuned or aligned to the sm couplings to reconcile with the strong constraints from @xmath1 factory measurements . bsm effects should appear in most cases through loop diagrams , while the tree diagrams are usually unaffected . to observe the effects on the unitarity triangle of different contributions of loop and tree diagrams , and also through other rare phenomena , it is again necessary to build a new facility with a further two orders of magnitude higher luminosity than now . this is the motivation for a `` super '' @xmath1 factory . # 110 ^ -#1 # 110^#1 the key measurement at a super @xmath1 factory will still be the measurement of the unitarity triangle , but with a precision an order of magnitude better than what we have now . there are also a large number of other potential measurements that are sensitive to bsm , and their importance will be more significant than has been the case in belle and babar . more thorough and detailed studies can be found in @xcite . the current world average on the angle @xmath5 of the unitarity triangle , @xmath6 , @xmath7 is used instead of @xmath8 , @xmath9 and @xmath10 . ] has a remarkable precision , @xmath11 @xcite , where the error is dominated by the results from the two @xmath1 factories using the sum of the datasets exceeding @xmath12 . measurement of this and other angles , and the sides of the triangle were made possible only after the @xmath1 factories came online . all the results so far are found to be consistent with each other . however , they are not precise enough yet to identify any possible discrepancy , for example in the @xmath0 violating phase measured in the loop diagrams in which bsm effects could reside , from the phase measured from tree diagrams in which bsm effects are not expected . the angle @xmath5 is measured in the form of @xmath13 , which is the coefficient of the sine term ( @xmath14 ) in the time - dependent @xmath0 asymmetry that appears due to an irreducible complex phase in the @xmath15 mixing when measured together with the @xmath16 transition into a @xmath0 eigenstate such as @xmath17 . at this moment the measurement of @xmath13 is still slightly statistical error dominated . an early dataset of @xmath18 from a super @xmath1 factory will turn this into an ultimate measurement of @xmath13 with an error of @xmath19 ( or @xmath20 in terms of @xmath5 ) . the precision of the measurement then matches the theoretical uncertainty , which appears due to the effects of subdominant diagrams with the same final state , a complex phase different from @xmath5 and an unknown size of the amplitude . the angle @xmath6 appears in the combination of the @xmath15 mixing and the @xmath21 transition into a @xmath0 eigenstate such as @xmath22 . unfortunately , @xmath23 can not be directly measured from the time - dependent asymmetry of @xmath22 due to the @xmath24 penguin contribution which has a different weak phase . the amplitude of such a contribution and the angle @xmath6 have to be extracted from measurements of all @xmath25 charge combinations ( @xmath26 , @xmath27 , @xmath28 ) by assuming isospin symmetry . the same is possible with @xmath29 , or a more elaborate technique can be performed to disentangle the amplitudes and phases of @xmath30 using a time - dependent dalitz analysis of @xmath31 . multiple solutions appear in these procedures and they make it difficult to pinpoint the value of @xmath6 ; the current @xmath32 interval from combined belle and babar results is @xmath33^\circ$ ] , e.g. , in @xcite . a super @xmath1 factory will significantly improve the situation as the correct solution will become unambiguous from the three independent measurements . the combined error for @xmath6 is about @xmath34 with @xmath18 data . this is the ultimate precision at which the experimental error matches the model uncertainty of the dalitz amplitudes or theoretical uncertainty due to isospin breaking effects . at a super @xmath1 factory , the set of angles @xmath5 and @xmath6 , already with @xmath18 , provides a reference `` point '' of the unitarity triangle with a 2% error . all of the other possible measurements will be used to search for a deviation of @xmath35 from the sm . the three most promising measurements to search for a deviation from expectation for the unitarity triangle variables are ; the angle @xmath5 from loop mediated @xmath36 decays , the angle @xmath7 from the tree process @xmath37 , and the length of the side @xmath38 using @xmath39 decays . any deviation from the reference point defined by @xmath40 becomes evidence of bsm . in the sm , the size of the time dependent @xmath0 asymmetry @xmath14 in the @xmath41 decay modes into a @xmath0 eigenstate is the same as that in @xmath16 , i.e. @xmath13 , as both diagrams have the same weak phase . since @xmath41 is a penguin loop , the phase can be modified by an additional bsm amplitude if it exists , while it is unchanged in @xmath16 . therefore , the deviation in @xmath14 from @xmath13 is the sign of bsm . for a given final state , there are always subdominant diagrams , e.g. , the tree diagram @xmath21 with @xmath42 , which are the source of possible modification of @xmath14 within the sm . three modes , @xmath43 , @xmath44 and @xmath45 , are considered to be the golden modes with the smallest of such pollutions , estimated to be about 0.02 . measurements will be statistical error dominated until @xmath46 or more integrated luminosity is obtained , as shown in fig . [ fig : btoscpv ] . the expected errors are , 0.02 , 0.03 and 0.04 for @xmath44 , @xmath43 and @xmath45 , respectively , at @xmath46 . , @xmath43 and @xmath45 decays as a function of the integrated luminosity at superkekb . , width=302 ] the angle @xmath7 is most cleanly measured using @xmath37 decays . there , @xmath7 appears as interference between the tree amplitude @xmath47 with @xmath48 and another tree amplitude @xmath21 with @xmath49 . the former gives @xmath50 in the final state while the latter gives @xmath51 ; common final states between @xmath51 and @xmath50 decays are the source of interference between these two decay channels . as there is no subdominant amplitude with a loop diagram , this provides the cleanest sm measurement . there are several methods depending on different @xmath51 decay modes , e.g. , @xmath51 decaying into a @xmath0 eigenstate , or into a doubly cabibbo suppressed final state . the most effective method is based on the dalitz analysis of the decay chain @xmath52 ; @xmath53 . the analysis depends on the model of the three - body @xmath2 decay amplitudes , and gains significantly from charm factory data with cp - tagged @xmath51 decays . the combined @xmath7 error will be @xmath54 at @xmath18 , or @xmath34 at @xmath46 . the @xmath38 measurement is performed using either an inclusive measurement of @xmath55 or an exclusive one such as @xmath56 . currently , both measurements have a similar size in the error , and have some tension between the results . both methods provide the cleanest results using a tagging technique of reconstructing a hadronic @xmath1 decay mode for the other @xmath1 meson decay . the tagging efficiency is not very high and demands huge statistics which is suitable at a super @xmath1 factory . since only a limited kinematical range can be measured due to the huge @xmath57 background , the inclusive measurement requires the operator product expansion technique to extrapolate into the full kinematical range . all the necessary information is available from data , and the total error will be 6% at @xmath18 and 4% at @xmath46 . the exclusive branching fraction will be more accurately measured at a super @xmath1 factory . if the @xmath58 form factor is calculated using lattice qcd more precisely , @xmath38 will be determined from the exclusive measurement with a smaller error . data at superkekb . , width=302 ] expected sensitivities of the unitarity triangle measurements at @xmath46 data at superkekb is shown in fig . [ fig : rhoeta50 ] using the current world average values for the central values . given these measurements , a 10% deviation in any of these measurement would be identified at a super @xmath1 factory with @xmath46 . there are many other key measurements which can be used to search for bsm in @xmath1 decays , and also in decays of @xmath2 mesons and @xmath3 leptons , which are equally abundantly produced at a super @xmath1 factory . the weak interaction which governs @xmath59 quark decays is based on a left - handed current in the limit of massless quarks . this is not necessarily the case in many bsm models . the right - handed current in the @xmath36 transition can be effectively identified as a non - zero @xmath14 value in a time - dependent cp measurement of @xmath60 . with @xmath46 the error will be less than @xmath61 and already better than the theoretical uncertainty on the deviation of @xmath14 from zero due to the finite mass of the @xmath62 quark . many bsm models also require more than one higgs doublet , including a charged higgs boson . the charged higgs boson can replace the weak boson in a tree diagram , and its effect is enhanced in the helicity suppressed purely leptonic decays and semi - leptonic decays with a @xmath3 lepton . the effect can be searched for through a deviation from expectation in the branching fraction of @xmath63 . similar measurements can be performed with @xmath64 and @xmath65 . if deviations are observed in all these modes , a comparison between them leads to a test of the universality of the coupling , and provides stronger evidence for the existence of the charged higgs boson . inclusive measurements such as @xmath66 , @xmath67 and @xmath68 are also sensitive to a wide range of bsm . especially , the zero - crossing point of the forward - backward asymmetry in @xmath68 has a very clean signature . the recently observed large values of the @xmath69 mixing parameters @xmath70 , of the order @xmath71 , suggest the possibility of a bsm contribution , while an explanation within the sm is not excluded because of a large hadronic uncertainty . a measurement of @xmath0 violation in @xmath69 mixing would be clear evidence for a bsm effect in the charm quark sector . finally , lepton flavor violating @xmath3 decay is also allowed in many bsm models , while it is not allowed at all in the sm . there are a large number of possible lepton flavor violating decay modes ( e.g. , @xmath72 , @xmath73 or @xmath74 ) which have been and will be searched for . if observed , it will be an unambiguous sign of new physics . there may be a question why we have to build a super @xmath1 factory while the next generation flavor physics can be studied at lhcb . in reality , it is almost impossible to measure modes with photons , @xmath75 and neutrinos , and perform inclusive measurements at lhcb . many of these are the key measurements to study bsm as already discussed . there are examples where lhcb has an excellent sensitivity : the unitarity triangle parameters , especially the angle @xmath7 , can be precisely measured at lhcb with a similar precision to that of at a super @xmath1 factory , provided that the systematic errors are under control . in order to search for a bsm @xmath0 phase in the @xmath76 transition , @xmath77 can be used ; in order to search for the right - handed current , @xmath78 can be used . these are different decay modes related to searches for the same type of bsm effects , and the searches at the two places are extremely helpful for gaining an unambiguous understanding of bsm physics . in order to collect an integrated luminosity of @xmath46 within a reasonable amount of running time , the instantaneous luminosity has to be above or at least close to @xmath79 . in addition , to keep synergy with energy frontier physics at the lhc and flavor physics at lhcb , it is crucial to operate the super @xmath1 factory in the next decade . currently , two projects are planned : the superkekb project in japan and the superb project in italy . if resources allow , it is definitely better to have both facilities for healthy competition and cross - checks , as was extremely helpful in the case of competition between belle and babar . however , under the current situation for high energy physics , it does not seem to be possible to have both of them at the same time . key parameters for a high luminosity are the beam current ( @xmath80 ) and the beam - beam parameter ( @xmath81 ) that are proportional to the instantaneous luminosity , and the vertical @xmath8 function at the interaction point ( @xmath82 ) which is inversely proportional to the luminosity . the two projects take different approaches for a higher luminosity . typical parameters for superkekb are beam currents of 9.4 a @xmath83 4.1 a , @xmath84 and @xmath85 mm for both rings . those for superb are 1.85 a @xmath83 1.85 a beam currents , @xmath86 and @xmath87 mm for the high / low energy ring . an example of the parameters is given in table [ tbl : superkekb - superb ] . these parameters give instantaneous luminosities of @xmath88 for superkekb and @xmath89 for superb , almost two orders of magnitude higher than the current kekb record , @xmath90 . # 1 .an example of machine parameters for superkekb and superb . [ cols="<,^,^,^,^,^,^ " , ] the superkekb project is an upgrade of the current kekb facility at the same place , reusing a large fraction of the existing components and infrastructure . one of the key components of superkekb is the `` crab '' cavity that rotates the envelope of the beam bunch and makes head - on collisions possible for beams incident at a finite angle ( crab - crossing ) . this will at least geometrically increase the effective volume of the collision . according to a simulation , the effect is more dramatic : the beam - beam force becomes nearly independent of the horizontal coordinate at a half integer tune . in the case of kekb with a 22 mrad crossing angle , @xmath81 becomes 0.15 and almost doubles the luminosity , and for superkekb with a 30 mrad crossing angle , @xmath84 is possible . with this simulation result , kekb has installed a crab cavity for each of high and low energy rings and has been commissioning since 2007 . under a low beam current operation , the specific luminosity reached the predicted value ( fig . [ fig : speclumi - crab ] ) , and an enhancement in the beam - beam parameter , @xmath91 , with respect to the case before the crab cavity , @xmath92 , was observed . however , there is a drop in the luminosity under a high beam current , and the reason is still being investigated . in addition , the current crab cavity has to be operated at a lower beam current than what was already achieved without , and this has prevented the expected boost in the instantaneous luminosity so far . the commissioning will continue for one more year until the end of the kekb running time . the other key issue is the higher beam current . in order to store a higher beam current without being affected by the electron cloud , the vacuum pipe will be replaced all over the ring with antechamber type beampipes , and the bellows with higher - current - proof ones . the kekb upgrade plan is already included in kek s five - year roadmap from 2009 to 2013 . this includes a three - year shutdown time for the construction . the superb project in italy is a completely new project , with an ultra low emittance approach inspired by the international linear collider ( ilc ) project . in order to achieve a very high luminosity with a moderate beam current , both beams have to be squeezed into an extremely small interaction region with a vertical size of 39 nm , and with a very small @xmath82 . such a beam could be possible , according to the studies for the ilc dumping ring . however , it does not necessary mean that such a small emittance can be sustained while making collisions at every turn with a very high luminosity . the other problem would be the very short beam lifetime of about only five minutes . the ultra - small beamsize also implies a huge hourglass effect , and it has to be suppressed either by making the bunch length extremely short , which is not realistic , or by moving the vertical waist positions of both beams along the @xmath93 axis with a proper phase . the latter is called the `` crab waist '' , and can be realized with a sextupole magnet . the concept of the crab waist has been successfully tested at da@xmath94ne at a low energy and low beam current , as shown in fig . [ fig : dafne-2008may ] . ne.,width=302 ] the other consideration at superb is to make the energy asymmetry smaller , in order to save on the electricity bill . this makes the resolution of the time difference between two @xmath1 meson vertices worse , but it gives a better hermeticity , which is crucial for measurements with neutrinos . the candidate site is at the tor vegarta campus of infn rome . here also , many components , including magnets , will be brought from pep - ii to minimize the cost . a high luminosity brings a high event rate . the @xmath1-pair events alone will be delivered at a rate of 1 khz under a luminosity of @xmath95 , and the physics trigger rate will become 10 khz excluding the bhabha events , which have an even higher rate . in order to keep nearly 100% efficiency for the @xmath1-pair events , the trigger and data acquisition system have to be drastically improved . the readout electronics also has to be as deadtime - free as possible . a high luminosity brings a high background rate at the same time . on the other hand , to keep the advantage of the clean @xmath4 environment especially for measurements with photons , @xmath75 and neutrinos , any performance drop with respect to the current belle or babar detector is not acceptable . since the existing belle and babar detectors already have excellent performance , it is not easy to drastically improve it , especially after coping with the high background . the detector at superkekb ( superbelle ) has to cope with the large beam - gas background due to the much higher beam currents . after a careful design of the beampipe and masks at the interaction region , it is found that the size of touschek background is moderate , and the radiative - bhabha background is not harmful except for the outmost @xmath96 and muon detector . the total background will be about 20 times that of the current conditions at belle , and therefore a significant amount of modification to the belle detector is necessary . in order to cope with the background , the following changes are planned . the silicon vertex detector will have a larger radius with six layers , with a possible option of a pixel detector for the innermost layer . the enlarged vertex detector will replace the inner part of the drift chamber and allow a larger volume for @xmath97 vertexing . the drift chamber will be replaced with one with smaller cell size to shorten the drift time and to reduce the occupancy . the outer radius of the drift chamber will be enlarged , thanks to the thinner outer detectors . the particle identification devices will be fully replaced from existing time - of - flight counters and the threshold - type aerogel cherenkov counter , with a detector to reconstruct the cherenkov ring image , such as a time - of - propagation counter for the barrel part and an aerogel ring - image counter for the forward endcap part . the endcap part of the calorimeter will be replaced with pure csi crystals that have a faster time response , while the thallium - doped csi crystals will be unchanged for the barrel part . the resistive plate counters for @xmath98 and muon detection will be replaced with scintillation fibres . thanks to the smaller beam current , the beam - gas background is expected to be moderate at the superb detector . on the other hand , a huge touschek background is expected , and also the very short lifetime could be harmful for the detector . the detector is based on reuse of the existing babar detector , which is already more immune to backgrounds than belle . however , similarly to belle s case , many of the components have to be replaced . these includes a new silicon vertex tracker , a new drift chamber , a new forward calorimeter and a new @xmath98 and muon detection system . the physics program at a super @xmath1 factory is very compelling in the integrated luminosity range between 5 to 50 @xmath99 . this includes the measurements of the precise reference point in the unitarity triangle and possible deviations from there , extensive searches for right handed currents , charged higgs , lepton - flavor violating decays , and many other search channels . both projects described in this paper are actively working on the accelerator and detector designs . the design for superkekb is already finalizing for the production of necessary components , while the superb design is also getting converged . we are looking forward to the exciting future of flavor physics that will be possible at a super @xmath1 factory .	after the establishment of the kobayashi - maskawa mechanism of @xmath0 violation at the two @xmath1 factories , possibilities to increase the integrated luminosity by two orders of magnitude have been investigated , since it seems to be the amount needed to find physics beyond the standard model through @xmath0 violating and other observables in rare @xmath1 meson decays , @xmath2 meson decays and @xmath3 lepton decays . this report reviews the physics sensitivities and status of such super @xmath1 factories , which are planned at two locations .
we consider a population of @xmath0 agents , representing for instance people , firms or nations , engaged in bilateral collaborative interactions . each interaction is described by a continuous snowdrift game @xcite , one of the fundamental models of game theory . in this game , an agent @xmath1 can invest an amount of time / money / effort @xmath2 into the collaboration with another agent @xmath3 . cooperative investments accrue equal benefits @xmath4 to both partners , but create a cost @xmath5 for the investing agent . assuming that investments from both agents contribute additively to the creation of the benefit , the payoff received by agent @xmath1 from an interaction with an agent @xmath3 can then be written as @xmath6 the game thus describes the generic situation in which agents invest their personal resources to create a common good shared with the partner . as an example of the snowdrift game , the reader may think of a scientific collaboration where two researchers invest their personal time in a project , while the benefit of the publication is shared between them . this example makes it is clear that the benefit of the collaboration must saturate when an extensive amount of effort is invested , whereas the cost to the an agents , measured for instance in terms of personal well - being , clearly grows superlinearly once the personal investment exceeds some hours per day . in the following we do not restrict the cost- and the benefit - functions , @xmath4 and @xmath5 , to specific functional forms , except in the numerical investigations . however , we assume that both are differentiable and , moreover , that @xmath4 is sigmoidal and @xmath5 is superlinear ( cf . fig . [ pocfig2 ] ) . these assumptions capture basic features of real - world systems such as inefficiency of small investments , saturation of benefits at high investments , as well as additional costs incurred by overexertion of personal resources and are widely used in the sociological and economic literature @xcite . to account for multiple collaborations per agent , we assume that the benefits received from collaborations add linearly , whereas the costs are a function of the sum of investments made by an agent , such that the total payoff received by an agent @xmath1 is given by @xmath7 where @xmath8 denotes the _ total investment _ of the agent @xmath1 while @xmath9 denotes the total investment made in the collaboration @xmath10 . this is motivated by considering that benefits from different collaborations , say different publications , are often obtained independently of each other , whereas the costs generated by different collaborations stress the same pool of personal resources of an agent . let us emphasize that we do not restrict the investment of an agent further . while investments can not be negative , no upper limit on the investments is imposed . furthermore , the agents are free to make different investments in collaborations with different partners . thus , to optimize its payoff , an agent can reallocate investments among its potential partners as well as change the total amount of resources invested . for specifying the dynamics of the network , we assume the agents to be selfish , trying to increase their total payoff @xmath11 by a downhill - gradient optimization @xmath12 every agent can cooperate with every other agent . thus , the network of potential collaborations is fully connected and the deterministic time - evolution of the model system is given by a system of @xmath13 ordinary differential equations of the form of eq . [ timeevolution ] . the network dynamics , considered in the following , is therefore only the shifting of link weights @xmath14 . note however that already the weight dynamics constitutes a topological change . as will be shown in the following , the agents typically reduce their investment in the majority of potential collaborations to zero , so that a sparse and sometimes disconnected network of non - vanishing collaborations is formed . therefore the terminology of graph theory is useful for characterizing the state that the system approaches . below , we use the term _ link _ to denote only those collaborations that receive a non - vanishing investment @xmath15 . a link is said to be _ bidirectional _ if non - vanishing investments are contributed by both connected agents , while it is said to be _ unidirectional _ if one agent makes a non - vanishing investment without reciprocation by the partner . likewise , we use the term _ neighbours _ to denote those agents that are connected to a focal agent by non - vanishing collaborations and the term _ degree _ to denote the number of non - vanishing collaborations in which a focal agent participates . in the following , the properties of the model are investigated mostly by analytical computations that do not require further specifications . only for the purpose of verification and illustration we resort to numerical integration of the ode system . for these we use the functions @xmath16 for studying the time - evolution of exemplary model realizations by numerical integration , all variables @xmath14 are assigned random initial values drawn independently from a gaussian distribution with expectation value @xmath17 and standard deviation @xmath18 constituting a homogeneous state plus small fluctuations . the system of differential equations is then integrated using euler s method with variable step size @xmath19 . in every timestep , @xmath19 is chosen such that no variable is reduced by more than half of its value in the step . if in a given timestep a variable @xmath14 falls below a threshold @xmath20 and the corresponding time derivative is negative , then @xmath21 is set to zero for one step to avoid very small time steps . we emphasize that introducing the threshold @xmath22 is done purely to speed up numerical integration and does not affect the results or their interpretation . in particular , we confirmed numerically that , the exact value of @xmath22 does not influence the final configuration that is approached . in all numerical results shown below @xmath23 the numerical exploration of the system reveals frustrated , glass - like behavior ; starting from a homogeneous configuration as described above , it approaches either one of a large number of different final configurations , which are local maxima of the total payoff . a representative example of an evolved network , and snapshots from the time - evolution of two smaller example networks are shown in figs . [ pocfig1],[pocfig1a ] , respectively . in the example networks only those links are shown that receive a non - vanishing ( i.e. above - threshold ) investment . most of these non - vanishing links are _ bidirectional _ , receiving investments from both of the agents they connect . only rarely , _ links appear , which are maintained by one agent without reciprocation by the partner . for further investigations it useful to define a _ bidirectionally connected component _ ( bcc ) as a set of agents and the bidirectional links connecting them , such that , starting from one agent in the set , every other agent in the set can be reached by following a sequence of bidirectional links . in the numerical investigations we observe that all bidirectional links within a bcc receive the same total investment in the final state . however , the investment @xmath15 made in every given link is in general not split equally among the two connected agents . furthermore , all agents within a bcc make the same total cooperative investment @xmath24 in the final state . however , the investments @xmath14 of one agent in different collaborations are in general different . the _ coordination _ of total investments @xmath15 , @xmath24 therefore arises although no agent has sufficient information to compute the total investment made by any other agent . we emphasize that the level of investments , which the agents approach is not set rigidly by external constraints but instead depends on the topology of the network of collaborations that is formed dynamically . this is evident for instance in differences of up to 20 % between the level of investment that is reached in different bccs of the same network . to understand how coordination of investment arises , we now formalize the observations made above . we claim that in our model in the final state the following holds : within a bcc ( i ) every agent makes the same total investment , and ( ii ) either all bidirectional links receive the same total investment or there are exactly two different levels of total investment received by bidirectional links . for reasons described below , the case of two different levels of total investment per link is only very rarely encountered . in this case every agent can have at most one bidirectional link that is maintained at the lower level of investment . we first focus on property ( i ) . this property is a direct consequence of the stationarity of the final state . consider a single link @xmath10 . since both investments , @xmath14 and @xmath25 , enter symmetrically into @xmath15 , the derivative of the benefit with respect to either investment is @xmath26 . thus , if @xmath27 , the stationarity conditions @xmath28 require @xmath29 this stipulates that the slope of the cost of the two interacting agents must match the slope of the shared benefit in the stationary state ( fig . [ pocfig2 ] ) . due to the symmetry of @xmath30 , @xmath31 holds for all @xmath32 . therefore , eq . implies @xmath33 . as we assumed @xmath5 to be superlinear , @xmath34 is injective and it follows that @xmath35 , such that @xmath1 and @xmath3 , are at a point of identical total investment . iterating this argument along a sequence of bidirectional links yields ( i ) . let us remark that the stationarity of vanishing investments may be fixed due to the external constraint that investments have to remain non - negative . the stationarity condition for vanishing and uni - directional links , analogous to eq . , is therefore @xmath36 because of the inequalities that appear in this equation , the argument given above does not restrict the levels of total investment found in different components . for similar reasons agents that are only connected by unidirectional links can sustain different levels of investment , which is discussed in sec . [ secexploitation ] . we note that , although the network of potential interactions is fully connected , no information is transfered along vanishing links . therefore , the equation of motion , eq . [ timeevolution ] , should be considered as a local update rule , in the sense that it only depends on the state of the focal agent and on investments received from a small number of direct neighbours . in order to understand property ( ii ) we consider multiple links connecting to a single agent @xmath1 . in an equilibrium the investment into each of the links has to be such that the slope of the benefit function of each link is identical . otherwise , the payoff could be increased by shifting investments from one link to the other . since the benefit function is sigmoidal , a given slope can be found in at most two points along the curve : one above and one below the inflection point ( ip ) . by iteration , this implies that if a stationary level of investment is observed in one link , then the investment of all other links of the same bcc is restricted to one of two values , which amounts to the first sentence of ( ii ) . for understanding why the case of two different levels of investments is rarely encountered the stability of steady states has to be taken into account . a local stability analysis , based on linearisation and subsequent application of jacobi s signature criterion , is presented in the appendix . we show that for a pair of agents @xmath10 connected by a bidirectional link , stability requires @xmath37 and every pair of links @xmath10 and @xmath38 connecting to the same agent @xmath1 has to satisfy @xmath39 note that eq . does not stipulate the sign of @xmath40 as it only implies @xmath41 . as eq . applies also to the link @xmath38 , the same holds for @xmath42 . we therefore have to consider three different cases when testing the compatibility of eq . with eq . : a ) : : @xmath43 and @xmath44 , ( both investments above the ip ) b ) : : @xmath45 and @xmath46 , ( both investments below the ip ) c ) : : @xmath45 and @xmath44 ( one investment above and one below the ip ) . in case a ) , is trivially fulfilled as the left hand side has positive and the right hand side negative sign . in case b ) , eq . and . are incompatible : estimating the lower bound of the right hand side of using the relation @xmath47 leads to the contradiction @xmath48 this shows that in a stable stationary state , every agent can at most have one link receiving investments below the ip . in case c ) , eq . can in principle be satisfied . however , the equation still imposes a rather strong restriction on a positive @xmath49 requiring high curvature of the benefit function close to saturation . the restriction becomes stronger , when the degree of agent @xmath1 increases @xcite . bilateral links with investments below the ip can be excluded entirely , if the benefit function approaches saturation softly , so that the curvature above the inflection point remains lower or equal than the maximum curvature below the inflection point . for such functions , every pair @xmath50 of solutions to the stationarity condition @xmath51 yields a pair of coefficients @xmath52 violating . in this case only configurations in which all links receive investments above the ip can be stable and hence all links produce the same benefit in the stable stationary states . this explains why the case of two different levels of cooperation is generally not observed in numerical investigations if realistic cost and benefit functions are used . for understanding the central role the ip plays for stability consider that in the ip the slope of @xmath4 is maximal . therefore , links close to the ip make attractive targets for investments . if the total investment into one link is below the ip then some disturbance raising ( lowering ) the investment increases ( decreases ) the slope , thus making the link more ( less ) attractive for investments . hence , below the ip , a withdrawal of resources by one of the partners , no matter how slight , will make the collaboration less attractive , causing a withdrawal by the other partner and thereby launching the interaction into a downward spiral . conversely , for links above the ip the gradual withdrawal of resources by one partner increases the attractiveness of the collaboration and is therefore compensated by increased investment from the other partner . in psychology both responses to withdrawal from a relationship are well known @xcite . the proposed model can therefore provide a rational for their observation that does not require explicit reference to long term memory , planning , or irrational emotional attachment . for our further analysis property ( ii ) is useful as it implies that , although our model is in essence a dynamical system , the bccs found in the steady states of this system can be analyzed with the tools of graph theory for undirected graphs . in the secs . [ secleaders ] , [ secgiantcomp ] we go one step further and treat not only the bcc but the whole network as an undirected graph . we thereby ignore the differences between directed and undirected links in order to study properties such as the degree- and component - size distributions before we continue in sec . [ secexploitation ] with a more detailed investigation of directed links and their topological implications . despite the coordination described above , the payoff extracted by agents in the final state can differ significantly . this is remarkable because the agents follow identical rules and the network of collaborations is initially almost homogeneous with respect to degree , link weights , and neighbourhood . because all bidirectional links in a bcc produce the same benefit , the total benefit an agent receives is proportional to the degree of the agent . by contrast , the cost incurred by an agent does not scale with the degree , but is identical for all agents in the bcc , because agents of high degree invest a proportionally smaller amount into their collaborations . topological positions of high degree thus allow agents to extract significantly higher benefits without requiring more investment . the payoff distribution in the population is governed by the degree distribution @xmath53 describing the relative frequency of agents with degree @xmath54 . figure [ pocfig3 ] shows a representative degree distribution of an evolved networks in the final state . while the finite width of the distribution indicates heterogeneity , the distribution is narrower , and therefore fairer , than that of an erds - rnyi random graph , which constitutes a null - model for randomly assembled network topologies . we verified that variance of the evolved network is below the variance of a random graph for the whole range of admissible mean degree @xmath55 in a network of given size . although the snowdrift game is not a zero - sum game , payoffs can not be generated arbitrarily . in order to sustain the extraction of high payoffs by agents of high degree , investments have to be redistributed across the network . in the definition of our model , we did not include the transport of resources directly . nevertheless , a redistribution of investments arises indirectly from the asymmetry of the agents investments . this is illustrated in fig . [ pocfig4 ] . consider for instance an agent of degree 1 . this agent necessarily focuses his entire investment on a single collaboration . therefore , the partner participating in this collaboration only needs to make a small investment to make the collaboration profitable . he is thus free to invest a large portion of his total investment into links to other agents of possibly higher degree . in this way investments flow toward the regions of high degree where high payoffs are extracted to explore the topological properties of the networks of collaborations in the final state further , we performed an extensive series of numerical integrations runs in which we varied all parameters in a wide range . these revealed that an important determinant of the topology is the mean degree @xmath56 , where @xmath57 denotes the number of links and @xmath0 the number of agents in the network . given two evolved networks with similar @xmath55 , one finds that the networks are also similar in other properties such as the component - size distribution , clustering coefficient , and the fraction of collaborations that are unidirectional . we therefore discuss the topological properties of the evolved networks as a function of @xmath55 , instead of the original model parameters . we first consider the expected size @xmath58 of a network component to which a randomly chosen agent belongs . in contrast to the bcc s discussed above , unidirectional collaborations are now taken into account in the computation of component sizes . the value of @xmath58 in the evolved network as a function of @xmath55 is shown in fig . [ pocfig5]a . the figure reveals that large components begin to appear slightly below @xmath59 . because of the difficulties related to integrating @xmath13 differential equations , our numerical investigations are limited to networks of up to 100 agents . while it is therefore debatable whether the observed behaviour qualifies as a phase transition , it can be related to the giant component transition commonly observed in larger networks . in the giant component transition a component is formed that scales linearly with network size . in the absence of higher correlations the transition occurs at @xmath60 @xcite , where @xmath61 is the mean excess degree of the network , i.e. , the number of additional links found connected to a agent that is reached by following a random link . in erds - rnyi random graphs , @xmath62 , therefore the giant component transition takes place at @xmath63 . in the present model the transition in @xmath58 is shifted to higher values of @xmath55 because of the nature of the underlying snowdrift game . the snowdrift game favors cooperation in the sense that for an agent of degree zero it is always advantageous to initiate an interaction . therefore @xmath63 is the lowest possible value that can be observed in evolved networks . further , any evolved network with @xmath63 invariably consists of isolated pairs , which precludes the existence of a giant component . finally , the relatively narrow degree distribution of the evolved networks implies @xmath64 and therefore @xmath65 at the transition . to estimate an upper limit for the connectivity at which the giant component transition occurs , it is useful to consider degree homogeneous networks . in these networks the degree distribution is a delta function and @xmath66 , so that the transition occurs at @xmath59 . in the networks evolved in the proposed model we can therefore expect a critical value of @xmath55 between one and two . based on numerical results we estimate that the giant component transition in the present model occurs at @xmath67 ( fig . [ pocfig5 ] ) . at this value a power - law distribution of component sizes , which is a hallmark of the giant - component transition , begins to show already in relative small networks with @xmath68 . while in sec . [ seccoordination ] we have mainly considered bidirectional links , and in sec.[secleaders ] and [ secgiantcomp ] only distinguished between vanishing and non - vanishing links , we will now focus on unidirectional links , which one partner maintains without reciprocation by the other . the presence of such links in collaboration networks was recently discussed in detail by @xcite . for the discussion below it is advantageous to consider the mean degree of agents in a connected component @xmath69 , where @xmath70 and @xmath71 are the number of agents and links in the component . note that in large components @xmath72 while the two properties can be significantly different in small components . in contrast to @xmath55 , @xmath73 allows us to infer global topological properties : components with @xmath74 are trees . components with @xmath75 contain exactly one cycle to which trees might be attached . and , components with @xmath76 contain more than one cycle , potentially with trees attached . as in the previous section , the term component refers to maximal subgraphs which are connected by bidirectional and/or unidirectional links . according to this definition a component may , beside one or more bccs , contain agents , which only have unidirectional links . in the following we denote the set of these agents as the non - bcc part of the component ( nbcc ) . for the sake of simplicity we focus on components which contain only one bcc , but note that the case of multiple bccs can be treated analogously . unlike the bcc , the nbcc is not a subcomponent but only a set of agents which are not necessarily connected . nevertheless , numerical results show that ( i * ) all nbcc agents make the same total investment @xmath77 and ( ii * ) all unidirectional links maintained by nbcc agents receive the same total investment @xmath78 . while property ( ii * ) can be understood analogously to property ( ii ) of bccs , property ( i * ) can not be ascribed to stationarity or stability conditions but seems to result from optimality restrictions . as a consequence of the properties ( i * ) and ( ii * ) the number of outgoing links @xmath79 is identical for all agents in the nbcc . so far we have decomposed a component into the bcc and the nbcc . within each subset , all agents make the same total investment , and all links receive the same total investment , therefore each subset can be characterized by two parameters , @xmath80 for bcc and @xmath81 for the nbcc . to recombine the subsets and infer properties of the whole component , we need to study the relation between these four parameters . the central question guiding our exploration is , why do agents not start to reciprocate the unidirectional investments . the lack of reciprocation implies that the unidirectional links are either less attractive or just as attractive as bidirectional links . we distinguish the two scenarios a ) : : @xmath82 , b ) : : @xmath83 . in case a ) the unidirectional collaborations are as attractive as targets for investments as bidirectional collaborations . in typical networks , where all remaining links receive investments above the ip this implies @xmath84 . furthermore , in case a ) the stationarity condition , eq . , requires that @xmath85 , which stipulates @xmath86 . therefore the whole component consists of agents making an investment @xmath87 and links receiving an investment @xmath88 . conservation of investments within a component implies @xmath89 and hence @xmath90 we know further that @xmath91 , where @xmath92 is the number of outgoing links of an agent in the nbcc . inserting @xmath93 in eq . yields @xmath94 , showing that unidirectional links that are as attractive as bidirectional links can only occur in components in which mean degree , @xmath95 , is an integer multiple of 2 . this matches the numerical data displayed in fig . [ pocfig6]a , which shows that @xmath96 is observed in components with @xmath75 and @xmath97 . it is remarkable that observing @xmath98 in a pair of collaborations is sufficient to determine the mean degree of the whole component . moreover components in which the mean degree is exactly 2 have to consist of a single cycle potentially with trees attached . in the numerical investigations we mostly observe cycles of bidirectional links to which trees of unidirectional links are attached , as shown in fig . [ pocfig7]b . in case b ) the bidirectional links are more attractive targets for investments than unidirectional links . in typical networks with @xmath99 this implies @xmath100 . now the stationarity condition , eq . , demands that @xmath101 , so that unidirectional links receive a higher investment than bidirectional links . by contrast the total investment made by an agent investing in bidirectional links is higher than the one made by agents investing in unidirectional links , i.e. @xmath102 this relationship restricts the connectivity in the bcc to @xmath103 , which implies @xmath76 , because the mean degree of the component can not be smaller than 2 if a subcomponent already has a degree greater than 2 . therefore , we find that unidirectional links that are less attractive than bidirectional links only occur in components in which the mean degree is larger than @xmath104 , but not an integer multiple of 2 ( cf . fig . [ pocfig6]a ) . as such links are only found at @xmath55 beyond the giant component transition they occur typically in large components as shown in fig . [ pocfig1 ] . in numerical investigations , we also observe some unidirectional links in components with @xmath74 ( cf . [ pocfig6]b ) . to explain these we have to consider case b ) but relax the assumption that both , @xmath78 and @xmath105 are above the ip . thus , we obtain case c ) , about which we know that the unidirectional links are less attractive than bidirectional links , @xmath106 , and that the unidirectional link only receives investments from one agent , i.e. , @xmath107 . moreover , @xmath74 implies @xmath108 and therefore @xmath109 . therefore @xmath110 which shows that unidirectional links can only appear in components with @xmath74 if the investment received by unidirectional links is smaller than the investment received by bidirectional links . satisfying @xmath111 and @xmath112 simultaneously requires @xmath113 . the components with @xmath74 , in which such links are found , are trees formed by a core of bidirectional links , to which individual agents are attached by unidirectional links ( fig . [ pocfig7]a ) . chains of unidirectional links , as we have observed in case a ) , can not appear for @xmath74 as this would mean that some agents would have one incoming and one outgoing link below the ip , which is ruled out by a trivial extension of the reasoning from sec . [ seccoordination ] . in this paper we have proposed a model for the formation of complex collaboration networks between self - interested agents . in this model the evolving network is described by a large system of deterministic differential equations allowing agents to maintain different levels of cooperation with different partners . we showed analytically that bidirectionally communities are formed , in which every agent makes the same total investment and every collaboration provides the same benefit . in contrast to models for cooperation on discrete networks , the present model thereby exhibits a high degree of coordination which can be interpreted as a precursor of a social norm . we emphasized that coordination is generally achieved although single agents possess insufficient information for computing the total investment made by any other agent and although the level of cooperation that is reached in a community is not fixed rigidly by external constraints . despite the high degree of coordination , we observed the appearance of privileged agents , reminiscent of the leaders emergind in @xcite . in the model proposed in the present paper , the privileged agents hold distinguished topological positions of high degree centrality allowing them to extract much higher payoffs than other agents , while making the same cooperative investment . however , we found that in the absence of further mechanism reinforcing differences the assembled topologies were fairer than random graphs . although our primary aim was to investigate the formation of social networks , some aspects of the behavior of social agents are reminiscent of results reported in psychology . for instance our investigation showed that agents can react to the withdrawal of investment by a partner either by mutual withdrawal of resources or by reinforcing the collaboration with increased investment . our analysis provides a rational which links the expected response to the withdrawal of resources to an inflection point of an assumed benefit function . furthermore , we investigated under which conditions non - reciprocated collaborations appear . here , our analysis revealed that such unidirectional collaborations can appear in three distinct scenarios , which can be linked to topological properties of the evolving networks . in particular exploited agents whose investments are not reciprocated invest less than the average amount of resources in their links when occurring in small components , but more than the average amount , when integrated in large components . we believe that the results from the proposed model can be verified in laboratory experiments in which humans interact via a computer network . such experiments may confirm the topological properties of the self - organized networks reported here and may additionally provide insights into the perceived cost and benefit functions that humans attach to social interactions . furthermore , results of the proposed model may be verified by comparison with data on collaboration networks between people , firms or nations . this comparison may necessitate modifications of the model to allow for instance for slightly different cost functions for the players . most of these extensions are straight forward and should not alter the predictions of the model qualitatively . for instance in the case of heterogeneous cost functions , players will make different total investments , but will still approach an operating point in which the slope of their cost function is identical . further , coordination should persist even if the network of potential collaborations is not fully connected . finally , but perhaps most importantly our analytical results do not rely heavily on the assumption that only two agents participate in each collaboration . most of the results can therefore be straight - forwardly extended to the case of multi - agent collaborations . our analytical treatment suggests that the central assumption responsible for the emergence of coordination is that the benefit of a collaboration is shared between the collaborating agents , but is independent of their other collaborations , whereas the cost incurred by an agent s investment depends on the sum of all of an agent s investments . because this assumption seems to hold in a relatively large range of applications we believe that also the emergence of coordination and leaders by the mechanisms described here should be observable in a wide range of systems . the analysis presented in this paper has profited greatly from the dual nature of the model , combining aspects of dynamical systems and complex network theory . in particular our analytical investigations were based on the application of jacobi s signature criterion to the system s jacobian matrix . apart from the symmetry of the jacobian , this ` double - jacobi ' approach does not depend on specific features of model under consideration . the same approach can therefore be used to address significant extensions of the present model . we therefore believe that also beyond the field of social interactions , the double - jacobi approach will prove to be a useful tool for the analytical exploration of the weighted adaptive networks that appear in many applications . to determine the local asymptotic stability of the steady states we study the jacobian matrix @xmath114 defined by @xmath115 . the terms contained in this matrix can be grouped into three different types @xmath116 albeit evaluated at different points . for reasons of symmetry @xmath117 and consequentially @xmath118 , and @xmath119 . ordering the variables according to the mapping @xmath120 the jacobian can be written in the form @xmath121 which is shown here for @xmath122 . as each cooperation @xmath10 is determined by a pair of variables @xmath123 , each @xmath124 occurs twice forming quadratic subunits with the corresponding entries @xmath125 and @xmath126 . subsequently , we restrict ourselves to the submatrix @xmath127 of @xmath128 , which only captures variables @xmath14 belonging to ` non - vanishing ' links . as argued before , ` vanishing links ' , i.e. links with @xmath129 , are subject to stationarity condition . if @xmath130 , their stability is due to the boundary condition @xmath131 and is independent of the second derivatives of @xmath5 and @xmath4 . hence , they can be omitted from the subsequent analysis . this means in particular that the spectra of different topological components of the network decouple and can thus be treated independently . all eigenvalues of the real , symmetric matrix @xmath127 are real . according to jacobi s signature criterion the number of negative eigenvalues equals the number of changes of sign in the sequence @xmath132 where @xmath133 is the rank of @xmath127 and @xmath134 , @xmath135 @xcite . in a stable system the sequence has to alternate in every step . a necessary condition for stability is therefore alternation in the first steps @xmath136 . by means of an even number of column and row interchanges the above stated form of @xmath127 can always be transformed such that the first @xmath137 block reads @xmath138 since we assume that @xmath10 is a non - vanishing link , and , hence , @xmath1 and @xmath3 to be in the same component , both agents make the same total investment @xmath87 . it follows from definition that @xmath139 and therewith that @xmath140 . thus , the sequence @xmath136 alternates if @xmath141 equation stipulates that @xmath142 and @xmath143 have the same sign . of the two possible scenarios @xmath144 the second is ruled out by eq . : if @xmath145 , it follows from eq . that @xmath146 , which contradicts @xmath147 . hence , the necessary conditions for stability , eqs . , , require @xmath148 if either agent @xmath1 or agent @xmath3 has another bilateral link , say @xmath38 , it is furthermore possible to transform @xmath127 by an even number of row and line interchanges such that the first @xmath137 block reads @xmath149 in this representation the sequence @xmath136 alternates if @xmath150 condition can then be written as @xmath151 inserting the definitions eqs.- in eqs . and yields the stability conditions cited in the main text as eqs . - . axelrod r and hamilton wd 1981 _ science _ * 211 * 13901396 doebeli m , hauert c and killingback t 2004 _ science _ * 306 * 859862 nowak ma and sigmund k 2004 _ science _ * 303 * 793799 nowak ma 2006 _ science _ * 314 * 15601563 axelrod r 1984 _ the evolution of cooperation _ ( new york : basic books ) nowak ma and may rm 1992 _ nature _ * 92 * 826829 burtsev m and turchin p 2006 _ nature _ * 440 * 10411044 hauert c and doebeli m 2004 _ nature _ * 428 * 643646 eguluz vm , zimmerman mg , cela conde cj and san miguel m 2005 _ am . j. soc . _ * 110(4 ) 9771008 santos fc and pacheco jm 2005 _ phys . lett . _ 95 * 09810414 ohtsuki h , hauert c , lieberman e and nowak ma 2006 _ nature _ * 441 * 502505 santos fc , santos md and pacheco jm 2008 _ nature _ * 454 * 213216 macy mw 1991 _ am . _ * 97(3 ) 808843 gould rv 1993 _ am . rev . _ 58(2 ) 182196 willers d 1999 _ network exchange theory _ ( westport : praeger ) fehr e and fischbacher u 2003 _ nature _ 425 * 785791 palla g , barabsi al and vicsek t 2007 _ nature _ * 446 * 664667 braha d and bar - yam y 2009 _ adaptive networks _ ( heidelberg : springer ) 3950 gross t and blasius b 2008 _ jrs interface _ * 5 * 259271 gross t and sayama h ( eds . ) 2009 _ adaptive networks : theory , models , and data _ ( heidelberg : springer ) ashlock d , smucker md , stanley ea and tesfatsion l 1996 _ biosystems _ * 37 * 99125 bala v and goyal s 2001 _ j. econ . theory _ * 17 * 101120 bornholdt s and rohlf t 2000 _ phys . lett . _ * 84 * 6114 pascuski m , bassler ke and corral a 2000 _ phys . rev . _ * 84 * 31853188 for a collection of respective publications see http://adaptive-networks.wikidot.com/publications skyrms b and pemantle r 2000 _ proc . usa _ * 97 * 93409346 zimmermann mg , eguluz vm san miguel m and spadaro a 2000 _ adv . complex syst . _ * 3 * 283297 zimmermann mg , eguluz vm and san miguel m 2004 _ phys . e _ * 69 * 065102 zimmermann mg and eguluz vm 2005 _ phys . e _ * 72 * 056118 fu f , wu t and wang l 2008 _ phys . e _ * 79 * 036101 pacheco jm , traulsen a and nowak ma 2006 _ j. theor . biol . _ * 243 * 437443 pacheco jm , traulsen a and nowak ma 2006 _ phys . _ * 97 * 25810314 szolnoki a , perc m and danku z 2008 _ euro . phys . _ * 84 * 50007 van segbroeck s , santos fc , lenaerts t and pacheco jm 2009 _ phys . _ * 102 * 058105 szolnoki a and perc m 2009 _ epl _ * 86 * 30007 poncela j , gmez - gardees j , flora lm , snchez a and moreno y 2008 _ plos one _ * 3 * e2449 poncela j , gmez - gardees j , traulsen a and moreno y 2009 _ new j. phys . _ * 11 * 083031 fu f , wu t and wang l 2009 _ phys . rev . e _ * 79 * 036101 suzuki r , kato m and arita t 2008 _ phys . e _ * 77 * 021911 ebel h and bornholdt s 2002 _ preprint _ arxiv : cond - mat/0211666 zschaler g , traulsen a and gross t 2009 _ preprint _ arxiv:0910.0940 biely c , dragosits k and thurner s 2007 _ physica d _ * 228 * 4048 szolnoki a and perc m 2009 _ new j. phys . _ * 11 * 093033 koenig md , battiston s , napoletano m and schweitzer f 2008 _ preprint _ cer - eth working paper 08/95 tomassini m , pestelacci e and luthi l 2010 _ biosystems _ * 99 * 5059 oliver p , marwell g and teixeira r 1985 _ am . j. sociol . _ * 91 * 522556 heckathorn dd 1996 _ am . rev . _ * 61 * 250277 this can be shown by taking determinants @xmath152 with @xmath153 into account . baxter la 1984 _ j soc . pers . relat . _ * 1 * 2948 newman me 2003 _ siam review _ * 45 * 167256 zeidler e , hackbusch w , schwarz hr and hunt b 2004 _ oxford user s guide to mathematics _ ( new york : oxford university press )	we study the self - assembly of a complex network of collaborations among self - interested agents . the agents can maintain different levels of cooperation with different partners . further , they continuously , selectively , and independently adapt the amount of resources allocated to each of their collaborations in order to maximize the obtained payoff . we show analytically that the system approaches a state in which the agents make identical investments , and links produce identical benefits . despite this high degree of social coordination some agents manage to secure privileged topological positions in the network enabling them to extract high payoffs . our analytical investigations provide a rationale for the emergence of unidirectional non - reciprocal collaborations and different responses to the withdrawal of a partner from an interaction that have been reported in the psychological literature . cooperation is the basis for complex organizational structures in biological as well as in social systems @xcite . the evolutionary and behavioural origin of cooperation is a subject of keen scientific interest , because the ubiquity of cooperation in nature seems to defy the often high costs incurred by the cooperating agent @xcite . evolutionary game theory has identified several mechanism allowing for the evolution and persistence of costly cooperation @xcite . in particular the emergence of cooperation is promoted if the interacting agents are distributed in some ( potentially abstract ) space , so that only certain agents can interact at any given time @xcite . in the context of social cooperation spatial structure can be appropriately modeled by a complex network , in which nodes represent agents , while the links correspond to collaborations . the topology of this network , i.e. , the specific configuration of nodes and links , has been shown to be of central importance for the level of cooperation that evolves @xcite . in social networks the topology is not static , but reacts to the behaviour of the agents @xcite . this defines an inherent dynamical interplay : while the agents s behaviour may depend on their topological neighbourhood , this neighbourhood is , at least in part , shaped through the agent s behavioural choices . networks containing such an dynamical interplay between the state of the nodes and the networks topology are called _ adaptive networks _ @xcite . while adaptive networks have been studied for some time in the social literature ( e.g. @xcite ) , pioneering work @xcite only recently triggered a wave @xcite of detailed dynamical investigations in physics . recent publications discuss simple cooperative games such as the one - shot prisoner s dilemma @xcite , the iterated prisoner s dilemma@xcite , and the snowdrift game @xcite on adaptive networks . they showed numerically and analytically that a significantly increased level of cooperation can be achieved if individuals are able rewire their links @xcite if links are formed and broken @xcite or if new agents are added to the network @xcite . moreover , it has been shown that the adaptive interplay between the agents strategies and the network topology can lead to the emergence of distinguished agents from an initially homogeneous population @xcite . while important progress has been made in the investigation of games on adaptive networks , it is mostly limited to discrete networks , in which the agents can only assume a small number of different states , say , unconditional cooperation with all neighbours and unconditional defection . by contrast , continuous adaptive networks have received considerably less attention @xcite . most current models therefore neglect the ability of intelligent agents to maintain different levels of cooperation with different self - chosen partners @xcite . in this paper we propose a weighted and directed adaptive network model in which agents continuously and selectively reinforce advantageous collaborations . after a brief description of the model , we show in sec . [ seccoordination ] that the network generally approaches a state in which all agents make the same total cooperative investment and every reciprocated investment yields the same benefit . despite the emergence of this high degree of coordination , the evolved networks are far from homogeneous . typically the agents distribute their total investment heterogeneously among their collaborations , and each collaborations receives different investments from the partners . in sec . [ secleaders ] , we show that this heterogeneity enables resource fluxes across the network , which allow agents holding distinguished topological positions to extract high payoffs . thereafter , in sec . [ secgiantcomp ] , we investigate further topological properties of the evolved networks and identify the transition in which large cooperating components are formed . finally , in sec . [ secexploitation ] , we focus on the appearance of unidirectional ( unreciprocated ) investments . specifically , we identify three distinct scenarios in which unidirectional collaborations can arise and discuss their implications for the interaction topology . our conclusions are summarized in sec . [ secdiscussion ] .
quantum criticality is currently being pursued in a number of strongly correlated materials ranging from intermetallic rare earth compounds to insulating magnets and even engineered systems . this interest is spurred in part by the observation that unconventional superconductivity commonly occurs near quantum critical points , as well as by mounting experimental evidence of the inadequacy of the traditional theoretical approach to continuous transitions at absolute temperature @xmath7 . this approach , commonly referred to as the spin - density - wave or hertz - millis - moryia picture , is based on an extension to zero temperature of landau s theory of order - parameter fluctuations.@xcite the evidence for the breakdown of the spin - density - wave picture is particularly compelling in the context of intermetallic rare - earth compounds near an antiferromagnetic instability . as a result , a major thrust in the quest for a theoretical framework of critical phenomena beyond the landau - ginzburg - wilson paradigm has centered around the question of how kondo screening can be destroyed at a @xmath7 magnetic - ordering transition . an extensively discussed alternative to the spin - density - wave picture is local quantum criticality,@xcite in which kondo screening itself becomes critical precisely at the magnetic - ordering transition . in the vicinity of such a kondo - destruction quantum critical point ( qcp ) , scaling properties of the order parameter are very different from those predicted by spin - density - wave theory . for example , inelastic neutron - scattering experiments have shown that the dynamical spin ( order - parameter ) susceptibility exhibits frequency - over - temperature ( or @xmath5 ) scaling , and displays a fractional exponent in the frequency and temperature dependences over a wide - range of momentum space.@xcite in addition , hall measurements have provided evidence that the single - particle relaxation rate is linear in temperature,@xcite a behavior that arises naturally from @xmath5 scaling of the single - particle green s function . a microscopic model of local quantum criticality in heavy - fermion metals has been provided@xcite through extended dynamical mean - field theory , which maps the kondo lattice model to a single magnetic impurity coupled to both a fermionic conduction band and to one or more bosonic baths representing collective spin fluctuations . the critical point in the lattice model is then captured in terms of a critical destruction of the kondo effect in the single - impurity problem . such a construction gives rise to a qcp that is spatially local but has interacting critical modes in imaginary time , and that correctly captures @xmath5 scaling in the order - parameter susceptibility . while the full extended dynamical mean - field treatment simulates the lattice through self - consistent determination of the band and bath densities of states , valuable insight can be gained by studying kondo - destruction qcps in pure impurity problems . this work investigates the quantum - critical destruction of the kondo effect in pseudogap variants of the ising - symmetric ( or easy - axis ) bose - fermi anderson ( bfa ) and bose - fermi kondo ( bfk ) models . in each model , a local degree of freedom couples to a band of conduction electrons having a density of states that vanishes as @xmath0 on approach to the fermi energy ( @xmath1 ) . the local degree of freedom is also coupled via the @xmath8 component of its spin to a bosonic bath having a density of states proportional to @xmath9 for frequencies up to some cutoff @xmath10 ( i.e. , for @xmath11 ) . in the presence of a metallic conduction band ( corresponding to an exponent @xmath12 ) , the bfa and bfk models with a sub - ohmic bosonic bath characterized by a bath exponent @xmath13 feature a second - order quantum phase transition ( qpt)@xcite between a kondo - screened ( fermionic strong - coupling ) phase and a localized ( kondo - destroyed ) phase in which the bosons asymptotically suppress spin - flip scattering and a residual impurity moment survives to @xmath7 . for @xmath14 , the ising - symmetry bfa and bfk models are thought@xcite to share the same critical properties as the corresponding sub - ohmic spin - boson model , which features an interacting qcp characterized by critical exponents that vary continuously with the bath exponent @xmath15 . in the absence of the bosonic bath , by contrast , the pseudogap bfa and pseudogap bfk models reduce , respectively , to the pseudogap anderson and pseudogap kondo models , in which the depression of the low - energy density of states impedes the formation of a many - body kondo screening cloud and gives rise to qpts@xcite between a kondo phase and a localized phase in which the impurity exhibits a free spin-@xmath16 at @xmath7 . the qpts in the pseudogap anderson and kondo models belong to the same universality class@xcite and are described by critical exponents that vary with the band exponent @xmath17 . the pseudogap versions of the bfa and bfk models offer fascinating possibilities for nontrivial interplay between two different mechanisms for destruction of the kondo effect ; an interplay that we investigate in this work . an su(2)-symmetric version of the pseudogap bfk model ( in which the cartesian components of the impurity spin couple to different bosonic baths sharing the same exponent @xmath15 ) has been studied via perturbative renormalization - group ( rg ) methods,@xcite while the ising - symmetric case of bosonic coupling to the @xmath8 component of the impurity spin has been the subject of preliminary investigation using the numerical renormalization group ( nrg).@xcite a spinless variant of the model ( coupling the impurity charge to a bosonic bath ) has been addressed by perturbative and numerical rg techniques.@xcite in all these previous studies , the order - parameter susceptibility has been found to exhibit an anomalous @xmath18 variation in the quantum - critical regime with an exponent @xmath19 independent of the band exponent @xmath17 . one of the objectives of the present work is to investigate more carefully the universality of this observation . another motivation for the present study is to expand our understanding of the conditions that lead to @xmath5- or dynamical scaling of critical correlation functions near unconventional quantum criticality . it is well known that a quantum impurity model with a bulk component that is conformally invariant can be described by a boundary conformal field theory.@xcite in any such theory , a conformal mapping can be used to obtain correlators at temperatures @xmath20 from their @xmath7 counterparts . in particular , the zero - temperature two - point correlator of a primary conformal field @xmath21 with scaling dimension @xmath22 exhibits a power - law decay @xmath23 . this gives rise , via a conformal mapping , to the scaling form@xcite @xmath24 the fourier transform of eq . can be performed analytically and yields a function of @xmath25 that can be analytically continued to real frequencies provided@xcite that @xmath26 . therefore , demonstration that a system obeys eq . with @xmath26 is sufficient to show the presence of @xmath5 scaling , a characteristic feature of interacting qcps . in the pseudogap anderson model , the conduction - band density of states breaks conformal invariance , while in the metallic ( @xmath12 ) bfk model , the conduction band is conformally invariant but the bosonic bath is not . nonetheless , local correlators of each model have been found@xcite to exhibit a boundary conformal scaling form in imaginary time , consistent with eq . ; such properties have been attributed to an enhanced symmetry at the qcps . it is an interesting question whether the scaling form eq . also applies to the critical correlators of the pseudogap bfa and bfk models , where the densities of states of the fermionic and bosonic baths both break conformal invariance in the bulk . we address these issues through a combination of techniques . for the particle - hole - symmetric pseudogap bfa model , we use a continuous - time quantum monte carlo ( ct - qmc ) method,@xcite which stochastically samples a perturbation expansion in the anderson hybridization , to probe static and dynamical quantities . we also use the bose - fermi extension of the nrg method@xcite to resolve the critical spectrum and extract static critical exponents of the pseudogap bfk model . our results show that the two models are in the same universality class . we find that within different ranges of the exponents @xmath17 and @xmath15 , the measured critical exponents are determined by the fermionic band alone , by the bosonic bath alone , or by both the fermions and the bosons . we show that both the single - particle green s function and the local spin susceptibility obey the scaling form of eq . with exponents @xmath27 , proving that each correlator obeys @xmath5 scaling in both the quantum coherent ( @xmath28 ) and the relaxational ( @xmath29 ) regimes . agreement with nrg results for static quantities confirms the ability of the ct - qmc approach to study quantum - critical properties of models involving bosons . the remainder of the paper is organized as follows : section [ sec : models ] introduces the pseudogap bose - fermi models and sec . [ sec : methods ] briefly describes the numerical methods used to solve these models . an overview of the phase diagram in sec . [ sec : phases ] is followed in sec . [ sec : critical ] by a detailed description of the quantum - critical properties . these results are summarized and discussed in sec.[sec : summary ] . the bose - fermi anderson impurity model with ising - symmetric bosonic coupling is described by the hamiltonian @xmath30 where @xmath31 annihilates a conduction - band electron with wave vector @xmath32 , energy @xmath33 , and spin @xmath8 component @xmath34 with @xmath35 ( or @xmath36 ) or @xmath37 ( or @xmath38 ) ; @xmath39 annihilates an impurity electron with energy @xmath40 and spin @xmath8 component @xmath34 ; @xmath41 ; @xmath42 annihilates a boson of energy @xmath43 ; and @xmath44 is the number of unit cells in the host ( i.e. , the number of distinct @xmath32 points ) . the other energy scales entering eq . are the coulomb repulsion @xmath45 between two electrons in the impurity level , the local hybridization @xmath46 between the impurity level and the conduction band , and the coupling @xmath47 between the @xmath8 component of the impurity spin and the bosonic bath . we focus in this paper on cases @xmath48 corresponding to particle - hole - symmetric impurities , but briefly discuss the effect of breaking this symmetry in sec . [ subsec : p - h - asymm ] . over a wide region of its parameter space , the low - energy properties of @xmath49 can be mapped@xcite via a schrieffer - wolff transformation onto the ising - symmetry bose - fermi kondo hamiltonian : @xmath50 where @xmath51 is the kondo coupling , @xmath52 parameterizes nonmagnetic potential scattering , and @xmath53 is the on - site conduction - band spin , with @xmath54 being a vector of pauli spin matrices . for @xmath55 , the potential scattering vanishes for electrons on the fermi surface , and we can set @xmath56 . although the bare kondo exchange term is su(2 ) symmetric , the bosonic coupling breaks spin rotational invariance . as a result , the rg description of the bfk model in terms of renormalized couplings requires consideration of an anisotropic exchange @xmath57 . for both models , we assume a conduction - band ( fermionic - bath ) density of states @xmath58 with a power - law pseudogap described by @xmath3 , and a sub - ohmic bosonic bath specified by @xmath59 with @xmath4 . the pseudogap density of states @xmath60 leads to a bfa model with the hybridization function @xmath61 , where @xmath62 . various limiting cases of eqs . and have been studied previously . for @xmath63 , the pseudogap bose - fermi models simplify to their pure - fermionic counterparts , in which a pseudogap critical point separates kondo and free - moment phases ; at particle - hole symmetry , this critical point exists only for @xmath3 ( ref . ) . in the absence of the conduction band , @xmath64 and @xmath65 both reduce to the sub - ohmic spin - boson model in zero transverse field , which has two degenerate ground states in which the bosonic coupling localizes the impurity either in its up- or down - spin configuration . for @xmath12 , the pseudogap bfa and bfk models reduce to their metallic counterparts where the critical properties for ising symmetry are thought@xcite to coincide with those of the spin - boson model for @xmath4 . in the sections that follow we show that the quantum criticality in the _ full _ pseudogap bose - fermi models described by eqs . and falls into one of three distinct types , depending on the values of the bath exponents @xmath17 and @xmath15 , with one of these types being governed by a mixed bose - fermi qcp unlike those seen in any of the limiting cases . this work uses a form of the ct - qmc method described in ref .. to make @xmath66 suitable for application of the approach , we apply a canonical transformation to eliminate the term linear in bosonic operators . this is achieved by the generator @xmath67 , which transforms eq . to @xmath68 where @xmath69 $ ] , @xmath70 , and @xmath71 . physically , the canonical transformation can be viewed as dressing the impurity with a bosonic cloud , which in turn renormalizes @xmath45 , @xmath40 , @xmath39 , and @xmath72 without breaking particle - hole symmetry . we are now in a position to calculate the partition function by expanding in the hybridization @xmath46 . this is similar to the approach in refs . and , except that we couple the bosonic bath to the @xmath8 component of the impurity s spin rather than to its occupancy . [ an extension of this method to the case of an su(2)-symmetric spin - boson coupling has recently been proposed.@xcite ] the resulting perturbation expansion is then sampled stochastically using a metropolis algorithm . tracing out the fermionic band , which is unchanged by the presence of bosons,@xcite allows the partition function to be recast as @xmath73 \ , w_k$ ] , where at perturbation order @xmath74 there are @xmath74 segments along the imaginary time axis , each one defined by a pair of renormalized impurity operators , @xmath75 and @xmath76 . the weight of a particular configuration of segments is @xmath77 , where @xmath78 is the weight of the band fermions , @xmath79 is the weight of the local configuration and @xmath80 is the weight of the bosonic bath . @xmath78 can be obtained as a product of determinants and @xmath79 is calculated in terms of the length and overlap of imaginary time segments ; for details see ref . . if the operators in a given segment configuration are time - ordered , with @xmath81 ( @xmath37 ) indicating that the @xmath82 operator acts at time @xmath83 to create ( annihilate ) an electron of spin @xmath8 component @xmath84 , then the bosonic weight is @xmath85 , where @xmath86 $ ] . tracing out the bosons and performing the momentum summation yields @xmath87 \biggr\}. \label{w_b}\ ] ] to calculate the function @xmath88 we replace the hard cut - off of the bosonic spectral function in eq . by an exponential cut - off , @xmath89 , which yields @xmath90,\end{aligned}\ ] ] where @xmath91 is the gamma function and @xmath92 is the hurwitz zeta function.@xcite now the local update procedure is identical to that in ref . . in all the ct - qmc calculations reported below , we have considered the low - temperature scaling limit ( @xmath93 ) and consequently dropped the terms @xmath94 from eq . . we have checked that retaining or discarding these terms does not change any of the universal features at the qcp such as critical exponents . we perform an unbiased ct - qmc study of the quantum - critical properties of the pseudogap bfa model by applying finite - temperature scaling to determine the location of the qcp . we measure powers of the local magnetization @xmath95^n \biggr\rangle,\ ] ] where @xmath96 $ ] , from which the binder cumulant @xmath97 is obtained . similar to the case without bosons@xcite we find that swap moves@xcite are essential to accurately calculate the binder cumulant . near classical phase transitions , the binder cumulant is a function of system size and temperature , and finite - size scaling allows one to obtain the critical temperature in the thermodynamic limit from classical monte carlo simulations of finite systems.@xcite for the current quantum - mechanical problem , the inverse temperature @xmath98 and coupling @xmath47 play the roles of system size and temperature , respectively , so taking the zero - temperature limit allows one to determine the critical coupling @xmath99 . in the vicinity of the critical point , the binder cumulant obeys the scaling form@xcite @xmath100 where @xmath101 is the correlation - length exponent defined in eq . below . therefore , finite - temperature scaling applied to the binder cumulant can be used to extract the correlation - length critical exponent . we also use ct - qmc to measure the single - particle green s function @xmath102 [ @xmath103 in zero magnetic field ] and the local spin susceptibility @xmath104 , as described in ref .. for the susceptibility calculations presented in this paper , we have used the segment representation , and have checked that the results are consistent with those calculated using a `` matrix '' formalism.@xcite the static local spin susceptibility can then be determined as @xmath105 , where we have set the land g factor and the bohr magneton to unity . for the noninteracting problem ( @xmath106 ) , with @xmath107 , we find agreement within numerical accuracy between the exact and ct - qmc results for both the green s function and the local spin susceptibility.@xcite the ct - qmc results reported below were obtained for @xmath107 , @xmath108 , @xmath109 , @xmath110 , and for two different bath exponents : @xmath111 and @xmath112 . it is known from previous work@xcite that in the absence of bosons , the qcp point for @xmath107 and @xmath109 occurs at @xmath113 . therefore , by fixing @xmath110 and adjusting the bosonic coupling @xmath47 , we are able to disentangle the pseudogap bose - fermi qcp from the pure - fermionic pseudogap qcp . the bose - fermi nrg@xcite treatment of the hamiltonian entails three key steps : ( i ) partition of the fermionic and bosonic baths described by @xmath60 and @xmath114 into logarithmic bins spanning the energy ranges @xmath115 , where @xmath116 , 2 , 3 , @xmath117 and @xmath118 is the wilson discretization parameter . within each logarithmic bin , the continuum is replaced by a single state , namely , the linear combination of states that couples to the impurity . ( ii ) tridiagonalization of the baths , yielding the hamiltonian @xmath119 \notag \\ & + \omega_c \sum_{m=0}^{\infty } \bigl [ e_m b_m^{\dag } b_m^{{\phantom{\dag } } } + t_m \bigl ( b_m^{\dag } b_{m-1}^{{\phantom{\dag } } } + \text{h.c } \bigr)\bigr ] \\ & + f^2 \rho_0 j \vec{s } \cdot \sum_{\sigma,\sigma ' } f^{\dag}_{0\sigma } \vec{\sigma}_{\sigma \sigma ' } f_{0\sigma'}^{{\phantom{\dag } } } + b k_0 g s_z \bigr ( b_0^{{\phantom{\dag } } } + b_0^{\dag } \bigr ) . \notag\end{aligned}\ ] ] all information about @xmath60 is encoded in @xmath120 and the dimensionless coefficients @xmath121 and @xmath122 , while @xmath114 is parametrized by @xmath10 and the dimensionless quantities @xmath123 and @xmath124 . for a particle - hole - symmetric @xmath60 such as that in eq . , @xmath125 for all @xmath126 . for large values of of @xmath126 , the remaining tight - binding coefficients satisfy @xmath127 and @xmath128 . ( iii ) iterative solution of eq . on fermionic chains restricted to sites @xmath129 with @xmath130 , 1 , 2 , @xmath117 . due to the faster decay of the bosonic hopping coefficients with increasing index @xmath131 , the bosonic chain is restricted to @xmath132 , meaning that a site is added to this chain only at even values of @xmath133 . as in the conventional ( pure - fermionic ) nrg , the @xmath134 many - body states of lowest energy are retained to form the basis for iteration @xmath135 . in problems involving bosonic baths , it is also necessary to truncate the fock space on each site of the bosonic chain . in this work we employed a basis of boson number eigenstates @xmath136 . schematic zero - temperature phase diagram for the pseudogap bfk model on the plane spanned by the bosonic coupling @xmath47 and the kondo exchange coupling @xmath51 . the system undergoes a qpt at any crossing of the solid line @xmath137 representing the boundary between the two stable phases : kondo and localized . in this work , we consider crossings achieved by varying just one model parameter , as exemplified by the horizontal and vertical dashed lines . the labels @xmath138 and @xmath139 mark the point along the phase boundary at which the temperature @xmath140 , the upper limit of the quantum - critical regime ( see fig . [ fig : crossover_scales ] ) , takes its maximum value . the terminus of the phase boundary at @xmath63 , @xmath141 corresponds to the qpt of the pure - fermionic kondo model . the pseudogap bfa model has a very similar phase diagram on the @xmath47-@xmath142 plane at any fixed , positive value of @xmath143.,height=172 ] the nrg calculation of critical exponents in the vicinity of a qcp is carried out as described previously for the metallic ( @xmath12 ) bfk model@xcite and as illustrated in a preliminary publication on the pseudogap bfk model.@xcite the local magnetization ( equivalent to @xmath144 above ) and the static local susceptibility are evaluated as @xmath145 where @xmath146 is a magnetic field that couples only to the impurity through an additional hamiltonian term @xmath147 . all results reported below were obtained for @xmath148 , using wilson discretization parameter @xmath149 and a bosonic truncation parameter @xmath150 , and retaining after each iteration @xmath151 many - body multiplets corresponding to approximately 900 eigenstates.@xcite experience from previous studies@xcite indicates that critical exponents calculated for these nrg parameter choices are well - converged with respect to discretization errors ( induced by the departure of @xmath152 from its continuum value 1 ) and truncation errors ( arising from restricting the values of @xmath134 and @xmath153 ) . over the range of bosonic exponents @xmath13 , the bose - fermi nrg yields critical exponents for the case @xmath12 that reproduce those obtained via nrg@xcite for the spin - boson model having the same exponent @xmath15 . it has been suggested in refs . and that , for @xmath154 , errors associated with the nrg calculation on a finite bosonic wilson chain lead to unphysical results with nontrivial exponents and hyperscaling . on the other hand , for this same range of @xmath15 , refs . have demonstrated an @xmath5 scaling for both the leading and sub - leading components of the self - energy , which provides evidence for the interacting nature of the fixed point . since this issue has yet to be fully resolved , we restrict ourselves in this paper to study of the range @xmath155 . in this section , we describe the two zero - temperature phases of the pseudogap bfa and bfk models that can be accessed by tuning one model parameter while holding all other parameters constant , as illustrated in the case of the bfk model by the horizontal and vertical dashed lines in fig . [ fig : phase_diagram ] . we will assume for the moment that the bosonic coupling @xmath47 is the parameter that is varied . it is important to emphasize that , if carried out at any temperature @xmath20 , such a variation invariably produces smooth crossovers in physical properties , represented schematically by the crossing of dashed lines on the @xmath156 vs @xmath47 diagram in fig.[fig : crossover_scales ] . only at @xmath7 is it possible to drive the pseudogap bfa or bfk model through a qpt at @xmath157 separating a kondo phase ( reached for @xmath158 ) from a localized phase ( accessed for @xmath159 ) . schematic @xmath156 vs @xmath47 diagram in the vicinity of the pseudogap bose - fermi qcp studied in this work . the qcp dominates the physics in a temperature window between crossover scales @xmath140 and @xmath160 ( dashed lines ) . for @xmath158 and @xmath161 , the system is in its kondo regime , whereas for @xmath159 and @xmath161 , the spin becomes decoupled from the conduction band and the system is in the localized regime . at @xmath7 ( only ) , cases @xmath158 and @xmath159 correspond to different phases , distinguished by the value of the order parameter @xmath162 defined in eq . .,height=144 ] the kondo phase can be characterized by the vanishing ( as @xmath163 ) of the long - imaginary - time value of the local spin susceptibility @xmath164 . correspondingly , the static local susceptibility @xmath165 approaches a constant at low temperatures ( see , for example , the squares in figs . [ fig : stat6 ] and [ fig : stat8 ] ) , signaling that the impurity spin is screened . since these behaviors ( and all other universal low - energy properties found for @xmath158 ) prove to be identical to those in the kondo phase of the pseudogap anderson and kondo models , and independent of the bosonic bath exponent @xmath15 , we associate them with an rg fixed point located at renormalized couplings @xmath63 and either @xmath166 or @xmath167 . in the localized phase , by contrast , @xmath164 approaches a constant @xmath168 in the limit @xmath163 and the static local susceptibility assumes the curie - law form @xmath169 ( triangles in figs . [ fig : stat6 ] and [ fig : stat8 ] ) characteristic of a free spin whose size is @xmath170 . in this low - temperature regime , the impurity spin is essentially decoupled from the conduction band and follows the fluctuations of the bosonic bath . the asymptotic low - energy properties are governed by an rg fixed point located at renormalized couplings @xmath171 and either @xmath172 or @xmath173 . the two phases described above are separated by a qpt occurring at bosonic coupling @xmath99 where the curie constant @xmath174 extrapolates continuously to zero as @xmath47 is decreased . as a result , @xmath175 can serve as an order parameter for the qpt,@xcite vanishing throughout the kondo phase and equaling the curie constant in the localized phase . however , it is conventional instead to take as the order parameter @xmath176 which rises continuously from zero on entry into the localized phase . if one fixes all parameters apart from @xmath142 and @xmath47 in the bfa model [ @xmath51 and @xmath47 in the bfk model ] , then the function @xmath177 [ @xmath178 defines the boundary between the kondo and localized phases . this boundary , shown for the kondo case as the solid line in fig . [ fig : phase_diagram ] , is anchored at @xmath63 by the qcp of the pure - fermionic pseudogap models . one of the central questions addressed in our work is whether the critical behavior reached by crossing the phase boundary at @xmath179 coincides with or differs from that for @xmath63 . our two numerical techniques lend themselves to different approaches for locating the phase boundary . in the ct - qmc treatment of the pseudogap bfa model , the low - temperature limit of the binder cumulant @xmath180 evolves continuously from @xmath181 in the kondo phase to @xmath182 in the localized phase . @xmath180 is independent of temperature at the critical bosonic coupling @xmath99 , but not at other nearby values of @xmath47 . thus , one can find @xmath99 via the intersection of curves @xmath180 vs @xmath47 for different ( low ) temperatures ( see figs.[fig : binder6 ] and [ fig : binder8 ] ) . within the nrg , one can identify the phase boundary of the pseudogap bfk model through examination of the asymptotic low - energy many - body spectrum , the @xmath163 values of thermodynamic properties such as the impurity contribution to the entropy ( @xmath183 in the kondo phase , and @xmath184 in the localized phase ) , or the static local spin susceptibility @xmath165 . in the present study of the pseudogap bfk model , rather than calculating @xmath99 as a function of @xmath51 , we have instead determined the critical kondo coupling @xmath185 for different values of @xmath47 . for given bath exponents @xmath17 and @xmath15 , points on the phase boundary at any @xmath179 are all found to share the same many - body spectrum and the same power laws in the quantum - critical regime @xmath186 ( see fig.[fig : crossover_scales ] ) . this universality strongly suggests that the quantum criticality is governed by a single qcp to which the system flows starting from any point on the curve @xmath187 . however , the upper temperature @xmath140 , which marks the energy scale at which rg flow first brings the system under the influence of the qcp , varies widely along the boundary . in order to provide the most accurate possible account of the quantum - critical properties , we focus below on nrg results obtained for the boundary point @xmath188 , @xmath189 that yields the highest value of @xmath140 for given @xmath6 , and can therefore be assumed to lie closest to the qcp . an advantage of this approach is that for certain @xmath6 pairs we find @xmath190 , making clear that in such cases the qcp is of pure - fermionic character . in what follows , we write @xmath191 for the bfa model and @xmath192 for the bfk model to denote the system s distance from the phase boundary , with @xmath193 describing the kondo phase and @xmath194 describing the localized phase . at any point lying on the phase boundary between the kondo and localized phases , the static susceptibility exhibits a temperature dependence @xmath195 for all temperatures in the quantum - critical window @xmath196 . here , the exponent @xmath197 is a universal property of the pseudogap bose - fermi qcp ( i.e. , @xmath197 depends only on the bath exponents @xmath17 and @xmath15 ) , whereas the crossover temperature @xmath140 and the prefactor @xmath198 can vary from point to point along the phase boundary ; as noted above , @xmath140 is highest for the boundary point that lies nearest to the qcp . close to but not precisely on the phase boundary , the system enters the quantum - critical regime once the temperature drops below roughly the same scale @xmath140 found for @xmath199 . however , further lowering the temperature produces a second crossover to an asymptotic regime @xmath200 governed either by the kondo fixed point ( for @xmath193 ) or by the localized fixed point ( for @xmath194 ) ; see fig . [ fig : crossover_scales ] . unlike @xmath140 , the lower crossover scale @xmath160 shows significant @xmath201 dependence , and vanishes continuously upon approach to the phase boundary according to @xmath202 where @xmath101 is the correlation - length exponent . in the absence of bosons , the exponent @xmath203 necessarily assumes the value @xmath204 found at the qcp of the pure - fermionic pseudogap anderson and kondo models . with a constant fermionic density of states ( i.e. , @xmath12 ) and isotropic , xy , or ising symmetry of the bosonic couplings , it is known@xcite that @xmath205 for @xmath206 reduces to the exponent @xmath207 of the spin - boson model . based on perturbative rg,@xcite it has been concluded that @xmath203 for a spin - isotropic version of the pseudogap bfk model is independent of @xmath17 , an observation that agrees with asymptotically exact results obtained in the dynamical large-@xmath133 limit@xcite where the symmetry group of the spin - isotropic pseudogap bfk model is generalized from su(2 ) to su(@xmath133).@xcite a preliminary nrg study@xcite of the easy - axis pseudogap bfk model also found @xmath208 . however , that study considered only exponent pairs @xmath6 for which @xmath209 , a regime in which it is quite plausible that the bosons should dominate the singular part of the spin response . in the present study we have also investigated cases where @xmath210 that offer better prospects for finding fermion - dominated spin dynamics . in what follows we show using both ct - qmc calculations and the nrg that the static magnetic critical exponent @xmath197 is simply the smaller of the exponents governing the cases of pure - fermionic and pure - bosonic critical kondo destruction : @xmath211.\ ] ] we focus primarily on the pseudogap exponent @xmath107 and two values of the bath exponent : @xmath111 and @xmath212 . for the pure - fermionic kondo model , the critical susceptibility is described by a temperature exponent @xmath213 obtained within the nrg,@xcite consistent with the value @xmath214 found using ct - qmc for the corresponding anderson model.@xcite ( here and throughout the remainder of the paper , a number in parentheses indicates our estimated uncertainty in the last digit . ) for the bose - fermi models , therefore , the cases @xmath215 and @xmath216 are representative of the regimes @xmath217 and @xmath218 , respectively . for each of these cases , we present ct - qmc results for the pseudogap bfa model and nrg results for the pseudogap bfk model , finding the critical behavior of the two models to be fully equivalent . we supplement the detailed results for @xmath107 , @xmath111 and @xmath212 with static critical exponents for a larger set of @xmath6 pairs as obtained for the bfk model using the nrg . in addition to @xmath197 and @xmath101 introduced in eqs . and , we consider exponents @xmath219 and @xmath220 defined via the relations@xcite @xmath221 if the qpt occurs below its upper critical dimension , yielding an interacting qcp , one expects the singular component of the free energy to take the form @xmath222 where @xmath223 is a scaling function . with this ansatz , the exponents @xmath219 and @xmath220 are related to @xmath101 and @xmath197 by hyperscaling relations @xmath224 in each of the cases we have studied , the value of @xmath197 is consistent with eq . , and wherever we have tested them , the hyperscaling eqs . and are well obeyed . based on the value of @xmath225 [ or alternatively , @xmath226as well as an analysis of the many - body spectrum at the qcp we are led to subdivide the region of the @xmath6 plane in which @xmath227 into two parts : one in which the fermions appear to play no role in the critical behavior , and another in which bosonic and fermionic fluctuations combine to produce critical behavior unlike that found in either the metallic ( @xmath12 ) bfk model or the pseudogap kondo model . figure [ fig : binder6 ] shows the variation with bosonic coupling @xmath47 of the binder cumulant @xmath180 of the pseudogap bfa model for @xmath107 , @xmath111 at different temperatures as calculated using ct - qmc . the intersection of the curves places the qcp at @xmath228 . ( color online ) binder cumulant @xmath180 vs bosonic coupling @xmath47 for @xmath107 and @xmath111 at different inverse temperatures @xmath229 . there is a clear intersection all curves at the critical coupling @xmath230 . error bars for binder cumulant calculations are obtained from a jackknife error analysis.,width=211 ] the scaling form eq . of the binder cumulant in the vicinity of the qcp can be used to extract the correlation - length exponent defined in eq . . as illustrated in fig . [ fig : nu6 ] , we obtain an excellent collapse of data taken at different temperatures with a fitted exponent @xmath231 . ( color online ) data collapse of the binder cumulant @xmath180 for @xmath107 and @xmath111 using eq . . an inverse correlation - length exponent @xmath232 produces an excellent collapse of data in the vicinity of the critical point.,width=240 ] having found @xmath99 , we are able to establish that the static local susceptibility has the expected temperature dependence in each phase and at the critical coupling , as illustrated in fig . [ fig : stat6 ] . in particular , for @xmath157 , @xmath233 follows eq . with @xmath234 over the lowest decade of temperature for which data were obtained : @xmath235 . ( for comparison purposes , we note that for @xmath63 , the kondo temperature is @xmath236 . ) to within numerical accuracy , we find that @xmath237 , in agreement with previous perturbative and numerical rg studies.@xcite ( color online ) static spin susceptibility @xmath233 from ct - qmc vs temperature @xmath156 for @xmath107 and @xmath111 in the kondo - screened phase ( squares ) , at the critical coupling ( circles ) , and in the localized phase ( triangles ) . at the critical coupling @xmath238 , @xmath233 diverges according to eq . with @xmath239.,height=211 ] in the bfk model , the convergence of the nrg many - body spectrum to the critical spectrum is fastest ( i.e. , the crossover scale @xmath140 is highest ) for @xmath240 , with @xmath241 . ( the value of @xmath185 was determined to roughly 10 significant figures . ) figure [ fig : stat6nrg ] superimposes the temperature dependence of the critical static local susceptibilities of the anderson and kondo models . the nrg results for the kondo model exhibit small oscillations around the dependence predicted in eq . . such oscillations , which are periodic in @xmath242 with period @xmath243 , are a known consequence of the nrg band discretization@xcite that can be reduced in amplitude by working with smaller values of @xmath152 . over the two decades of temperature shown in the figure , the nrg data are described by an exponent @xmath244 . however , a fit over the range @xmath245 yields an improved estimate @xmath246 , consistent to within numerical error with the ct - qmc result for the anderson model . like the ct - qmc estimate for the anderson model , this value is consistent with the hypothesis that for @xmath209 spin fluctuations are primarily driven critical by the bosonic bath , and @xmath237 . ( color online ) static spin susceptibility @xmath233 vs temperature @xmath156 for @xmath107 and @xmath111 calculated within ct - qmc for the bfa model and using the nrg for the bfk model at the qcp . the curves run parallel over the temperature range shown and are fitted by consistent exponents.,height=211 ] the reciprocal of the correlation - length exponent extracted from the crossover in the nrg many - body spectrum is @xmath247 , again consistent with the value obtained for the anderson model . this value differs from that found in two other cases : in the metallic bfk model,@xcite @xmath248 , while in the pseudogap kondo model,@xcite @xmath249 . although the critical spin fluctuations ( and hence the exponent @xmath197 ) for @xmath215 seem to be dominated by the bosonic bath , the rg flow away from the critical point described by the exponent @xmath101 is clearly different from that in cases of pure - bosonic or pure - fermionic criticality . figure [ fig : binder8 ] shows the variation of the bfa - model binder cumulant for @xmath107 , @xmath112 as calculated using ct - qmc . the intersection of curves representing different temperatures places the qcp at @xmath250 . this critical value is larger than that ( @xmath251 ) for @xmath111 because the bosonic interaction between time segments falls off faster with increasing @xmath15 . figure [ fig : nu6 ] shows that the binder cumulant scales according to eq . with an excellent collapse of data taken at different temperatures with a fitted exponent @xmath252 . ( color online ) binder cumulant @xmath253 vs bosonic coupling @xmath47 for @xmath107 and @xmath112 at different inverse temperatures @xmath229 . there is a clear intersection of the curves at the critical coupling @xmath250.,width=211 ] ( color online ) data collapse of the binder cumulant for @xmath107 and @xmath112 . a correlation length exponent @xmath254 produces an excellent collapse in the vicinity of the critical point.,width=240 ] figure [ fig : stat8 ] illustrates the temperature dependence of the static local susceptibility in each phase and at the critical coupling . for @xmath157 , @xmath233 follows eq . with @xmath255 over the lowest decade of temperature for which data were obtained , a clear departure from the behavior @xmath237 seen above for @xmath215 . ( color online ) static spin susceptibility @xmath233 from ct - qmc vs temperature @xmath156 for @xmath107 and @xmath112 in the kondo - screened phase ( squares ) , at the critical coupling ( circles ) and in the localized phase ( triangles ) . at the critical coupling @xmath250 , @xmath233 diverges according to eq . with @xmath256.,height=211 ] in the bfk model , the convergence of the nrg many - body spectrum to the critical spectrum is fastest ( i.e. , the crossover scale @xmath140 is highest ) for @xmath190 , with @xmath257 . the value @xmath190 means that the impurity is entirely decoupled from the bosons , and the qcp must be of pure - fermionic character . indeed , the asymptotic low - energy many - body spectrum in the quantum - critical regime can be reproduced by taking every possible combination of ( i ) one state from the spectrum of free @xmath112 bosons , and ( ii ) one state from the critical spectrum of the @xmath107 pseudogap kondo model . we summarize this spectral decomposition in the shorthand ( bf critical ) = ( b free ) @xmath258 ( f critical ) and refer to it below simply as f - type criticality . figure [ fig : stat8nrg ] compares the temperature dependence of the critical static local susceptibilities of the bfa and bfk models . over the two decades of temperature shown in the figure , the nrg data for the bfk model are described by an exponent @xmath259 , identical to the ct - qmc value for the anderson model . however , since the qcp corresponds to @xmath63 , we know that the exponent must coincide exactly with that of the pseudogap kondo model:@xcite @xmath213 . this value clearly differs from the one @xmath260 found within the spin - boson - model@xcite and the metallic bfk model.@xcite ( color online ) static spin susceptibility @xmath233 vs temperature @xmath156 for @xmath107 and @xmath112 calculated within ct - qmc for the bfa model and using the nrg for the bfk model at the qcp . the curves run parallel over the temperature range shown and are described by consistent exponents.,height=211 ] similarly , we can be confident that the correlation - length exponent must be identical to that of the pseudogap kondo model , @xmath261 , a result that is consistent with the ct - qmc value @xmath252 quoted above . the preceding results suggest that in cases @xmath6 where @xmath262 , the physics at and near the qcp is determined primarily by fermionic fluctuations , and that critical properties should coincide with those of the pseudogap anderson and pseudogap kondo models . dddcdddd & & & critical + & & & spectrum & & & & + + 0.1 & 0.5 & & b & 0.499 & & 0.48 & 0.475 + & 0.7 & & b & 0.700 & 0.296 & 0.506 & 0.506 + & 0.75 & & b & & 0.254 & 0.493 & 0.493 + & 0.8 & 5.0(5 ) & b & 0.800 & 0.213 & 0.46 & 0.470 + & 0.85 & 3.58(3 ) & m & & 0.176 & 0.426 & 0.433 + & 0.9 & 2.8(3 ) & m & 0.900 & 0.142 & 0.352 & 0.376 + & & & & 0.989 & 0.060 & 0.093 & + + 0.2 & 0.5 & & b & 0.499 & 0.526 & 0.48 & 0.475 + & 0.6 & & b & & 0.394 & 0.508 & 0.509 + & 0.65 & 3.40(5 ) & m & & 0.347 & 0.504 & 0.512 + & 0.7 & 2.98(3 ) & m & & 0.313 & 0.479 & 0.506 + & 0.75 & 2.63(3 ) & m & & 0.282 & 0.443 & 0.493 + & 0.8 & 2.30(5 ) & m & 0.800 & 0.251 & 0.398 & 0.470 + & 0.9 & 1.4 & m & 0.900 & 0.189 & 0.265 & 0.376 + & & & & 0.948 & 0.160 & 0.161 & + + 0.3 & 0.5 & 2.5 & m & 0.500 & 0.582 & 0.427(4 ) & 0.475 + & 0.6 & 2.05(5 ) & m & 0.600 & 0.493 & 0.405(2 ) & 0.509 + & 0.7 & 1.65(5 ) & m & 0.700 & 0.422 & 0.356 & 0.506 + & 0.8 & 1.1 & m & 0.800 & 0.365 & 0.270 & 0.470 + & 0.9 & 0 & f & 0.862 & 0.355 & 0.194 & 0.376 + & & & & 0.862 & 0.355 & 0.194 & + + 0.4 & 0.5 & 1.15(5 ) & m & 0.500 & 0.930 & 0.269 & 0.475 + & 0.6 & 0.84(2 ) & m & 0.600 & 0.853 & 0.233 & 0.509 + & 0.7 & 0 & f & 0.688 & 0.914 & 0.171 & 0.506 + & 0.8 & 0 & f & 0.688 & 0.914 & 0.171 & 0.470 + & 0.9 & 0 & f & 0.688 & 0.914 & 0.171 & 0.376 + & & & & 0.688 & 0.914 & 0.171 & + in order to investigate more systematically the different types of quantum criticality exemplified in the cases @xmath215 and @xmath263 , we have studied the particle - hole - symmetric pseudogap bfk model for 23 different @xmath6 pairs spanning the ranges @xmath264 and @xmath265 . table [ tab : nrg_exponents ] summarizes the critical properties , one line per @xmath6 pair . on each line , an estimate of @xmath138the value of @xmath47 that produces the highest temperature @xmath140 of entry into the quantum - critical regime for @xmath266 , is followed by the assignment of the critical many - body spectrum to one of three categories ( f , b , or m ) described further below . the remaining columns of table [ tab : nrg_exponents ] list critical exponents : @xmath197 , @xmath219 , and @xmath267 [ defined in eqs . , , and , respectively ] are values calculated or inferred for the pair @xmath6 in question . for purposes of comparison , we also list @xmath268 , the reciprocal of the order - parameter exponent at a pure - bosonic qcp with the same @xmath15 , as determined within the metallic bfk model . the exponent @xmath197 has been calculated for all but a small number of @xmath6 pairs , and in all cases its value is consistent with eq . . the order - parameter exponent @xmath219 can generally be evaluated to higher accuracy than @xmath101 ( because the latter depends on @xmath160 values obtained by interpolation of data collected at a discrete set of temperature points ) , and has also been calculated for almost all @xmath6 pairs . by contrast , we have explicitly computed @xmath267 for only about half the pairs . in every case where it can be tested , spanning the full range of bath exponents @xmath269 and @xmath4 , hyperscaling [ eq . ] holds to within our estimated numerical uncertainty . for this reason , we can confidently apply eq . to predict the value of @xmath267 in cases where it has not been computed directly . finally , we note that wherever it has been calculated , @xmath220 [ defined in eq . , but not listed in table [ tab : nrg_exponents ] ] obeys eq . to high precision . based on an examination of the critical spectra and the exponents @xmath197 and @xmath219 ( and hence @xmath101 via hyperscaling ) , we are able to identify three distinct types of quantum criticality : @xmath270 fermionic ( f)the asymptotic low - energy critical spectrum exhibits su(2 ) spin symmetry and decomposes into a direct product of the spectrum of free bosons with bath exponent @xmath15 and the critical spectrum of the pseudogap kondo model with band exponent @xmath17 , i.e. , ( bf critical ) = ( b free ) @xmath258 ( f critical ) . all static critical exponents that have been calculated are identical to those of the pure - fermionic pseudogap anderson and kondo models with the same @xmath17 . @xmath270 bosonic ( b)the asymptotic low - energy critical spectrum exhibits su(2 ) spin symmetry and decomposes into a direct product of the critical spectrum of the spin - boson model with bath exponent @xmath15 and the strong - coupling spectrum@xcite of the pseudogap kondo model with band exponent @xmath17 , i.e. , ( bf critical ) = ( b critical ) @xmath258 ( f strong - coupling ) . all static static critical exponents that have been calculated in region b are identical to those of the spin - boson model and of the metallic ( @xmath12 ) bfa and bfk models with the same @xmath15 . however , despite the asymptotic decomposition of the low - energy spectrum , the single - particle spectral function has a non - fermi liquid form ( at least for @xmath12 ) , as discussed in sec . [ subsec : dynamics ] . this may be related to the fact that the fermionic strong - coupling spectrum is not only found at the kondo fixed point where the exchange couplings take renormalized values @xmath271 , but also is approached for @xmath272 , @xmath273 finite in the limit of energy scales much smaller than @xmath274 . we will return to this observation in sec . [ subsec : flows ] . @xmath270 mixed ( m)the critical spectrum exhibits broken su(2 ) spin symmetry and does not decompose into a direct product of bosonic and fermionic parts . the exponents satisfy @xmath275 but the order - parameter exponent lies between the values for the spin - boson model and the pseudogap kondo / anderson models , i.e. , @xmath276 . the three types of quantum criticality are clearly revealed in plots of @xmath197 and @xmath267 vs @xmath15 at fixed @xmath17 . as can be seen in fig . [ fig : x_vs_s ] , @xmath203 coincides with @xmath207 until the latter value exceeds its counterpart @xmath204 in the pure - fermionic pseudogap problem . the break in the slope of @xmath197 vs @xmath15 marks the transition from m - type to f - type criticality . [ fig : nu_vs_s ] shows that the bose - fermi correlation - length exponent equals the bosonic value @xmath277 for sufficiently small @xmath15 ( b - type criticality ) and equals the fermionic value @xmath278 over precisely the range of @xmath15 where @xmath279 ( f - type criticality ) , but between these regimes , @xmath225 takes values that differ from both @xmath277 and @xmath278 ( m - type criticality ) . ( color online ) static magnetic critical exponent @xmath197 vs bath exponent @xmath15 for @xmath280 , @xmath281 , and @xmath282 . the diagonal line represents the pure - bosonic exponent @xmath207 , while each horizontal line segment shows the pure - fermionic value @xmath204 . for each @xmath6 pair , the bose - fermi exponent satisfies @xmath283$].,height=211 ] ( color online ) reciprocal of the correlation length exponent @xmath267 vs bath exponent @xmath15 for the metallic case @xmath12 and for pseudogaps described by @xmath284@xmath282 . the @xmath12 exponents coincide with those of the corresponding spin - boson model and represent the values @xmath268 describing pure - bosonic criticality . each horizontal line segment shows a pure - fermionic value @xmath285 . for each @xmath6 , the bose - fermi exponent coincides with @xmath268 for @xmath286 and with @xmath285 for @xmath287 . for @xmath288 , @xmath289 lies in between the bosonic and fermionic values.,height=211 ] figure [ fig : r - s_plane ] summarizes the type of criticality found at different locations on the @xmath17@xmath15 plane , including points studied for the metallic case @xmath12 where the behavior is always of the b type . it is seen that each type of criticality ( f , b , or m ) occupies a contiguous region . all the results are consistent with there being a boundary @xmath290 between the f and m regions ( shown as a solid curve in fig . [ fig : r - s_plane ] ) . as argued at the beginning of this section , such a boundary arises from the assumption that the spin response at the bose - fermi critical point is dominated by the bath ( bosonic or fermionic ) that has the more singular dynamical spin fluctuations , corresponding to the smaller value of @xmath197 . the results in fig . [ fig : r - s_plane ] are also consistent with there being a boundary @xmath291 between the m and f regions ( the straight line in the figure ) . such a boundary marks the line of equality of the frequency exponents of the bare bosonic propagator and the fermionic particle - hole bubble,@xcite although the significance of this observation in the present context remains to be established . ( color online ) summary of the pairs of bath exponents @xmath6 studied in this work . results for @xmath12 describe a metallic conduction band , while those for @xmath292 correspond to pseudogapped problems . squares , triangles , and circles respectively correspond to quantum criticality of the f , b , and m types , as described in the text . filled symbols summarize nrg results for the bfk model while open symbols represent ct - qmc results for the bfa model . solid lines show the conjectured boundaries @xmath291 and @xmath293 between the different types of criticality.,height=211 ] we now turn to the finite - temperature dynamics of the impurity green s function and the local susceptibility calculated at the critical bosonic coupling @xmath157 . since the nrg is unreliable in the regime @xmath294 , we rely mainly on the ct - qmc for this part of the study , focusing once more on the type - m case @xmath107 , @xmath111 and on the type - f case @xmath107 , @xmath112 . figs . [ fig : g_vs_zeta ] and [ fig : chi_vs_zeta ] plot @xmath295 and @xmath296 , respectively , as functions of the combination @xmath297 , where @xmath298 renders the scaling function dimensionless . for both @xmath6 pairs , the critical correlation functions at temperatures well below the bare kondo temperature , @xmath299 , exhibit excellent scaling collapse over two decades of @xmath300 . the scaling collapse leads to the important conclusions that in the long - time limit @xmath301 , @xmath302 we obtain @xmath303 , @xmath304 for @xmath111 and @xmath303 , @xmath305 for @xmath112 . these exponents are consistent with the relations @xmath306 where @xmath203 is defined in eq . . in the zero - temperature limit , eq . and the first eq . give @xmath307 , reproducing an exact result.@xcite the quality of the scaling collapse in figs . [ fig : g_vs_zeta ] and [ fig : chi_vs_zeta ] , as well as the reproduction of the correct zero - temperature limit , provide significant evidence that our results have reached the asymptotic low - energy scaling regime . since @xmath308 and @xmath309 are both less then one , each correlator obeys @xmath5 scaling per the discussion of eq . . the observation that @xmath310 implies that vertex corrections can not be neglected , in line with the fully interacting nature of the qcp . ( color online ) single - particle green s function @xmath295 vs @xmath311 for @xmath107 , @xmath111 , @xmath312 ( lower data ) and for @xmath107 , @xmath112 , @xmath313 ( upper data , all @xmath314 values multiplied by 1.75 to avoid overlap with the @xmath111 data ) . one observes excellent collapse of the data for just under two decades of @xmath300 . temperature labels are shared between this figure and fig . [ fig : chi_vs_zeta].,width=211 ] ( color online ) local spin susceptibility @xmath315 vs @xmath297 for @xmath107 , @xmath111 , @xmath312 ( lower data ) and for @xmath107 , @xmath112 , @xmath313 ( upper data , all @xmath316 values multiplied by 1.25 to avoid overlap with the @xmath111 data ) . one observes excellent collapse of the data for over two decades of @xmath300 . temperature labels are shared between this figure and fig . [ fig : g_vs_zeta].,width=211 ] for reasons discussed in sec . [ sec : summary ] , we have not been able to access low enough temperatures using ct - qmc to study cases of b - type quantum criticality . in this region of the @xmath17-@xmath15 plane , we must rely on information from previous nrg studies of the @xmath12 bfk and bfa models,@xcite which have demonstrated that at the critical coupling , the local susceptibility calculated for @xmath317 is consistent with the existence of @xmath5 scaling , while the zero - temperature impurity spectral function shows a non - fermi liquid form . the latter is also seen for @xmath12 results in a dynamical large-@xmath133 limit,@xcite as a function of both frequency and temperature . it is reasonable to attribute this non - fermi - liquid behavior to the existence of an rg - irrelevant coupling between the fermionic and bosonic sectors of the hilbert space . schematic rg flow diagrams for the pseudogap bfk model projected onto the plane spanned by the impurity - boson coupling @xmath47 and the spin - flip exchange scattering @xmath274 , valid for bath exponents @xmath6 such that the quantum criticality is of the ( a ) b or m , and ( b ) f types . not shown in these diagrams is a third axis describing the longitudinal exchange coupling @xmath318 . the dashed line marks the boundary between the kondo and localized phases . arrows show the direction of rg flow of the effective couplings between fixed points represented by circles : free moment ( fm , at @xmath319 ) , kondo ( k , at @xmath272 ) , localized ( l , a line of fixed points spanning @xmath320 ) , fermionic critical ( fc , at @xmath321 ) , and bose - fermi critical ( bfc , at @xmath272 in region b but at finite @xmath322 in region m ) . very similar rg flow diagrams can be constructed for the pseudogap bfa model.,title="fig:",width=158 ] schematic rg flow diagrams for the pseudogap bfk model projected onto the plane spanned by the impurity - boson coupling @xmath47 and the spin - flip exchange scattering @xmath274 , valid for bath exponents @xmath6 such that the quantum criticality is of the ( a ) b or m , and ( b ) f types . not shown in these diagrams is a third axis describing the longitudinal exchange coupling @xmath318 . the dashed line marks the boundary between the kondo and localized phases . arrows show the direction of rg flow of the effective couplings between fixed points represented by circles : free moment ( fm , at @xmath319 ) , kondo ( k , at @xmath272 ) , localized ( l , a line of fixed points spanning @xmath320 ) , fermionic critical ( fc , at @xmath321 ) , and bose - fermi critical ( bfc , at @xmath272 in region b but at finite @xmath322 in region m ) . very similar rg flow diagrams can be constructed for the pseudogap bfa model.,title="fig:",width=158 ] we conclude this section with a discussion of the rg structure of the ising - anisotropic pseudogap bfk model in the different regions of the @xmath17-@xmath15 plane . it is important to recognize that the bare couplings @xmath323 entering eq . are insufficient to describe this rg structure . since a coupling @xmath179 breaks spin - rotation symmetry , su(2 ) symmetry of the kondo exchange need not be preserved under renormalization , and a minimal description involves keeping track of a triad of effective couplings @xmath324 . the stable fixed points of the pseudogap bfk model are the kondo fixed point at @xmath325 , @xmath326 and a line of localized fixed points at @xmath327 , @xmath328 , and ( at least for @xmath12 ; see ref . ) @xmath329 where @xmath47 is the bare bosonic coupling and @xmath219 is the order - parameter exponent . the qcp of the pseudogap kondo model is located at @xmath330 and @xmath331 , where @xmath332 for @xmath333 . a perturbative rg treatment of the model valid for @xmath12 and @xmath334 has shown@xcite that the bose - fermi qcp is located at @xmath335 , @xmath336 , and @xmath337 , with the nrg having confirmed the result @xmath338 over a wider range of @xmath15 values.@xcite what remains to be established is the manner in which @xmath339 evolves with increasing @xmath17 and/or @xmath15 to reach @xmath340 upon entry into the region f of the @xmath17-@xmath15 plane . since region f is bounded by the line @xmath341 , and the exponent @xmath204 is very well described@xcite by @xmath342 ^ 2=1-[\rho_0 j_{\perp}^*]^2 $ ] , one is led to conclude that for a given value of @xmath15 satisfying @xmath334 , the relation @xmath343 holds true both at @xmath12 ( on the left - most edge of region b in fig . [ fig : r - s_plane ] ) and at the border between regions m and f. this observation , when combined with the asympototic spectral decomposition ( bf critical ) = ( b critical ) @xmath258 ( f strong - coupling ) that holds throughout region b , naturally leads to the conjecture that upon increasing @xmath17 from 0 to the point of entry into region f , ( 1 ) @xmath344 remains constant ( or very nearly so ) ; ( 2 ) @xmath345 decreases monotonically from a value of order unity to reach zero ; and ( 3 ) @xmath346 is infinite throughout region b , and decreases monotonically to @xmath347 on crossing region m. the presence in region m of finite exchange couplings satisfying @xmath348 is consistent with the observation of broken su(2 ) spin symmetry in the asymptotic low - energy critical spectrum . by contrast , the value @xmath338 in region b ensures that the spectrum appears to be su(2 ) invariant at energy scales much below @xmath344 . support for this picture comes from table [ tab : nrg_exponents ] , which shows a clear trend with increasing @xmath17 at fixed @xmath15 ( or with increasing @xmath15 at fixed @xmath17 ) in the value @xmath138 of the bosonic coupling that brings the model into its quantum - critical regime at the highest temperature . throughout region b in fig . [ fig : r - s_plane ] , @xmath138 and the corresponding kondo coupling @xmath349 are very large in order to achieve rapid flow to @xmath338 and the large fixed - point value of @xmath345 . while region m is crossed , @xmath138 and @xmath350 decrease in line with @xmath351 and @xmath346 . in region f , the value @xmath190 shows that the critical point can be reached without any coupling of the impurity to the bosons , meaning that @xmath351 is necessarily zero and that @xmath352 . based on the preceding considerations we propose the schematic rg flow diagrams shown projected onto the @xmath47-@xmath274 plane in fig.[fig : rg - flows ] . in regions b and m , a qcp labeled bfc in fig.[fig : rg - flows](a ) lies on the separatrix between the basins of attraction of the kondo and localized fixed points ( k and l , respectively ) . rg flow along the separatrix is toward bfc , and on the small-@xmath47 side , away from the fermionic pseudogap critical point fc . with increasing @xmath17 at fixed @xmath15 ( or increasing @xmath15 at fixed @xmath17 ) , bfc moves to smaller values of @xmath47 , and merges with fc at the boundary between regions m and f. throughout the latter region of the @xmath17-@xmath15 plane , the rg structure is as shown in fig . [ fig : rg - flows](b ) , with flow along the separatrix toward fc . to this point , we have focused exclusively on conditions of strict particle - hole ( @xmath353-@xmath354 ) symmetry , i.e. , @xmath143 for the bfa model , which maps to a bfk model with potential scattering @xmath56 . in this section , we consider the effects of breaking this symmmetry , supporting our arguments with nrg results for the critical spectrum and for the critical exponent @xmath197 entering eq . , as obtained in a preliminary study of the bfk model with @xmath355 . under the conditions of @xmath353-@xmath354 symmetry defined in the previous paragraph , the pseudogap kondo and anderson models exhibit a kondo - destruction qcp only over the range of band exponents @xmath3 ; for @xmath356 , the kondo phase disappears and over the entire parameter space the system approaches the fm fixed point at low temperatures . in these models , it is known that away from @xmath353-@xmath354 symmetry , a strong - coupling or kondo phase is present for all @xmath292 , and that for @xmath357 this phase is separated from the free - moment phase by an interacting qcp . for exponents @xmath358 , @xmath353-@xmath354 asymmetry is irrelevant at the qcp and the quantum criticality is identical to that at @xmath353-@xmath354 symmetry , whereas for @xmath359 , quantum criticality away from @xmath353-@xmath354 symmetry is governed by a distinct asymmetric qcp.@xcite combining these well - established properties of the pure - fermionic pseudogap models with the results presented in sec . [ sec : critical ] allows informed speculation about the effects of @xmath353-@xmath354 asymmetry in the pseudogap bfa and bfk models . we expect the persistence of a region of b type criticality in which the critical spectrum has the product form ( bf critical ) = ( b critical ) @xmath258 ( f strong - coupling ) , with @xmath353-@xmath354 asymmetry affecting the fermionic strong - coupling spectrum but not the critical properties studied in this work , all of which are determined solely by the bosonic spectrum . for @xmath12 , @xmath353-@xmath354 asymmetry should be marginal ( as it is in the metallic anderson and kondo models ) , giving rise to a line of qcps sharing the same critical properties . for @xmath292 , by contrast , @xmath353-@xmath354 asymmetry should be relevant , with the fermionic spectrum being that of the asymmetric strong - coupling fixed point . these conjectures are consistent with preliminary nrg studies of the cases @xmath360 and @xmath361 . in the presence of @xmath353-@xmath354 asymmetry , we also expect a region in which the dynamical spin response arising from the fermions is more singular than that from the bosonic bath . this region of what we will dub f@xmath362-type criticality is likely to span the range of exponents @xmath357 and @xmath363 in which @xmath364 . here , @xmath365 is the value of the exponent @xmath197 in eq . at the asymmetric pseudogap qcp . for @xmath366 , @xmath367 ( ref . ) so f@xmath362 criticality should be identical to f - type . for @xmath368 , by contrast , @xmath369 ( ref . ) , and the f@xmath362 critical exponents should belong to the universality class of the asymmetric pseudogap qcp . we have found an example of f@xmath362 criticality for @xmath370 . finally , it seems probable that the @xmath123 and @xmath371 regions will be separated by one in which the critical spectrum does not have a simple direct product form . at @xmath353-@xmath354 symmetry , the m - type region covers the range @xmath288 . as discussed in sec . [ subsubsec : other - r , s ] , the lower bound on the range of @xmath15 seems to be defined by the equality of the frequency exponents of the bare bosonic propgator and the fermionic particle - hole bubble . since @xmath353-@xmath354 asymmetry is irrelevant at the fm fixed point,@xcite there seems to be no reason to expect the boundary to be affected by this breaking of symmetry . on the other hand , the upper bound @xmath290 seems likely to be replaced by @xmath372 , the condition discussed in the previous paragraph for entry into a region of f@xmath362 criticality . in this work , we have applied a combination of continuous - time quantum monte - carlo ( ct - qmc ) and numerical renormalization - group ( nrg ) methods to study systematically the interplay between bosonic and fermionic baths , each of which on its own can induce critical kondo destruction at a continuous zero - temperature transition . we have shown that at particle - hole symmetry , the quantum critical point ( qcp ) in the easy - axis pseudogap bfa model belongs to the same universality class as the qcp of the corresponding kondo model . we have further shown the surprising result that the value of the exponent @xmath197 for the temperature dependence of the critical local spin susceptibility [ defined in eq . ] of either model is sensitive to the exponents @xmath17 and @xmath15 characterizing the vanishing of the fermionic and bosonic densities of states . in the region of the @xmath17-@xmath15 plane where @xmath287 ( @xmath373 being the thermal critical exponent of the pseudogap anderson and kondo models without bosons ) , all critical exponents of the bose - fermi models that we have calculated are identical to those of the pure - fermionic models , and the critical many - body spectrum decomposes into a direct product of a free bosonic spectrum and a critical pseudogap fermionic spectrum ; this regime has eluded all previous studies . for @xmath374 , the critical spin fluctuations are instead dominated by the bosonic bath , leading to @xmath275 . however , the correlation - length exponent @xmath101 [ defined in eq . ] coincides with that of the spin - boson model , and the asymptotic low - energy critical spectrum decomposes into a direct product of a critical bosonic spectrum and a pseudogap kondo fermionic spectrum , only for @xmath375 . within an intermediate region @xmath288 , @xmath225 takes a value lying between those for the spin - boson model and for the pseudogap anderson and kondo models , and bosonic and fermionic degrees of freedom can not be disentangled in the critical spectrum . in all three regions , other static critical exponents are related to @xmath197 and @xmath101 via hyperscaling relations that are expected to hold only at an interacting critical point . we have also shown that at the qcp , the imaginary - time correlation functions @xmath376 and @xmath377 scale as functions of @xmath378 and that their real - frequency counterparts obey @xmath5 scaling , consistent with the notion that the qcp is fully interacting . scaling collapse of imaginary - time correlators as functions of @xmath300 has previously been reported for the sub - ohmic bfk model,@xcite where the kondo effect is critically destroyed by the bosonic bath , and in the pseudogap anderson model at and away from particle - hole symmetry,@xcite where criticality is driven by fermionic fluctuations of the band . that it generalizes to the more complex case considered in the present work suggests that the scaling collapse may very well be a general feature of local quantum criticality . the scaling collapse in terms of @xmath300 implies ( under conditions spelled out in sec.[subsec : dynamics ] ) that the associated real - frequency correlator displays @xmath5 scaling , a property that has been reported in several experiments on unconventional quantum criticality in @xmath379-electron based magnets.@xcite a scaling collapse of the form observed here is natural for boundary - conformal quantum impurity systems . however , conformal symmetry is broken both by the bosonic bath ( as in the sub - ohmic bfk model ) , and the fermionic bath ( as in the pseudogap anderson model ) . symmetry restoration frequently accompanies criticality . the case discussed here , however , has to be distinguished from this more standard situation of irrelevant symmetry - breaking fields , as the broken symmetry in the bulk induces boundary criticality with local correlators that are compatible with a boundary conformal critical theory . as discussed earlier,@xcite a deeper understanding of this observation should help identify a critical field theory of unconventional quantum criticality . from a methodological viewpoint , this work has shown that the ct - qmc method can attain sufficiently low temperatures in the presence of a bosonic bath to identify a quantum critical point lying between stable phases , and to obtain critical properties in agreement with those given by the nrg . such study is possible using the ct - qmc only in cases where entry into the critical power - law regime takes place at a fairly high temperature @xmath140 . the presence of a pseudogap in the fermionic density of states helps in this regard : the value of @xmath140 decreases as @xmath17 decreases , so larger @xmath17 values are optimal . decreasing the bath exponent @xmath15 also reduces @xmath140 , making it difficult to study the range @xmath380 without fine - tuning of the hamiltonian . we thank m. t. glossop , p. ribeiro , m. vojta , f. zamani , and j .- x . zhu for useful discussions . this work has been supported by nsf grants no . dmr-0710540 , dmr-1309531 , and dmr-1107814 , the robert a. welch foundation grant no . c-1411 , and the alexander von humboldt foundation . we acknowledge the hospitality of the max planck institute for the physics of complex systems ( k.i . and j.h.p . ) , the aspen center for physics ( nsf grant no . 1066293 ; and q.s . ) , the institute of physics of chinese academy of sciences , and karlsruhe institute of technology ( q.s . ) . in problems that exhibit strict particle - hole symmetry , the u(1 ) symmetry that leads to conservation of total charge ( electron number measured from half - filling ) @xmath383 is promoted to an su(2 ) isospin symmetry generated by @xmath384 and @xmath385 . the nrg calculations were performed on a basis of multiplets labeled by quantum numbers @xmath386 ( the @xmath8 component of the total spin ) and @xmath387 , with each multiplet representing @xmath388 degenerate eigenstates .	we address the phenomenon of critical kondo destruction in pseudogap bose - fermi anderson and kondo quantum impurity models . these models describe a localized level coupled both to a fermionic bath having a density of states that vanishes like @xmath0 at the fermi energy ( @xmath1 ) and , via one component of the impurity spin , to a bosonic bath having a sub - ohmic spectral density proportional to @xmath2 . each bath is capable by itself of suppressing the kondo effect at a continuous quantum phase transition . we study the interplay between these two mechanisms for kondo destruction using continuous - time quantum monte carlo for the pseudogap bose - fermi anderson model with @xmath3 and @xmath4 , and applying the numerical renormalization - group to the corresponding kondo model . at particle - hole symmetry , the models exhibit a quantum critical point between a kondo ( fermionic strong - coupling ) phase and a localized ( kondo - destroyed ) phase . the two solution methods , which are in good agreement in their domain of overlap , provide access to the many - body spectrum , as well as to correlation functions including , in particular , the single - particle green s function and the static and dynamical local spin susceptibilities . the quantum - critical regime exhibits the hyperscaling of critical exponents and @xmath5 scaling in the dynamics that characterize an interacting critical point . the @xmath6 plane can be divided into three regions : one each in which the calculated critical properties are dominated by the bosonic bath alone or by the fermionic bath alone , and between these two regions , a third in which the bosonic bath governs the critical spin response but both baths influence the renormalization - group flow near the quantum critical point .
the total inclusive , semileptonic @xmath2 decay rate admits a clean theoretical analysis based on the heavy - quark expansion . the optical theorem is used to relate the decay rate to the forward matrix element of the product of two hadronic currents , which is evaluated using the operator product expansion @xcite . the total rate is presently known including corrections of order @xmath6 @xcite and @xmath7 ( with @xmath8 a typical hadronic scale ) . an accurate knowledge of the total rate does not immediately translate into a precise determination of @xmath0 , since experimentally one has to impose cuts on the kinematical variables in order to discriminate the @xmath9 transitions against the much larger @xmath10 background . the first inclusive determination of @xmath0 was based on a measurement of the charged - lepton energy spectrum in the endpoint region , which is kinematically forbidden for decays with a charm meson in the final state . this cut eliminates about 90% of the signal . it is difficult to reliably calculate the remaining event fraction , since the operator product expansion breaks down in this part of the phase space and must be replaced by a twist expansion @xcite . recently , it was shown that the situation is improved when a cut @xmath11 with @xmath12 is applied to the dilepton invariant mass @xcite . this keeps about 20% of the signal events , and the fraction of the remaining events can be calculated in a model - independent way using the operator product expansion . after the cut , the typical momentum flow @xmath13 through the hadronic tensor is of order the charm - quark mass or less , so that the problem at hand involves the three scales @xmath14 . in the standard heavy - quark expansion , the charm and the @xmath4-quark masses are treated on the same footing , but it is known that the convergence of this expansion at the charm scale can be poor . in this letter we follow the strategy proposed in @xcite and perform a two - step expansion ( the `` hybrid expansion '' ) in the ratios @xmath15 and @xmath16 . the first step is to integrate out the degrees of freedom above the scale @xmath13 and to sum up logarithms of the form @xmath17 using the renormalization group ( rg ) . this is achieved by expanding the weak currents and the qcd lagrangian to order @xmath18 in the heavy - quark effective theory ( hqet ) @xcite , and calculating the wilson coefficients of the pertinent operators at next - to - leading order ( nlo ) in rg - improved perturbation theory . whereas the wilson coefficients of the leading - order currents @xcite and of the @xmath19 corrections to the hqet lagrangian @xcite were known for a long time , the nlo calculation of the coefficients entering the expansion of the currents at order @xmath18 was completed only recently @xcite . in the present work , we perform the second step in the hybrid expansion and construct the operator product expansion in the resulting low - energy effective theory . we calculate the decay rate @xmath20 at subleading order in the ratio @xmath15 and to nlo in rg - improved perturbation theory . we also resum leading logarithmic corrections to the power corrections of order @xmath21 . in contrast with previous analyses @xcite we calculate the decay rate directly , rather than calculating the fraction of events that passes the @xmath22-cut and then multiplying this fraction with an independent result for the total semileptonic rate . since the characteristic scale of the total rate is significantly higher than ( and parametrically different from ) the typical hadronic scale @xmath13 after the @xmath22-cut , the theoretical error can not be reduced by separating the two calculations . based on a detailed analysis of the residual renormalization - scale dependence and the sensitivity to the definition of the @xmath4-quark mass , we conclude that the @xmath2 branching ratio after the dilepton invariant - mass cut can be calculated reliably , allowing for a determination of @xmath0 at the 10% level . a cut @xmath11 on the dilepton invariant mass squared @xmath23 implies @xmath24 for the hadronic invariant mass @xmath25 and energy @xmath26 . for @xmath27 both variables vary between @xmath28 and @xmath29 , and thus no charmed hadrons are allowed in the final state . if the cutoff is chosen higher ( as may be required for experimental reasons ) , the maximal values of @xmath25 and @xmath26 become less than the @xmath30-meson mass . the kinematical variables that enter the partonic calculation of the decay rate are the parton invariant mass and energy , @xmath31 and @xmath32 , where @xmath33 with @xmath34 being the velocity of the @xmath1-meson . it is convenient to work with the dimensionless quantities @xmath35 and @xmath36 , the phase space for which is @xmath37 while @xmath38 is of order unity , independent of the value of @xmath39 , the quantity @xmath40 is parametrically suppressed by @xmath41 with the optimal value @xmath42 the characteristic scale of the parton momenta is @xmath43 . the strong - interaction dynamics of the inclusive decay @xmath2 is encoded in the imaginary part of the hadronic tensor @xmath44 defined as @xmath45 where the scalar functions @xmath46 depend on @xmath47 and @xmath32 , or equivalently , on @xmath38 and @xmath40 . the weak current @xmath48 mediates the @xmath9 transition . to separate the long- and short - distance physics one usually performs the operator product expansion of the hadronic tensor and calculates the corresponding wilson coefficient functions in perturbation theory . however , since in the present case the characteristic scale @xmath13 of the momentum flow through the operator product is lower than the @xmath4-quark mass , the perturbative expansion of the wilson coefficients involves logarithms of the ratio @xmath15 . to improve the quality of the perturbative results these logarithms should be summed . this can be achieved by absorbing the physics above the scale @xmath13 into the couplings of a low - energy effective theory . the relevant effective theory is the hqet , and the expansion of the weak current at nlo reads @xcite @xmath49 the two operators contributing at leading order are @xmath50 where @xmath51 , and @xmath52 is the effective heavy - quark spinor of the hqet . at nlo in @xmath18 , the operator basis contains ten independent operators . six of these are local operators involving a gauge - covariant derivative on one of the quark fields : @xmath53 the remaining four involve time - ordered products of the leading - order currents with the operators @xmath54 and @xmath55 appearing at order @xmath18 in the hqet lagrangian : @xmath56 these nonlocal operators are present because we work with eigenstates of the lowest - order hqet lagrangian , so that all dependence on the heavy - quark mass is explicit . it is convenient to treat the operators @xmath57 on the same footing as the local operators @xmath58 , because the nonlocal operators mix into the local ones under renormalization . we have recently completed the calculation of the two - loop anomalous dimension matrix for the operators @xmath59 and solved the rg equations for their coefficients @xmath60 @xcite . when the expansion ( [ jexp ] ) is inserted into the operator product in ( [ tprod ] ) , all dependence on the @xmath4-quark mass becomes explicit and is contained in the wilson coefficients @xmath61 , @xmath62 and the @xmath18 prefactor of the higher - dimension operators . in the next step , we perform the operator product expansion and obtain forward @xmath1-meson matrix elements of local operators of the form @xmath63 in the low - energy theory . to order @xmath21 , these matrix elements can be expressed in terms of the parameters @xcite @xmath64 the anomalous dimension of the kinetic operator vanishes because of reparameterization invariance @xcite . up to higher orders in the heavy - quark expansion , @xmath65 is determined from the mass splitting @xmath66gev@xmath67 . by dimensional analysis , the contributions from time - ordered products of higher - dimension operators in ( [ jexp ] ) are suppressed by powers of @xmath68 . at nlo in @xmath69 , it suffices to evaluate time - ordered products with operator insertions of the type @xmath70 or @xmath71 . the decay rate is then obtained by contracting the result for the hadronic tensor with the lepton tensor and integrating over the portion of phase space with @xmath11 : @xmath72 where @xmath73 from ( [ eq : zmax ] ) . using the fact that the hadronic tensor is an analytic function of @xmath38 apart from discontinuities on the positive real axis , we have written the expression for the rate @xmath74 as a contour integral in the complex @xmath38-plane . the integrand of the contour integral is @xcite @xmath75\,t_1 + 6(4 - 4z+z^2\xi)\,t_2 \nonumber\\ & & \hspace{3.1cm}\mbox{}+ 2 z^2(1-\xi)(t_3 + 2 m_b\,t_4 + m_b^2\,t_5 ) \bigg\ } \,,\end{aligned}\ ] ] where @xmath76 is the tree - level decay rate . + at nlo in rg - improved perturbation theory , the matrix elements of the hadronic tensor must be evaluated including corrections of order @xmath77 in the low - energy theory . these are obtained by evaluating the contributions of all physical cuts of the diagrams shown in figure [ fig : graphs ] . these diagrams contain infrared divergences for @xmath78 from soft and collinear gluons , which cancel once the differential rate is integrated over phase space , because the rate @xmath74 is an infrared - safe quantity . the contour - integral representation in ( [ eq : phaseint ] ) avoids the singularity at @xmath79 . after integrating over the loop momentum , the function @xmath80 is given by an integral over feynman parameters . at this point we exchange the order of integration and integrate over the phase space first . the @xmath40-integration is trivial , because the dependence of the loop integrals on @xmath40 is fixed by their mass dimension . note that the upper limit @xmath81 of the integration depends on @xmath38 . after expanding to nlo in @xmath69 , the @xmath38-integrals can all be reduced to @xmath82 where the denominator @xmath83 with @xmath84 is the result of the feynman parameterization . some of the diagrams have a pole at @xmath79 due to the propagation of a single @xmath85-quark . the integral @xmath86 evaluates to @xmath87 the second integral picks up contributions for @xmath88 and @xmath89 . it is needed only for @xmath90 and @xmath91 ( with @xmath92 the dimensional regulator ) . we find @xmath93 - 2\epsilon \ln(1+\sqrt{1-a } ) + o(\epsilon^2 ) \bigg\ } \nonumber\\ & & \mbox{}+ \frac{\theta(a-1)}{a^{1+\epsilon } } \,.\end{aligned}\ ] ] we have not expanded the result for @xmath89 , since we will have to integrate it from @xmath94 to @xmath95 when performing the integrals over feynman parameters . the corresponding expression for the function @xmath96 can be obtained by integrating the above result with respect to @xmath97 . we now present the results of our calculation . we expand the rate @xmath74 in powers of @xmath69 and define @xmath98 where the dimensionless expansion coefficients @xmath99 collect terms of different order in the hybrid expansion . the leading - order contribution @xmath100 was calculated in @xcite ; the nlo term @xmath101 is the main result of the present work . we find that @xmath102 \ , , \nonumber\\ \widehat\gamma_4 & = & \sum_{i=1,2}\,c_1(\mu)\,c_i(\mu)\,f_4[c_1,c_i ] + \sum_{j=4,6,7,9}\,c_1(\mu)\,b_j(\mu)\,f_4[c_1,b_j ] \,,\end{aligned}\ ] ] where ( @xmath103 is a color factor ) @xmath104 & = & 1 + \frac{c_f\alpha_s(\mu)}{4\pi } \left ( 6\ln\frac{\mu}{2\mu_c } + 19 - \frac{8\pi^2}{3 } \right ) - \frac32\,\frac{\lambda_2(\mu)}{\mu_c^2 } + \dots \ , , \nonumber\\ f_4[c_1,c_1 ] & = & - \bigg\ { 1 + \frac{c_f\alpha_s(\mu)}{4\pi } \left ( 6\ln\frac{\mu}{2\mu_c } + \frac{29}{2 } - \frac{8\pi^2}{3 } \right ) \bigg\ } + \frac32\,\frac{\lambda_1 + 3\lambda_2(\mu)}{\mu_c^2 } + \dots \ , , \nonumber\\ f_4[c_1,c_2 ] & = & \frac12 + \dots \ , , \nonumber\\ f_4[c_1,b_4 ] & = & \frac34\,\bigg\ { 1 + \frac{c_f\alpha_s(\mu)}{4\pi } \left ( 6\ln\frac{\mu}{2\mu_c } + \frac{37}{2 } - \frac{8\pi^2}{3 } \right ) \bigg\ } - \frac34\,\frac{\lambda_2(\mu)}{\mu_c^2 } + \dots \ , , \nonumber\\ f_4[c_1,b_6 ] & = & \frac14\,\bigg\ { 1 + \frac{c_f\alpha_s(\mu)}{4\pi } \left ( 6\ln\frac{\mu}{2\mu_c } + \frac{39}{2 } - \frac{8\pi^2}{3 } \right ) \bigg\ } + \frac34\,\frac{\lambda_2(\mu)}{\mu_c^2 } + \dots \ , , \nonumber\\ f_4[c_1,b_7 ] & = & \frac{c_f\alpha_s(\mu)}{4\pi } \left ( 6\ln\frac{\mu}{2\mu_c } + \frac92 \right ) - \frac32\,\frac{\lambda_1}{\mu_c^2 } + \dots \ , , \nonumber\\ f_4[c_1,b_9 ] & = & \frac13\,\frac{c_f\alpha_s(\mu)}{4\pi } \left ( 6\ln\frac{\mu}{2\mu_c } + \frac{25}{2 } \right ) - \frac92\,\frac{\lambda_2(\mu)}{\mu_c^2 } + \dots \,.\end{aligned}\ ] ] note that the wilson coefficient @xmath105 starts at nlo in @xmath77 , so there is no need to include @xmath106 corrections to the quantity @xmath107 $ ] . the coefficients @xmath108 and @xmath109 are determined in terms of products of coefficients appearing in the leading - order expansion of the currents and in the hqet lagrangian . @xmath110 is the coefficient of the chromo - magnetic operator @xmath111 , and @xmath112 is the coefficient of the kinetic operator @xmath113 . the remaining coefficients , @xmath114 and @xmath115 , were calculated at nlo in @xcite . the wilson coefficients , as well as the matrix elements in the low - energy theory , depend on the choice of the renormalization scheme . the scheme dependence cancels in the result for the decay rate @xmath74 . in @xcite we have considered a class of renormalization schemes parameterized by the quantity @xmath116 . our results ( [ fres ] ) refer to the @xmath117 renormalization scheme ( with anticommuting @xmath118 ) . the corresponding results in a different scheme rs are obtained by replacing @xmath119 that way , it can be checked that the scheme - dependent terms proportional to @xmath120 indeed cancel in the final result . the above expressions are given in terms of the @xmath4-quark pole mass @xmath121 , which is the expansion parameter that enters in the hqet . however , this mass parameter is plagued by infrared renormalon ambiguities , and it is better to eliminate it from the final result for the decay rate . it is well known that the convergence of the perturbative series of inclusive decay rates can be improved by replacing the pole mass in favor of a low - scale subtracted quark mass , which is obtained from the pole mass by removing a long - distance contribution proportional to a subtraction scale @xmath122 @xcite . in addition to @xmath123 , the characteristic scale @xmath124 of the hybrid expansion and the associated parameter @xmath68 depend on the definition of the quark mass . it has been shown in @xcite that one should choose @xmath125 with a parameter @xmath126 so as not to upset the power - counting in the hybrid expansion . specifically , we use the potential - subtracted ( ps ) mass @xmath127 introduced by beneke @xcite . at nlo , the relation between the pole mass and the ps mass reads @xmath128 + \dots \right\ } , \ ] ] where @xmath129 is the first coefficient of the @xmath130 function . with @xmath125 it follows that @xmath131 .\ ] ] from now on , we use the ps mass @xmath132 in all our equations unless indicated otherwise . the parameters @xmath13 and @xmath69 are defined with respect to this choice . this results in extra terms in the perturbative expansion proportional to @xmath133 . the results in the pole scheme can be recovered by setting @xmath134 . in @xcite , the authors have eliminated the pole mass in favor of the so - called upsilon mass @xmath135 @xcite , which up to a small nonperturbative contribution is one half of the mass of the @xmath136 resonance . the relation between the upsilon mass and the pole mass is @xmath137 + \dots \right\ } .\ ] ] in higher orders in this expansion the logarithms @xmath138 $ ] appear in the form @xmath139\sim\exp(\ell)$ ] , which yields a factor @xmath140 $ ] . hence , schematically the relation between the masses becomes @xmath141 \,,\ ] ] which shows that despite of the unusual form of ( [ 1smass ] ) the upsilon mass is a low - scale subtracted mass in the sense described earlier . ( one may , however , question whether this argument can consistently be applied in low - order calculations . ) based on this observation , the authors of @xcite have suggested to treat the @xmath142 term on the right - hand side of ( [ 1smass ] ) as an @xmath106 correction when the upsilon mass is used in the context of perturbative expansions in the @xmath1 system . therefore , in our case we should replace @xmath143 $ ] and treat the @xmath142 term as a term of order @xmath144 , as in ( [ psfirst ] ) . this means that our formulae apply without modification to the upsilon scheme , provided we set @xmath145 and consider this to be a parameter of order unity . inserting the explicit expressions for the wilson coefficients and matrix elements into ( [ g3g4 ] ) , and evaluating the results for @xmath146 colors and @xmath147 light quark flavors , we obtain @xmath148 \nonumber\\ & & \mbox{}+ \ln x\,\left [ \frac{12}{25 } + k_7\,\frac{\alpha_s(m_b)}{\pi } + \left ( \frac{24}{25}\,\ln\frac{\mu}{2\mu_c } + \frac{64}{25}\,\zeta + k_8 \right ) \frac{\alpha_s(\mu)}{\pi } \right ] \nonumber\\ & & \mbox{}+ \left ( \frac{37}{18 } - \frac{12}{25}\,\ln x + \frac{10}{3}\,x^{-6/25 } - \frac{97}{18}\,x^{-9/25 } \right ) x^{9/25}\,\frac{\lambda_2(m_b)}{\mu_c^2 } + \dots \,,\end{aligned}\ ] ] where @xmath149 , and we have used that @xmath150 . we have extracted an overall factor @xmath151 from the expressions for the quantities @xmath99 , which reflects the leading - order anomalous dimensions of the heavy light currents in the time - ordered product in ( [ tprod ] ) . the coefficients @xmath152 summarizing the nlo corrections are @xmath153 since we have solved the rg equation for the wilson coefficients to obtain their running from the @xmath4-quark mass down to the scale @xmath154 , all logarithms of the ratio @xmath155 are resummed into the running couplings @xmath156 and @xmath157 . note that the expressions ( [ master ] ) are scale- and scheme - independent at nlo . as a check of our result , we expand @xmath156 in terms of @xmath157 in order to recover the known one - loop expression for the decay rate . this yields @xmath158 in agreement with the findings of @xcite . the numerical effects of the nlo resummation of logarithms achieved in @xcite and in the present work are rather significant , despite the fact that there is limited phase space for operator evolution between the @xmath4-quark mass and the scale @xmath159gev . in the pole scheme ( i.e. , with @xmath134 ) and with @xmath27 , the running from @xmath160 down to a characteristic scale @xmath161 decreases the value of @xmath100 by a factor 0.47 , and it changes the value of @xmath101 by a factor @xmath162 . the resummation effects are less pronounced in the ps scheme with @xmath163 , where we find that the value of @xmath100 increases by a factor 1.10 , whereas the value of @xmath101 decreases by a factor 0.66 . finally , we quote the expression for the remainder @xmath164 in ( [ hybrid ] ) , which contains all terms suppressed by at least five powers of @xmath69 , and for which we do not perform a rg improvement . as we will see below these contributions are numerically very small , so that a leading - order treatment is indeed justified . we obtain @xmath165 where the scale @xmath166 is undetermined and will be chosen between @xmath167 and @xmath123 in our analysis below . the function @xmath168 can be obtained from the results of @xcite and is given by @xmath169 \nonumber\\ & & \mbox{}- \frac{1}{6\delta^4 } \left [ \frac{1}{12 } - \frac{89}{12}\,\delta + 21\delta^2 + \frac{25}{3}\,\delta^3 + \left ( 1 + \delta + \frac43\,\delta^2 - 14\delta^3 \right ) \ln2\delta \right ] \nonumber\\ & & \mbox{}+ \frac{4(1-\delta)}{3\delta^2 } \left [ l_2(2\delta ) - l_2(1 - 2\delta ) + \frac{\pi^2}{6 } \right ] + 4\zeta \nonumber\\ & = & \frac{8}{15}\ln 2\delta + 4\zeta - \frac{544}{225 } - \frac{76}{135}\,\delta + o(\delta^2 ) \,.\end{aligned}\ ] ] for the optimal choice @xmath27 , corresponding to the largest value of @xmath69 , and in the ps scheme with @xmath170 and @xmath171 , we find that the relative contributions of the three terms involving @xmath100 , @xmath101 , and @xmath164 in ( [ hybrid ] ) are approximately @xmath172 , indicating a rapid convergence of the hybrid expansion , and supporting our claim that resummation effects are unimportant for the quantity @xmath164 , especially since more than 80% of the contributions to this quantity come from power corrections . we now turn to the numerical evaluation of our result for the decay rate @xmath74 and investigate the perturbative uncertainties in the calculation . the uncertainties related to higher - order power corrections @xmath173 have been estimated previously @xcite . the value of the ps mass at the subtraction scale @xmath174gev has been determined using a sum - rule analysis of the @xmath175 production cross section near threshold @xcite . the result is @xmath176gev , which corresponds to @xmath177gev in the @xmath117 scheme . the most recent value for the upsilon mass as reported in the second paper in @xcite is @xmath178gev and corresponds to @xmath179gev . this paper also contains a sum - rule analysis of the @xmath175 threshold region , yielding @xmath180gev . whereas charm - quark mass effects were included in @xcite they were neglected in @xcite . when comparing the results for the @xmath2 decay rate obtained using different mass definitions , one should use numerical values that correspond to the same value of the @xmath117 mass . combining the values reported above , and applying a small charm - quark mass correction in the case of @xcite , we choose @xmath181gev as a reference value . the correspondingly shifted values for the low - scale subtracted quark masses are @xmath182 below we will indicate how our results would change if different mass values were used . given a value of @xmath183 , we use ( [ eq : psmass ] ) to compute the running ps mass at the scale @xmath125 by solving the nlo equations @xmath184 , \nonumber\\ { m_b^{\text{ps}}}(\mu_f ) & = & \mu_c + \sqrt{\mu_c^2+q_0 ^ 2 } \,.\end{aligned}\ ] ] for instance , with the optimal choice @xmath27 we obtain @xmath185gev , @xmath186gev and @xmath187 for @xmath163 . the next question that arises is which values of the parameter @xmath133 we should consider . a priori any value @xmath126 is reasonable , but in practice the choice is constrained by the following considerations . first , it is known that the characteristic scale for the total @xmath2 decay rate , while parametrically of order @xmath123 , it is numerically much smaller than the @xmath4-quark mass , of order 12gev . a theoretical argument supporting this finding has been presented in @xcite . because the cut on the dilepton invariant mass reduces the phase space for the hadronic invariant mass and energy , the characteristic scale for the decay rate @xmath74 is lower than that for the total rate . hence , the largest reasonable value of @xmath133 is about 2 , corresponding to a subtraction scale @xmath188gev . likewise , we should not consider renormalization scales larger than about 2gev . this conclusion is also borne out by closer inspection of the perturbative expansion coefficients in ( [ master ] ) . whereas for @xmath170 the coefficients of the various terms proportional to @xmath156 in the expressions for @xmath100 and @xmath101 take moderate values ( and are of much smaller magnitude than in the pole scheme ) , some of these coefficients become large for @xmath189 . on the other hand , @xmath133 can also not be taken too small . the formal limit @xmath190 corresponds to the pole scheme and reintroduces the bad perturbative behavior of expansions based on using the pole mass . in addition , there are significant uncertainties in the calculation of the running ps mass when the subtraction scale is much below 1gev , where nnlo corrections become important . based on these observation we will employ the ps mass subtracted at the scale @xmath191 ( @xmath163 ) as our default choice , and consider the choices @xmath192 and @xmath193 as extreme variations . to present our results for the branching ratio we define a function @xmath194 such that @xmath195 where @xmath196 is the @xmath1-meson lifetime . in figure [ fig : mudep ] , we show the renormalization - scale dependence of @xmath194 for the optimal choice @xmath27 and different mass definitions . we scan the parameters @xmath197 and @xmath166 entering the expression for @xmath164 in ( [ remainder ] ) in the ranges @xmath198gev and @xmath199 . the sensitivity to the values of these parameters is very small , as indicated by the widths of the bands . the vertical gray lines indicate the `` perturbative window '' between @xmath161 and @xmath200 ( calculated with @xmath163 ) . in this range there are no large logarithms , and the coupling @xmath156 is reasonably small . we observe good stability of the results in this region , and an excellent agreement between the ps scheme and the upsilon scheme for scales above 1.7gev . for lower scales , the results in the upsilon scheme become unstable , whereas the stability interval for the ps scheme extends down to @xmath201gev . remarkably , the curves corresponding to the different mass definitions all cross at a point close to the center of the window . we take the central value of the band corresponding to the ps mass with @xmath163 , evaluated at the center of the window , as our central value , and use the results obtained with @xmath163 , @xmath161 and @xmath192 , @xmath202 to get an estimate of the perturbative uncertainty . the error bar at the default point reflects the sensitivity to the value of the @xmath4-quark mass . the branching ratio scales as approximately the tenth power of @xmath123 @xcite , and because of this strong dependence this is the dominant theoretical uncertainty in our prediction . the dependence on the @xmath4-quark mass can be parameterized as @xmath203 where @xmath204 , and the exponent @xmath205 increases with @xmath39 ( see table [ tab : results ] below ) . alternatively , one may substitute the ratio @xmath206gev with @xmath207gev or @xmath208gev . the @xmath209mev error assumed in the figure is less conservative than the @xmath210mev advocated in @xcite , but larger than the @xmath211mev found in @xcite . = 0.5 cm [ cols="^,^,^,^,^,^",options="header " , ] table [ tab : results ] shows our final theoretical predictions for @xmath194 as a function of the cutoff on the dilepton invariant mass . we separately list the errors due to the uncertainty in the @xmath4-quark mass , the perturbative uncertainty , and higher - order power corrections . each of the quantities @xmath99 in ( [ hybrid ] ) receives corrections of order @xmath212 , whose contributions are essentially unknown . because of the prefactor @xmath213 in this equation , the corresponding contribution to the decay rate is to first order independent of @xmath39 and scales like @xmath214 . correspondingly we assign an error of @xmath215 , where we have used @xmath216mev as a typical hadronic scale . recently the possibility has been entertained that it might be possible to lower the @xmath22 cutoff below the charm threshold , because the contributions from low - lying charm states can be subtracted using experimental data @xcite . we have therefore included a result for @xmath217gev@xmath67 in the table . the results in table [ tab : results ] compare well with the ones obtained in @xcite , where a value corresponding to @xmath218 was obtained for @xmath27 . a new element of the present analysis is the study of different mass definition schemes , which allows us to obtain a more reliable estimate of the perturbative uncertainty . the authors of @xcite have calculated the decay rate in the upsilon scheme using fixed - order perturbation theory at the scale @xmath160 . with this choice of scale and scheme we find @xmath219 , significantly less than the central value shown in the table . these authors have also included a partial calculation of higher - order corrections ( terms of order @xmath220 ) and used @xmath221 for the upsilon mass ( with no uncertainty ) , thereby neglecting the nonperturbative contribution estimated in @xcite . their result reported in @xcite corresponds to @xmath222\pm 1.7\,[\mbox{power}])$ ] . since the @xmath220 terms are related to the choice of scale in the leading - order correction , it is not clear to us why the quoted perturbative error is smaller than the contribution of these terms . more importantly , however , the dominant theoretical uncertainty due to the strong sensitivity to the @xmath4-quark mass has been neglected in @xcite . we have presented a calculation of the branching ratio for the inclusive decay @xmath2 with a cut @xmath223 on the dilepton invariant mass . since the typical parton momenta @xmath224 after the cut are of order the charm - quark mass , we have performed a two - step expansion in the ratios @xmath15 and @xmath16 and summed logarithms of the form @xmath17 at next - to - leading order . to improve the quality of the perturbative result , we have eliminated the pole mass in favor of a low - scale subtracted @xmath4-quark mass . we find that both the potential subtracted and the upsilon mass lead to consistent results for the rate . considering the residual renormalization - scale dependence and the variations between different definitions of the heavy - quark mass , we estimate the perturbative uncertainty to be of order 10% . this is comparable to the size of unknown higher - order power corrections , but smaller than the error arising from the uncertainty on the value of the @xmath4-quark mass . if a value of the cutoff not far above the optimal choice @xmath27 can be achieved experimentally , we conclude that @xmath0 can be determined with a precision of about 10% . this would be considerably less than the current theoretical uncertainty in the value of this important standard model parameter . bigi , n.g . uraltsev and a.i . vainshtein , phys . b * 293 * ( 1992 ) 430 [ hep - ph/9207214 ] , ibid . * 297 * ( 1993 ) 477 ( e ) ; + i.i . bigi , m. shifman , n.g . uraltsev and a. vainshtein , phys . * 71 * ( 1993 ) 496 [ hep - ph/9304225 ] ; + b. blok , l. koyrakh , m. shifman and a.i . vainshtein , phys . d * 49 * ( 1994 ) 3356 [ hep - ph/9307247 ] , ibid . * 50 * ( 1994 ) 3572 ( e ) . a. hoang , _ bottom quark mass from upsilon mesons : charm mass effects _ , preprint cern - th-2000 - 227 [ hep - ph/0008102 ] ; _ light quark mass effects in bottom quark mass determinations _ , preprint cern - th-2001 - 48 [ hep - ph/0102292 ] .	we reduce the perturbative uncertainty in the determination of @xmath0 from inclusive semileptonic @xmath1 decays by calculating the rate of @xmath2 events with dilepton invariant mass @xmath3 at subleading order in the hybrid expansion , and to next - to - leading order in renormalization - group improved perturbation theory . we also resum logarithmic corrections to the leading power - suppressed contributions . studying the effect of different @xmath4-quark mass definitions we find that the branching ratio after the cut is @xmath5 , where the dominant error is due to the uncertainty in the @xmath4-quark mass . this implies that @xmath0 can be determined with a precision of about 10% . clns 01/1737 + hep - ph/0105217 + may 21 , 2001 * improved determination of @xmath0 from inclusive semileptonic @xmath1-meson decays * thomas becher and matthias neubert + _ newman laboratory of nuclear studies , cornell university + ithaca , ny 14853 , usa _
lvy statistics applies to the description of a particular class of random walks obeying a ` heavy - tailed ' probability distribution with power law dependence @xcite . lvy statistics has been studied in great details owing to its importance in numerous scientific areas where such random walks can occur , and where the long tail of lvy distributions is essential for the prediction of rare events , be that in chemistry @xcite , biology @xcite , or in economics @xcite . in physics , lvy statistics appears often as a result of a complex dynamics , in various transport processes of heat , sound , or light diffusions , in chaotic systems , and in laser cooling where it has broad applications for sub - recoil cooling techniques @xcite . it was shown long ago that in a gas the distribution of energy shifts resulting from the van der waals interaction is given by a lvy - type distribution , known as van der waals profile @xcite . however , lvy statistics has so far not been applied to ensembles of rydberg atoms , typically interacting via pair - wise long - range potentials . being able to simulate very large systems of interacting rydberg atoms with simple statistics would be very valuable when the size of the hilbert space becomes too large to allow for full quantum simulation . moreover , a statistical approach to the simulation of the laser excitation of atoms to rydberg state would be significantly faster than other approximate computational techniques such as monte carlo rate equations ( mcre ) sampling @xcite . most importantly , any approach for simulating the laser excitation must take into account the effect of the interaction blockade @xcite . the interaction blockade prevents the excitation of pairs of rydberg atoms with internuclear separations shorter than the so - called blockade distance @xmath5 , which may give rise to highly correlated many - body states and tends to favor the excitation of ordered structures @xcite . more precisely , the dynamics of the excitation of atoms highly depends on the shifts of the individual rydberg energy levels . for a given atom @xmath6 and in case of isotropic interactions , this shift due to all the surrounding rydberg atoms @xmath7 at @xmath8 distance writes in frequency unit , @xmath9 , where @xmath10 is the planck constant , and @xmath11 is the interaction strength coefficient . the most common interaction is of van der waals type with @xmath3 exponent , while another possibility is the resonant dipole - dipole interaction @xmath12 , and in practice both of them can be isotropic for well - chosen experimental conditions . in order to develop a statistical treatment of the laser excitation , it is critical to determine the probability distribution @xmath13 of energy level shifts @xmath14 , at a particular location @xmath15 independently chosen from the position of any excited rydberg atoms . the primary purpose of this article is to show @xmath16 is mostly given by lvy stable distributions , and that lvy statistics is suitable for the description of the dynamics of an interacting rydberg gas . in section [ interacting ] , a general statistical analysis of ensembles of interacting particles is reported , valid when the space dimension @xmath0 and the interaction exponent @xmath17 satisfy the condition @xmath2 . in subsection [ interactingindependent ] , we first remind the case of a randomly distributed and interacting gas of point - like particles . this result is interesting for rydberg atom physics if the bandwidth of the laser excitation is sufficiently large , whereby the blockade distance becomes vanishingly small in comparison to the interatomic distances . then in subsection [ interactinghardball ] , we simulate @xmath13 in the framework of the hard ball model with random distributions of impenetrable interacting particles of radii @xmath18 , for which the pair - correlation function closely resembles that of the real system @xcite . we demonstrate with numerical simulations that the distribution @xmath16 remains of lvy stable type to a very good approximation , and confirm this result analytically at the limit of small @xmath5 . in section iii , the actual laser excitation of a cold gas to rydberg state is studied more specifically for @xmath19 in electromagnetically induced transparency - like condition . in subsection [ levylaser ] , we use the mcre approach in order to simulate the two - photon excitation with near - resonant lasers , and confirm the distribution @xmath16 yielded by the hard ball model . finally in subsection [ rydbergeit ] , we develop a model of the probe light transmission through an atomic gas driven in configuration of electromagnetically induced transparency involving a rydberg state ( rydberg eit ) , with statistical modeling of the interactions based on lvy distributions . the simulated transmission spectra match closely those obtained directly by the mcre approach , and are in good agreement with experimental data acquired as in reference @xcite this subsection provides a reminder on the distributions of energy level shifts @xmath13 for randomly distributed point - like particles interacting via @xmath1 pair - wise potentials . we assume @xmath20 for convenience , although all the calculations presented in this section are also valid for @xmath21 , as long as the change of sign is properly taken into account in all the expressions of the distribution functions . as demonstrated in appendix [ tmodel ] for any isotropic interaction of @xmath1 type and any dimension @xmath0 , @xmath2 , the distribution @xmath16 is equal to the lvy stable distribution @xmath22 , whose laplace transform takes the simple form @xmath23 . the two parameters are given by @xmath24 and @xmath25 , where @xmath26 is the density of interacting particles , @xmath27 stands for the gamma function and @xmath28 is the unit surface in @xmath0-dimension , namely @xmath29 , @xmath30 or @xmath31 . in the case @xmath19 , the distribution of shifts @xmath16 is therefore given by the following distribution , obtained long ago and known as lvy distribution or van der waals profile @xcite @xmath32 with @xmath33 . this distribution is highly peaked at @xmath34 and possesses a very slow decreasing tail which is mostly due to the interaction with the nearest neighbor particle . simple analytical expressions of the distribution functions exist for other choices of the parameters @xmath17 and @xmath0 as well , especially for @xmath35 and @xmath36 as given in apendix [ tmodel ] . the outstanding question addressed in this subsection is how the distributions of energy level shifts @xmath37 obtained in the previous subsection are modified when the particles become impenetrable . we consider a system composed of interacting hard balls of radii @xmath18 , hence the pair - correlation function is stepwise , equal to 0 at distances shorter than @xmath5 and 1 otherwise . the numerical simulations are performed by markov chain monte carlo ( mcmc ) computations of hard balls @xcite . remarkably , we find with this method that the distribution of energy shifts @xmath37 remains of lvy stable - type to a very good approximation , up to @xmath38 . as illustrated in fig . [ comparisonlevyhs ] for the particular case of @xmath19 , the numerical simulations are very well fitted by @xmath39 lvy stable distributions , where @xmath24 , and @xmath40 is slightly changed compared to the case of point - like particles . computed by markov chain monte - carlo ( mcmc ) simulation of hard balls ( black solid line ) , and the lvy distribution function @xmath41 with @xmath42 where @xmath43 and @xmath44 ( red dashed line ) . the system is composed of a cube of @xmath45 m edge length , containing hard balls of diameter @xmath46 m with wigner - seitz radius @xmath47 m , and periodic boundary conditions are assumed . the histogram of @xmath13 is realized with the computed data of the total shift at the center of the cube , @xmath48 , accumulated over 8000 different trajectories , with a different initial state and 200000 markov chain steps for each trajectory . [ comparisonlevyhs],width=302 ] the exact form of the distribution @xmath13 can be recovered analytically in the limit of small @xmath5 as shown in appendix [ tmodel ] . we find for small @xmath5 that for any interaction exponent @xmath17 and dimension @xmath0 , @xmath2 , the distribution of energy shifts @xmath16 is given by a lvy stable distribution @xmath49 of parameter @xmath50 , with @xmath51 and where @xmath52 is the @xmath0-dimensional volume of a ball of radius @xmath5 . this analytical approximation is in good agreement with monte - carlo simulations at small @xmath53 ratios , as illustrated in fig . [ correctioncoefhs ] for the case @xmath19 . the @xmath40 parameter extracted from fitting of the simulated distribution verifies the relation @xmath54 , where @xmath55 is a correction which is indeed equal to the analytical finding at small @xmath53 ratios , i. e. @xmath56 . there is a higher order correction for large @xmath53 ratios , which will invite further investigation in the future . the lowest order dependence on the exclusion volume @xmath57 is not surprising as it appears also in the derivation of the second virial coefficient of the equation of state of a van der waals gas @xcite . the shift of the distribution peak value to higher energies is a manifestation of the ordering at higher @xmath5 , as correlations tend to decrease the probability for a region locally empty of hard balls . the long tail of the distribution , which varies as @xmath58 for large shifts in fig . [ comparisonlevyhs ] , is due mostly to the presence of the nearest neighbor . indeed , since the level shift is calculated at a particular position chosen independently from the positions of the hard balls , the tail of the distribution still varies as in the case of point - like particles . , as a function of @xmath59 . the analytical approximate curve ( red solid line ) is given by @xmath60 with @xmath61 . the data points ( blue circles ) are extracted from fitting of the simulated energetic distributions @xmath13 to a lvy distribution @xmath62 of parameter @xmath63 . the distributions @xmath13 are simulated with mcmc computations as in fig . [ comparisonlevyhs ] , at fixed particles density , and with various @xmath5 . the fit of the data points to a model curve of the form @xmath64 gives @xmath65 and @xmath66 ( black dashed line ) . [ correctioncoefhs],width=302 ] as shown in the previous section from the hard ball model , the distribution of energy shifts @xmath13 is given by a lvy stable distribution in the approximation of a stepwise pair - correlation function . while restricting the discussion in this section to the case @xmath19 , which is of most relevance for rydberg atom experiments , we show that the distribution of shifts @xmath67 remains a lvy distribution in the real physical system , where the laser excitation results into less ideal pair - correlation functions . a common method for producing rydberg atoms , shown in fig . [ setuprydeit ] , consists of a ladder scheme of excitation with two lasers ( 1 ) and ( 2 ) , coupling the @xmath68 and @xmath69 transitions , respectively , where @xmath70 , @xmath71 and @xmath72 are three atomic levels , ground , intermediate and rydberg , respectively . without loss of generality , we will only consider @xmath73 atoms , and the three energy levels @xmath74 , @xmath75 whose decay rate is @xmath76mhz , and @xmath77 , where the principal quantum number @xmath78 can be varied for tuning the @xmath79 parameter . in this subsection , we calculate the distribution @xmath13 in an homogeneous system of a gas of interacting atoms coupled to two laser fields @xmath80 and @xmath81 as shown in fig . [ setuprydeit ] , for a wide range of excitation conditions . the mcre approach is a well - known computation method of the excitation dynamics in such a system , which can provide approximate results for the case of interest here @xcite . thus , the distribution of energy level shift is computed using this approach , and according to the method described in reference @xcite and summarized in appendix [ levymodel ] . and @xmath82 are detuned @xmath83 and @xmath84 from the resonance of the @xmath85 and @xmath86 transitions , respectively . [ setuprydeit],width=151 ] computed using the mcre approach ( black solid line ) and the lvy distribution @xmath87 ( red dashed line ) . the system is composed of a cube containing 1000 atoms at a ground state density of @xmath88 and the computation is carried out with periodic boundary conditions . @xmath13 corresponds here to the distribution for an equilibrium excitation of @xmath73 atoms , with @xmath89 mhz , @xmath90 mhz , @xmath91 , rydberg state @xmath92 , and decay rate of atomic coherences @xmath93 and @xmath94 mhz . the wigner - seitz radius is equal to @xmath95 , where @xmath96 m . the histogram is plotted with the rydberg energy level shifts of all the atoms excluded the ones that are in the rydberg state , computed over 23 different trajectories , with nearly 1300 monte carlo time steps each . [ comparisonlevymc],width=302 ] as can be seen on fig . [ comparisonlevymc ] , the distribution of energy shifts indeed closely resembles a lvy distribution @xmath97 , where @xmath98 , @xmath26 is the density of atoms excited to the rydberg state obtained from the mcre simulation , and @xmath5 is the blockade distance . here , the blockade distance @xmath5 is defined as the distance that satisfies @xmath99 , where @xmath100 is the single - atom rydberg population calculated for a given rydberg energy level shift @xmath101 , as obtained from the solution @xmath102 of the single - atom master equation reminded in appendix [ levymodel ] . we have varied independently the total atomic density @xmath103 in the range of 0 to @xmath104 , @xmath105 in the range of @xmath106 to @xmath107 , the detuning @xmath108 from -8 to + 8 mhz and finally the rabi frequency @xmath109 from 0.5 to 3 mhz , while keeping the rabi frequency @xmath110mhz and @xmath111 . we find @xmath112 is nearly constant in this range of parameters except for a small variation from 0.25 to 0.35 versus probe detuning . this variation may be a consequence of the higher order correction term in the expansion of the parameter @xmath40 that is not considered in the fits of the numerical distributions . it may also be due to the pair correlation function enhancement on the blue two - photon detuning @xcite , which may induce a small departure from the hard ball picture . an important conclusion drawn from this calculation is that the distribution @xmath13 of level shifts in this system closely matches the simple lvy distribution . the effect of the dipole blockade is accounted almost entirely in a single coefficient @xmath113 given by eq . . in principle , it is possible to probe this type of distribution @xmath13 by measuring the fraction of atoms excited to a rydberg state @xmath114 in conditions of weak driving , and in presence of atoms already excited to another state @xmath115 , where the @xmath116 energy levels of the probe atoms are strongly affected by the van der waals interaction with @xmath115 atoms . many experimental and theoretical studies have investigated the effects of dipolar interactions on rydberg eit - like spectra @xcite . recently , it has been demonstrated that the energy level shifts in an interacting rydberg gas are responsible for a readily observable spectral shift of the eit transparency window , in addition to the strong dephasing that was observed in previous experiments @xcite . the purpose of this section is to show that a light propagation model based on lvy statistics can capture most of the features observed experimentally . rydberg eit consists in driving the two transitions of fig . [ setuprydeit ] , with weak probe ( 1 ) and strong coupling ( 2 ) near - resonant lasers . for clarity , subscripts 1 and 2 of the previous subsection will be replaced by subscripts @xmath17 and @xmath117 , respectively . the propagation of the probe light in an eit medium is generally well described with the framework of coupled maxwell - bloch equations . in the following analysis , we only consider the case @xmath118 . moreover , the kinetics of the atoms is neglected , while steady state equilibrium of the excitation is still assumed , as it is reached on a timescale much shorter than the duration of the typical experiment . finally , the system is assumed to be locally homogeneous , and the gas sufficiently dilute to neglect possibilities of light modulation instabilities @xcite . consequently , maxwell s equations reduce to the following one - dimensional differential equation for the probe field rabi frequency @xcite @xmath119 where @xmath120 is the atomic susceptibility along the probe beam propagation direction @xmath121 . then , the probe intensity @xmath122 is simply obtained from the differential equation @xmath123 i_p$ ] . in principle , the calculation of the susceptibility involves the solution of the full master equation of the many - body system . in our system , the susceptibility @xmath124 can be approximated as a local spatial average over the optical responses of the different atoms . in a simple manner , the susceptibility can be written as the following average over interaction configurations , the rydberg fraction @xmath125 , and the two - level susceptibility @xmath126 , was found from monte - carlo simulations @xcite , which is however verified only for zero probe detuning and when @xmath127 . ] @xmath128 where @xmath129 is the distribution function of energy level shifts due to the van der waals interaction , @xmath125 is the averaged rydberg population fraction , and @xmath130 is the single - atom 3-level susceptibility calculated at a given rydberg energy level shift @xmath131 or equivalently for an effective coupling laser detuning @xmath132 ( cf . appendix [ levymodel ] for a more detailed definition of @xmath133 ) . as @xmath129 is given in a very good approximation by the lvy distribution @xmath134 with @xmath135 , only the knowledge of @xmath125 is needed in eq . in order to calculate the susceptibility . according to the super - atom model @xcite , a reasonable analytical approximation for @xmath125 is obtained from the following equation @xmath136 where @xmath137 is the volume of the blockade sphere of radius @xmath5 , and @xmath138 is the rydberg population fraction in absence of interactions . rydberg state population fraction @xmath125 for an ensemble of interacting atoms at a density of @xmath139 , driven in configuration of eit with probe field rabi frequency @xmath140 mhz and coupling field rabi frequency @xmath141 mhz . the plotted curves are as follows : @xmath142 $ ] obtained from eq . , where the rydberg fraction @xmath125 is calculated with eq . ( red dotted line ) ; @xmath143 $ ] computed with the mcre simulation over 800 atoms in a square box ( black thin solid line ) ; the population fraction in the rydberg state @xmath125 obtained directly with the mcre simulation ( black thick solid line ) ; and the approximate rydberg population fraction @xmath125 as given by eq . ( green dashed line ) . [ appendixfig],width=321 ] the imaginary part of the susceptibility as calculated with eqs . and , @xmath142 $ ] , may be directly compared to that obtained from mcre simulations , @xmath143=\frac{\hbar n_{at}}{i_p \gamma_e } f_e$ ] , where @xmath144 is the average intermediate state population fraction , and @xmath103 is the atomic density . although none of the two approaches yield exact solutions for the susceptibility , they lead to very similar results , as shown in fig . [ appendixfig ] , except for a small discrepancy near the autler - townes resonances . there is actually a larger mismatch on the blue detuning side of the @xmath145 transition when we compare the rydberg population fractions given by eq . and that yielded by the mcre simulation , as shown in fig . [ appendixfig ] . the susceptibility , proportional to the intermediate state population fraction , is however not very sensitive to the rydberg population fraction at such detuning , where the rate of single - photon excitation is high . this explains the good agreement between @xmath142 $ ] and @xmath143 $ ] even for probe detuning on the blue side of the @xmath145 transition . we have seen so far how lvy statistics may be used successfully for describing rydberg eit by comparison to mcre computations . in order to fully validate the approach , we provide in the following paragraph a direct comparison between the simulation and some experimental data acquired in conditions as described in reference @xcite . in the experiment of reference @xcite , transmission spectra of the probe light passing through an atomic cloud driven in conditions of rydberg eit were acquired as a function of probe detuning @xmath146 , while keeping @xmath118 . a couple of such experimental spectra are shown in fig . [ figure2levy ] along with spectra calculated using eq . . , an input probe field rabi frequency of @xmath147 , the rydberg state @xmath148 , and in the case of two different gaussian atomic clouds . the first one is of @xmath149 radius @xmath150 with peak atomic density , @xmath151 ( @xmath152 ) , while the second one is of @xmath149 radius @xmath153 with peak atomic density @xmath154 ( @xmath155 ) . the dashed lines are generated by the lvy statistics model with the corresponding experimental parameters as inputs while the solid lines are obtained using the mcre approach ( see text).[figure2levy],width=321 ] the spectra simulated from eq . with the approximate susceptibility of eq . are in good agreement with the experimental ones as shown in fig . [ figure2levy ] . they are also very similar to the spectra computed from eq . with the input of @xmath156 . in conclusion , eqs . , , and yield reliable and easily computable results , which capture most of the experimental features , and are just as good as the results of mcre computations . there is a slight discrepancy between experimental and simulated data for larger atomic densities , that was also observed with the enhanced mean - field model of reference @xcite . it is mostly due to motion induced dephasing , which should be accounted for with a dynamical model beyond the scope of the frozen gas approximation assumed in this section . the distribution of energy shifts in a gas of interacting rydberg atoms was shown to follow a simple lvy - type distribution which depends mostly on the rydberg atom density @xmath26 , the interaction strength @xmath11 , and the blockade distance @xmath5 . based on this result , we have demonstrated that the recent observation of spectral shift and dephasing of rydberg eit spectra in presence of strong rydberg interactions is well captured by a model where the susceptibility is calculated as a statistical average . spectroscopic measurements of the distribution of energy level shifts in an interacting rydberg gas should provide a complete test of the theory presented in this article . we are grateful to benot grmaud for helpful discussions . this work is supported by the national research foundation , prime minister s office , singapore and the ministry of education , singapore under the research centres of excellence programme , as well as by singapore ministry of education academic research fund tier 2 ( grant no . moe2015-t2 - 1 - 085 ) . a lvy stable distribution @xmath157 with exponent @xmath158 and scale @xmath159 is a distribution on @xmath160 with laplace transform @xmath161 for notational convenience , we introduce the parameter @xmath162 , write the laplace transform as @xmath163 , and define @xmath164 . the generic example is the lvy ( van der waals ) distribution @xmath165 other simple examples are @xmath166 , where ai stands for the airy function , and @xmath167 $ ] , expressed with modified bessel functions of the second kind @xcite . we consider an energy of the type @xmath168 where @xmath169 are the positions of @xmath170 ( independent ) atoms uniformly distributed in the ball @xmath171 of radius @xmath172 and volume @xmath173 ; the notation @xmath28 indicates the surface of the unit sphere in @xmath174 , and we assume that @xmath175 . the laplace transform of the energy reads ( for @xmath176 ) @xmath177 where the limit was taken for @xmath178 at fixed density @xmath26 . the final result @xmath179 is the laplace transform of a lvy stable distribution with exponent @xmath180 and parameter @xmath181 . this result is a particular case of the more general theory of the sum of independent and random variables having ` heavytail ' probability distributions with power - law dependence @xcite . the probability distribution of the energy is indeed that of a sum of random variables @xmath182 with power - law dependent probability distributions @xmath183 . a full analytical derivation of the distribution of energy shifts for non - independent distributions of particles is beyond the scope of this article . we emphasize here the main arguments leading to the derivation of the coefficient @xmath184 of eq . in case where the parameter @xmath185 is small and considering a simplified system of hard balls of radii @xmath18 in @xmath174 . the calculation is based on the cluster expansion used for example for the derivation of the first corrections to the equation of state of a van der waals gas @xcite . again , the analytical derivation of the distribution of energy shifts relies on the computation of its laplace transform as @xmath186 where @xmath187 is equal to 0 for @xmath188 and 1 otherwise . @xmath189 is the partition function given by : @xmath190 the first order approximation consists in dropping all the products involving two @xmath191 functions from the integrands , that means @xmath192 . with this approximation , the partition function reduces to @xmath193 . this result is valid in the limit of small enough @xmath194 and @xmath78 such that the correction to the partition function remains relatively small . further , the number of atoms @xmath78 will be assumed sufficiently large in order to calculate asymptotic behaviors while still having @xmath195 . this approximation yields that @xmath196 where on the one hand we have @xmath197 and on the other hand and very similarly , we have @xmath198 substitution of the last two results back into @xmath199 , and using the assumptions @xmath200 and @xmath201 yield that @xmath202 in other words , in @xmath174 and with interaction potentials varying as @xmath203 , finite size effects are entirely accounted for as a shift of the parameter @xmath204 , @xmath205 with @xmath206 and @xmath57 the exclusion volume of a ball of radius @xmath5 . this analytical result seems to hold for somewhat larger @xmath5 than given by the range of validity of the present calculation . indeed , the condition @xmath207 for large @xmath78 up to 10 is equivalent to @xmath208 , i. e. @xmath5 can not be larger than a small fraction of the interatomic distance @xmath209 , whereas the simulation by markov chain monte - carlo of hard balls in fig . [ correctioncoefhs ] gave a similar correction @xmath210 up to @xmath211 . the aim of this appendix is to come back to the different approaches for computing the imaginary part of the susceptibility as in fig . [ appendixfig ] . mcre is based on the computation of single atom transition rates that depend on the rydberg energy shift of each atom in the presence of surrounding atoms already excited to rydberg state . with the ladder configuration of excitation with near - resonant lasers of fig . [ setuprydeit ] , performing the adiabatic elimination of atomic coherences on the master equation in order to derive rate equations suitable for monte carlo simulations fails as it may yield negative transition rates even for a non interacting system . for simulations that are interested only in steady state solutions , this problem can be overcome by considering strictly positive transitions rates that evolve a system of non - interacting atoms into the correct single atom steady states but does not reproduce properly the states of the system at intermediate time steps @xcite . more precisely , the state vector of the single atom populations , @xmath212 is assumed to obey rate equations @xmath213 where @xmath214 is a well - designed matrix having positive off diagonal elements ensuring transition rates between two different states are always positive , and @xmath215 equals the steady state populations of the single atom master equation . this rate equations solution is readily extended to the case of an ensemble of interacting atoms by calculating the effect of energy shifts on the matrix @xmath214 , locally and at each time step of the computation , i. e. computing the effect of the detuning change @xmath216 of laser ( 2 ) in fig . [ setuprydeit ] . in fig . [ appendixfig ] , the intermediate state population fraction @xmath144 is used in order to calculate the susceptibility from the simple formula @xmath217=f_e\ , \frac{\hbar c n_{at } } { i_p \gamma_e}$ ] , and is obtained directly from mcre computation with the @xmath214 matrix of reference @xcite @xmath218 this simple form of the matrix tends to define very large transition rates . other matrices exist , exhibiting transition rates closer to the actual ones @xcite . in fig . [ appendixfig ] is also plotted @xmath219 $ ] , where @xmath220 $ ] writes @xmath221=\left ( 1- f_r \right ) \int_0^{\infty } \mathrm{im}\left[\chi_0 ( \nu ) \right ] f ( \nu ) d\nu , \label{chiexact2}\ ] ] in this expression of @xmath222 $ ] , @xmath223 is a weighting distribution depending on @xmath224 , the local shift of the rydberg energy level , @xmath125 is the rydberg atom fraction , while @xmath133 is the non interacting 3-level atom susceptibility for the probe light . @xmath133 is given as below where @xmath226 is the dipole matrix element for the @xmath227 transition , @xmath103 is the atomic density , assumed locally uniform , and @xmath228 is the atomic coherence . @xmath228 is itself given by the solution of the single - body master equation @xmath229 + { \ensuremath{\mathcal{l } } } { \varrho}$ ] where @xmath10 is the hamiltonian of interaction with the probe and coupling lights . @xmath230 is a set of liouvillians simulating the effects of the dissipative processes . their expressions , which are not reminded here and can be found in reference @xcite , include notably the decay rate of atomic coherence @xmath231 which is of the order of @xmath232 mhz in absence of interactions . in fig . [ figure2levy ] , we use a slightly larger value for the intrinsic decay rate of atomic coherence , @xmath233 mhz , in order to account for residual effects of dipolar forces . the rydberg atom fraction @xmath125 in eq . may be obtained via eq . from the knowledge of @xmath234 , the rydberg atom fraction absent any interaction , and @xmath235 the volume inside a blockade sphere . in our simulations , are solved with @xmath133 , @xmath236 , and @xmath235 calculated to all orders in @xmath237 . 36ifxundefined [ 1 ] ifx#1 ifnum [ 1 ] # 1firstoftwo secondoftwo ifx [ 1 ] # 1firstoftwo secondoftwo `` `` # 1''''@noop [ 0]secondoftwosanitize@url [ 0 ] + 12$12 & 12#1212_12%12@startlink[1]@endlink[0]@bib@innerbibempty @noop _ _ ( , ) @noop * * , ( ) @noop * * , ( ) @noop _ _ , vol . ( , ) @noop _ _ , vol . ( , ) @noop _ _ ( , ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) link:\doibase 10.1103/physreva.87.053414 [ * * , ( ) ] link:\doibase 10.1103/physreva.89.063407 [ * * , ( ) ] @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop ( ) @noop _ _ , vol . ( , ) @noop * * , ( ) @noop ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( )	a statistical analysis of the laser excitation of cold and randomly distributed atoms to rydberg states is developed . we first demonstrate with a hard ball model that the distribution of energy level shifts in an interacting gas obeys lvy statistics , in any dimension @xmath0 and for any interaction @xmath1 under the condition @xmath2 . this result is confirmed with a monte carlo rate equations simulation of the actual laser excitation in the particular case @xmath3 and @xmath4 . with this finding , we develop a statistical approach for the modeling of probe light transmission through a cold atom gas driven under conditions of electromagnetically induced transparency involving a rydberg state . the simulated results are in good agreement with experiment .
the d0 collaboration has recently reported a new measurement of the like - sign dimuon charge asymmetry in semileptonic @xmath0-hadron decays @xcite which disagrees with the standard model ( sm ) prediction by about three standard deviations . although this finding still needs to be confirmed by future experiments , it might be a clue to the presence of unexpectedly substantial @xmath1-violating new physics in the mixing of @xmath2 mesons . subsequently , we have suggested as one of the possible scenarios for the new physics that the d0 result could be attributed to a nonstandard spin-1 particle lighter than the @xmath0 quark , with flavor - changing couplings to the @xmath0 and @xmath3 quarks @xcite . specifically , we showed that the effect of the new particle can lead to a prediction which is consistent with the new data within its one - sigma range . new - physics scenarios involving nonstandard spin-1 particles with masses of a few gev or less have been discussed to some extent in various other contexts in the literature . their existence is generally still compatible with currently available data and also desirable , as they may offer possible explanations for some of the recent experimental anomalies and unexpected observations . for instance , a spin-1 boson having mass of a few gev and couplings to both quarks and leptons has been proposed to explain the measured value of the muon @xmath11@xmath122 and the nutev anomaly simultaneously @xcite . as another example , @xmath13(mev ) spin-1 bosons which can interact with dark matter as well as leptons may be responsible for the observed 511-kev emission from the galactic bulge @xcite . if its mass is at the gev level , such a particle may be associated with the unexpected excess of positrons recently observed in cosmic rays , possibly caused by dark - matter annihilation @xcite . in the context of hyperon decay , a spin-1 boson with mass around 0.2gev , flavor - changing couplings to quarks , and a dominant decay mode into @xmath14 can explain the three anomalous events of @xmath15 reported by the hypercp experiment several years ago @xcite . although in these few examples the spin-1 particles tend to have suppressed couplings to sm particles , it is possible to test their presence in future high - precision experiments @xcite . it is therefore also of interest to see if the light spin-1 boson that could be responsible for the anomalous d0 result contributes to some other @xmath0-meson processes , perhaps with detectable effects , which we will attempt to do in this paper . here we consider the possibility that the new spin-1 particle , which we shall refer to as @xmath16 , has flavor - conserving couplings to the electron and muon , besides its flavor - changing couplings to the @xmath0 and @xmath3 quarks . accordingly , it can contribute to a number of rare @xmath5 transitions involving the charged leptons , via the quark - level process @xmath17 , where @xmath4 . in particular , we will deal with the impact of @xmath16 on the inclusive @xmath0-meson decay @xmath18 and the exclusive ones @xmath19 and @xmath20 , all of which have been observed @xcite , as well as the leptonic decay @xmath21 . using the existing experimental information on the @xmath0-meson decays as well as the recent data from d0 on @xmath2 mixing , we will explore constraints on the couplings of @xmath16 . since @xmath16 has flavor - conserving interactions with the leptons and hence contributes to their anomalous magnetic moments , we will also take their measured values into account . in the following section , we write down the relevant lagrangians and derive the amplitudes for the processes of interest . in sec . [ numbers ] , we present our numerical results for the constraints on the @xmath16 couplings and provide some predictions which may be tested at lhcb or future @xmath8 factories . we conclude in sec . [ concl ] and collect some of the formulas in appendixes . we adopt a model - independent approach , assuming in addition that @xmath16 carries no color or electric charge and its couplings to the fermions have both vector and axial - vector parts . the lagrangian for its flavor - changing interactions with the @xmath0 and @xmath3 quarks is then @xcite @xmath22 where @xmath23 and @xmath24 parametrize the vector and axial - vector couplings , respectively , and in general can be complex , which would be new sources of @xmath1 violation . moreover , the flavor - conserving couplings of @xmath16 with a charged lepton @xmath25 is described by @xmath26 where @xmath27 and @xmath28 are real parameters because of the hermiticity of @xmath29 . we will study only the @xmath30 and @xmath31 cases due to lack of the relevant data for @xmath32 at present . in the absence of model specifics , @xmath33 and @xmath34 are not necessarily related . in principle , @xmath16 can have additional interactions , flavor - conserving and/or flavor - violating , with other fermions which are parametrized by more coupling constants . we assume that these additional parameters already satisfy other experimental constraints to which they are subject , but which we do not cover in this paper . together @xmath35 and @xmath36 generate the contributions of @xmath16 to the above - mentioned @xmath5 transitions . for the inclusive decay @xmath18 , the resulting amplitude is @xmath37b\,\bar\ell\gamma_5^{}\ell } { m_x^2\bigl(q^2-m_x^2+i\gamma_x^{}m_x^{}\bigr ) } ~,\end{aligned}\ ] ] where @xmath38 is the combined momentum of the dilepton and @xmath39 the total width of @xmath16 . we remark that the presence of the @xmath40 dependence in the denominators distinguishes this new - physics scenario from others involving heavy particles , which would induce four - fermion operators with coefficients independent of @xmath40 , as have been studied in the literature @xcite . the sm counterpart of @xmath41 is well known and given by @xcite @xmath42 ~,\end{aligned}\ ] ] where @xmath43 and @xmath44 denote the usual fine - structure and fermi constants , respectively , @xmath45 are cabibbo - kobayashi - maskawa matrix elements , @xmath46 are wilson coefficients evaluated at a scale @xmath47 , and @xmath48 . from the sum of the sm and @xmath16-induced amplitudes follows the decay rate @xmath49 where @xmath50 \bigl(m_b^2-q^2\bigr)^2 ~,\end{aligned}\ ] ] @xmath51 \bigl(m_b^2 + 2q^2\bigr ) \nonumber \\ & & \hspace{19ex } + \ ; 6\,{\rm re } \biggl ( \kappa\ , c_7^{\rm eff}\ , \frac{g_{vs}^{}+g_{as}^{}}{\delta_x } \biggr ) g_{v\ell}^{}\,m_b^2 \biggr\ } \bigl(m_b^2-q^2\bigr)^2 \nonumber \\ & & + \ ; \frac{\alpha^2g_{\rm f}^2\,\|\lambda_t^{}\|^2}{768\,\pi^5\,m_b^3}\ , \frac{\|\kappa\|^2}{2 } \bigl(\|g_{vs}^{}\|^2+\|g_{as}^{}\|^2\bigr)\bigl(g_{v\ell}^2+g_{a\ell}^2\bigr)\int dq^2\ ; \frac{\bigl(m_b^2 + 2q^2\bigr)\bigl(m_b^2-q^2\bigr)^2}{\|\delta_x\|^2 } ~,\end{aligned}\ ] ] with @xmath52 and @xmath53 having been set to zero and @xmath54 evidently @xmath55 contains not only the @xmath16-induced amplitude alone , but also its interference with the sm one . for @xmath56 , the amplitude follows from the effective hamiltonians @xmath57 which yield eqs . ( [ hb2sx2sll ] ) and ( [ smhb2sll ] ) . adding the @xmath16 and sm contributions , we can write it as @xmath58 \bar\ell\gamma_\mu^{}\gamma_5^{}\ell \bigr\ } ~,\end{aligned}\ ] ] where @xmath59 , @xmath60 } { \delta_x^ { } } ~,\end{aligned}\ ] ] with @xmath61 denoting the form factors for the @xmath62 matrix elements of the @xmath5 quark operators in @xmath57 and being defined in appendix [ ff ] . one can see that the @xmath56 amplitude is independent of @xmath24 . the corresponding decay rate is given in appendix [ rates ] . it is worth mentioning here that the @xmath63-meson decay @xmath64 involving a pseudoscalar meson @xmath65 containing an @xmath66 component in its quark content , such as @xmath67 and @xmath68 , also has an amplitude independent of @xmath24 . consequently , the observation of @xmath64 will provide additional information on the @xmath16 contributions involving the products @xmath69 and @xmath70 . from the squared amplitude for @xmath71 , one can arrive at two additional observables which have been measured besides the branching ratio . they are the @xmath79 longitudinal polarization fraction @xmath80 and the lepton forward - backward asymmetry @xmath81 , which are defined from @xcite @xmath82 where @xmath83 is the angle between the directions of @xmath84 and @xmath85 in the dilepton rest frame . since these observables are ratios of squared amplitudes , their dependence on the hadronic form - factors is partly canceled , which reduces the theoretical uncertainties associated with the form factors . especially @xmath81 is predicted with good precision in the sm to have a zero - crossing point at @xmath86 @xcite , which makes this asymmetry very sensitive to the signals of new physics that can shift the location of the point . in appendix [ rates ] we have written down @xmath80 and @xmath81 in terms of @xmath87 . the amplitude for the contribution of @xmath16 to @xmath10 can be obtained from eq . ( [ hb2sx2sll ] ) after making the approximation @xmath88 in the center - of - mass frame of @xmath89 and neglecting the @xmath39 part . thus , using the matrix elements @xmath90 , @xmath91 , and @xmath92 , we arrive at @xcite @xmath93 since @xmath94 , the sm yields @xmath95 the sum of the two results in the decay rate @xmath96 hence @xmath97 does not probe either @xmath23 or @xmath27 . the @xmath16 interaction with the lepton @xmath25 as described by @xmath36 in eq . ( [ llx ] ) affects the anomalous magnetic moment @xmath98 of @xmath25 at one loop . the @xmath16 contribution to @xmath98 can be expressed as @xcite @xmath99 where @xmath100 , @xmath101 since the anomalous magnetic moments of the electron and muon , @xmath102 and @xmath103 , have been measured precisely , we need to take their constraints into account . to obtain the first set of constraints on the contributions of @xmath16 , we employ the data on the inclusive decay @xmath18 . the babar and belle collaborations have measured its branching ratios @xmath104 for different ranges of the squared dilepton invariant mass , @xmath40 . we take their results for the low- and high-@xmath40 ranges @xmath105 and @xmath106 , respectively . the numbers from babar , @xmath107 and @xmath108 @xcite , and from belle , @xmath109 and @xmath110 @xcite , average out to @xmath111 we also need the sm predictions @xcite @xmath112 upon comparing these experimental and sm values , we can require that the @xmath16 contributions , from @xmath55 in eq . ( [ gb2sx ] ) , satisfy @xmath113 where @xmath114 with @xmath115 being the @xmath8 lifetime . it is clear from eq . ( [ gb2sx ] ) that @xmath116 contains both the @xmath16-mediated amplitude and its interference with sm one . numerically , we use @xmath117ps , the average of the @xmath118 and @xmath119 lifetimes @xcite , @xmath120 , @xmath121 , @xmath122 @xcite , and the lepton and meson masses from ref . @xcite , as well as the wilson coefficients @xmath123 , @xmath124 , and @xmath125 from ref . @xcite , the expression for the complex function @xmath126 given therein . these input parameter values are also used in the rest of the paper . since @xmath127 in eq.([gb2sll ] ) behave as @xmath128 if @xmath129 gets large , they have sizable uncertainties depending on the choice of @xmath129 value . in our numerical treatment of @xmath55 , we set @xmath130gev , as its application in @xmath131 leads to @xmath132 and @xmath133 , which are compatible with the predicted ranges in eq.([smb2sll ] ) from more refined calculations . the second set of constraints comes from the data on @xmath7 . the babar , belle , and cdf collaborations have measured several different observables in these decays @xcite . we will choose the branching - ratio results provided by belle and cdf , as they are available for the specific @xmath40 ranges for which the most recent predictions in the sm with detailed estimates of the uncertainties are also available . their numbers are listed in tables [ expb2kll ] and [ smb2kll ] , respectively . in view of the currently sizable experimental and theoretical errors , we will ignore the numerical effects of @xmath118-@xmath119 and @xmath134-@xmath31 differences on these processes . furthermore , we will take the more precise of each pair of experimental values in table [ expb2kll ] and then compare it to the corresponding sm number in table [ smb2kll ] in order to estimate the allowed range of the @xmath16 contribution @xmath135 . accordingly , we can impose the limits @xmath136\rm\,gev^2 } \,\,\le\,\ , 0.4\times10^{-7 } ~,\end{aligned}\ ] ] @xmath137\rm\,gev^2 } \,\,\le\,\ , 0.5\times10^{-7 } ~,\end{aligned}\ ] ] @xmath138\rm\,gev^2 } \,\,\le\,\ , 0.7\times10^{-7 } ~,\end{aligned}\ ] ] @xmath139 where @xmath140 with @xmath141 being the rates of @xmath19 , whose formulas are given in appendix [ rates ] , minus the purely sm part . the @xmath79 longitudinal polarization fraction @xmath80 and lepton forward - backward asymmetry @xmath81 in @xmath142 have also been measured by babar , belle , and cdf . although the babar @xcite and belle @xcite data on @xmath143 exceed the sm expectation , the most recent cdf measurement @xcite is consistent with it . it is expected that more definitive information on this and other observables will be available with the upcoming results from lhcb in the near future . since most of the current data on @xmath80 and @xmath81 still have significant errors @xcite , of order 40% or greater , we do not use them in exploring the constraints on @xmath16 . .experimental branching - ratios of @xmath7 from belle @xcite and @xmath144 from cdf @xcite , in units of @xmath145 , used to constrain the @xmath16 contributions , for different @xmath40 ranges . the statistical and systematic errors have been combined in quadrature.[expb2kll ] [ cols="^,^,^,^,^",options="header " , ] [ ffparameters ] the general expressions for the ( double ) differential decay rates of @xmath19 arising from various @xmath146 four - fermion operators within and beyond the sm are known in the literature @xcite . here , for completeness , we write down the specific formulas resulting from our amplitudes of interest . the absolute square of the @xmath56 amplitude in eq . ( [ mb2kll ] ) , summed over the lepton spins , is @xmath147 with @xmath148 and @xmath149 . this leads to the differential decay rate @xmath150 \nonumber \\ & & \hspace{19ex } + \ ; the absolute square of the @xmath71 amplitude in eq . ( [ mb2ksll ] ) , summed over the @xmath79 polarization and lepton spins , can be written as @xmath155r\,z + 4\|{\cal c}\|^2 r\bigl(3-v^2\bigr)z + 8\|{\cal f}\|^2 r v^2 z \nonumber \\ & & \hspace{9ex } + \ ; \bigl[\|{\cal c}\|^2+\|{\cal f}\|^2 + \bigl(\|{\cal d}\|^2+\|{\cal g}\|^2\bigr)\lambda\,m_b^4 + 2\,{\rm re}({\cal c}^{\cal d}+{\cal f}^{\cal g})\,m_b^2(r+z-1)\bigr ] \bigl[\lambda-\bigl(\hat t-\hat u\bigr)^2\bigr ] \nonumber \\ & & \hspace{9ex } -\ ; \bigl [ 2\|{\cal e}\|^2r+\|{\cal g}\|^2(z-2 - 2r ) - \|{\cal h}\|^2 z + 2\,{\rm re}({\cal g}^{\cal h})\,(r-1)\bigr ] \lambda\,m_b^4\bigl(1-v^2\bigr)z \nonumber \\ & & \hspace{9ex } -\ ; 2\,{\rm re}({\cal f}^{\cal g}+{\cal f}^{\cal h})\,\lambda\,m_b^2\bigl(1-v^2\bigr)z \,+\ , 8\,{\rm re}({\cal a}^{\cal f}+{\cal c}^{\cal e})\,r(t - u)z \bigr\ } ~,\end{aligned}\ ] ] where @xmath156 , @xmath157 , @xmath158 , and @xmath159 . the corresponding differential decay rate is @xmath160 \bigl(3-v^2\bigr ) \,+\ , 4\,\|{\cal e}\|^2 r\ , v^2\ , z \,+\ , 3\,\|{\cal h}\|^2 \bigl(1-v^2\bigr ) z^2 \nonumber \\ & & \hspace{11ex } + \ ; \frac{\|{\cal f}\|^2}{m_b^4 } \biggl(3-v^2+\frac{24r\,v^2\,z}{\lambda}\biggr ) \,+\ , \|{\cal g}\|^2 \bigl [ ( 1-r)^2 \bigl(3-v^2\bigr)+2(z-2 - 2r)v^2z\bigr ] \nonumber \\ & & \hspace{11ex } + \ ; \frac{2\,{\rm re}({\cal c}^{\cal d})}{m_b^2 } ( r+z-1 ) \bigl(3-v^2\bigr ) \,+\ , \frac{2\,{\rm re}({\cal f}^{\cal g})}{m_b^2 } \bigl[(1-r)\bigl(v^2 - 3\bigr)+2v^2\,z\bigr ] \nonumber \\ & & \hspace{11ex } + \ ; 6\biggl [ \frac{{\rm re}({\cal f}^{\cal h})}{m_b^2 } + { \rm re}({\cal g}^{\cal h})\,(r-1 ) \biggr ] \bigl(v^2 - 1\bigr)z \biggr\ } ~.\end{aligned}\ ] ] from eqs . ( [ d2gb2ksll ] ) and ( [ m2b2ksll ] ) , we can derive the @xmath79 longitudinal - polarization fraction @xmath161 \bigl(1-v^2\bigr)z \nonumber \\ & & \hspace{28ex } + \ ; \frac{\|{\cal c}\|^2}{m_b^4 } \biggl [ 2r\bigl(3-v^2\bigr)\frac{z}{\lambda } + \frac{1+v^2}{2 } \biggr ] \nonumber \\ & & \hspace{28ex } + \ ; \biggl [ \frac{\|{\cal d}\|^2 \lambda}{2 } + \frac{{\rm re}({\cal c}^{\cal d})}{m_b^2}(r+z-1 ) \biggr ] \bigl(1+v^2\bigr ) \nonumber \\ & & \hspace{28ex } + \ ; \frac{\|{\cal f}\|^2}{m_b^4 } \biggl ( \frac{4 r v^2 z}{\lambda } + \frac{1+v^2}{2 } \biggr ) + \nonumber \\ & & \hspace{28ex } + \ ; \frac{{\rm re}({\cal f}^{\cal g})}{m_b^2}\left[r - 1 + ( r+2z-1)v^2\right ] \biggr\}\end{aligned}\ ] ] and lepton forward - backward asymmetry @xmath162 where @xmath163 and @xmath164 . we remark that the @xmath165 term in the last line of eq . ( [ m2b2ksll ] ) contains @xmath81 for @xmath71 . since such a term is absent in eq . ( [ m2b2kll ] ) , there is no @xmath81 for @xmath56 from the sm or from the @xmath16 contributions under consideration . abazov _ et al . _ [ d0 collaboration ] , arxiv:1005.2757 [ hep - ex ] ; arxiv:1007.0395 [ hep - ex ] . s. oh and j. tandean , phys . lett . b * 697 * , 41 ( 2011 ) [ arxiv:1008.2153 [ hep - ph ] ] . gninenko and n.v . krasnikov , phys . b * 513 * , 119 ( 2001 ) [ arxiv : hep - ph/0102222 ] ; c. boehm , phys . rev . d * 70 * , 055007 ( 2004 ) [ arxiv : hep - ph/0405240 ] . d. hooper , phys . d * 75 * , 123001 ( 2007 ) [ arxiv : hep - ph/0701194 ] ; p. fayet , phys . d * 75 * , 115017 ( 2007 ) [ arxiv : hep - ph/0702176 ] . r. foot , x.g . he , h. lew , and r.r . volkas , phys . d * 50 * , 4571 ( 1994 ) [ arxiv : hep - ph/9401250 ] ; p.f . yin , j. liu , and s.h . zhu , phys . b * 679 * , 362 ( 2009 ) [ arxiv:0904.4644 [ hep - ph ] ] . he , j. tandean , and g. valencia , phys . b * 631 * , 100 ( 2005 ) [ arxiv : hep - ph/0509041 ] ; c.h . chen , c.q . geng , and c.w . kao , phys . b * 663 * , 400 ( 2008 ) [ arxiv:0708.0937 [ hep - ph ] ] ; s. oh and j. tandean , jhep * 1001 * , 022 ( 2010 ) [ arxiv:0910.2969 [ hep - ph ] ] . m. pospelov , phys . d * 80 * , 095002 ( 2009 ) [ arxiv:0811.1030 [ hep - ph ] ] . m. reece and l.t . wang , jhep * 0907 * , 051 ( 2009 ) [ arxiv:0904.1743 [ hep - ph ] ] ; b. aubert _ et al . _ [ babar collaboration ] , phys . * 93 * , 081802 ( 2004 ) [ arxiv : hep - ex/0404006 ] . m. iwasaki _ et al . _ [ belle collaboration ] , phys . d * 72 * , 092005 ( 2005 ) [ arxiv : hep - ex/0503044 ] . b. aubert _ et al . _ [ babar collaboration ] , phys . d * 79 * , 031102 ( 2009 ) [ arxiv:0804.4412 [ hep - ex ] ] ; phys . lett . * 102 * , 091803 ( 2009 ) [ arxiv:0807.4119 [ hep - ex ] ] . wei _ et al . _ [ belle collaboration ] , phys . rev . * 103 * , 171801 ( 2009 ) [ arxiv:0904.0770 [ hep - ex ] ] . t. aaltonen _ et al . _ [ cdf collaboration ] , arxiv:1101.1028 [ hep - ex ] . c. greub , a. ioannisian , and d. wyler , phys . b * 346 * , 149 ( 1995 ) [ arxiv : hep - ph/9408382 ] ; s. fukae , c.s . kim , t. morozumi , and t. yoshikawa , phys . d * 59 * , 074013 ( 1999 ) [ arxiv : hep - ph/9807254 ] ; t.m . aliev , c.s . kim , and y.g . kim , phys . d * 62 * , 014026 ( 2000 ) [ arxiv : hep - ph/9910501 ] ; c.w . chiang , r.h . li , and c.d . lu , arxiv:0911.2399 [ hep - ph ] ; q. chang , x.q . li , and y.d . yang , jhep * 1004 * , 052 ( 2010 ) [ arxiv:1002.2758 [ hep - ph ] ] ; a.k . alok , a. datta , a. dighe , m. duraisamy , d. ghosh , d. london , and s.u . sankar , arxiv:1008.2367 [ hep - ph ] . w. altmannshofer , p. ball , a. bharucha , a.j . buras , d.m . straub , and m. wick , jhep * 0901 * , 019 ( 2009 ) [ arxiv:0811.1214 [ hep - ph ] ] . deshpande and j. trampetic , phys . rev . lett . * 60 * , 2583 ( 1988 ) . b. grinstein , m.j . savage , and m.b . wise , nucl . b * 319 * , 271 ( 1989 ) ; w. jaus and d. wyler , phys . d * 41 * , 3405 ( 1990 ) ; a. ali , t. mannel , and t. morozumi , phys . b * 273 * , 505 ( 1991 ) . g. burdman , phys . d * 52 * , 6400 ( 1995 ) [ arxiv : hep - ph/9505352 ] ; a. ali , p. ball , l.t . handoko , and g. hiller , phys . d * 61 * , 074024 ( 2000 ) [ arxiv : hep - ph/9910221 ] ; m. beneke , t. feldmann , and d. seidel , nucl . b * 612 * , 25 ( 2001 ) [ arxiv : hep - ph/0106067 ] . f. kruger and j. matias , phys . d * 71 * , 094009 ( 2005 ) [ arxiv : hep - ph/0502060 ] . leveille , nucl . b * 137 * , 63 ( 1978 ) . t. huber , e. lunghi , m. misiak , and d. wyler , nucl . b * 740 * , 105 ( 2006 ) [ arxiv : hep - ph/0512066 ] ; t. huber , t. hurth , and e. lunghi , nucl . b * 802 * , 40 ( 2008 ) [ arxiv:0712.3009 [ hep - ph ] ] . k. nakamura _ [ particle data group ] , j. phys . g * 37 * , 075021 ( 2010 ) . c.h . chen and h.n . li , phys . d * 71 * , 114008 ( 2005 ) [ arxiv : hep - ph/0504020 ] . gao , c. meng , and k.t . chao , eur . j. a * 28 * , 361 ( 2006 ) [ arxiv : hep - ph/0606044 ] . p. ball and r. zwicky , phys . d * 71 * , 014015 ( 2005 ) [ arxiv : hep - ph/0406232 ] . p. ball and r. zwicky , phys . d * 71 * , 014029 ( 2005 ) [ arxiv : hep - ph/0412079 ] . c. bobeth , g. hiller and g. piranishvili , jhep * 0712 * , 040 ( 2007 ) [ arxiv:0709.4174 [ hep - ph ] ] ; c. bobeth , g. hiller and d. van dyk , jhep * 1007 * , 098 ( 2010 ) [ arxiv:1006.5013 [ hep - ph ] ] . f. jegerlehner and a. nyffeler , phys . rept . * 477 * , 1 ( 2009 ) [ arxiv:0902.3360 [ hep - ph ] ] . t. onogi , pos * lat2006 * , 017 ( 2006 ) [ arxiv : hep - lat/0610115 ] . geng and c.c . liu , j. phys . g * 29 * , 1103 ( 2003 ) [ arxiv : hep - ph/0303246 ] .	the anomalously large like - sign dimuon charge asymmetry in semileptonic @xmath0-hadron decays recently measured by the d0 collaboration may be hinting at the presence of @xmath1-violating new physics in the mixing of @xmath2 mesons . it has been suggested that the effect of a nonstandard spin-1 particle lighter than the @xmath0 quark with flavor - changing couplings to @xmath0 and @xmath3 quarks can reproduce the d0 result within its one - sigma range . here we explore the possibility that the new particle also couples to charged leptons @xmath4 and thus contributes to rare @xmath5 processes involving the leptons . we consider in particular constraints on its couplings from existing experimental data on the inclusive @xmath6 and exclusive @xmath7 decays , as well as the anomalous magnetic moments of the leptons . we find that there is parameter space of the particle that is allowed by the current data . future measurements of these @xmath8 transitions and rare decays of the @xmath2 meson , such as @xmath9 and @xmath10 , at lhcb and next - generation @xmath8 factories can probe its presence or couplings more stringently .
many years ago kubo @xcite has proven the general sum rule for diagonal dynamic ( frequency dependent ) conductivity @xmath1 , which holds for any system of charged particles irrespective of interactions , temperature or statistics . this sum rule is usually written as : @xmath2 where @xmath3 specifies the type of charged particles , @xmath4 and @xmath5 are the respective densities and charges . for the system of electrons in a solid eq . ( [ kub_sum ] ) takes the form : @xmath6 where @xmath7 is the density of electrons and @xmath8 is the plasma frequency . however , in any real experiment we are not dealing with an infinite range of frequencies . if one considers electrons in a crystal and limits himself to the electrons in a particular ( e.g. conduction ) band , neglecting interband transitions , the general sum rule ( [ f_sum_el ] ) reduces to the single band sum rule of kubo @xcite : @xmath9 where @xmath10 is the bare dispersion as defined by the effective single band hamiltonian , while @xmath11 is the momentum distribution function ( occupation number ) , which is in general defined by the _ interacting _ retarded electronic green s function @xmath12 @xcite : @xmath13 where @xmath14 is the usual fermi distribution . in eq . ( [ 1 ] ) @xmath15 represents an ultraviolet cut - off , frequency , which is assumed to be larger than the bandwidth of the low energy band , but smaller than the gap to other bands . the function @xmath16 accounts for the cut - off dependence , which arises from the presence of drude spectral weight beyond @xmath15 @xcite and is unity if we formally set @xmath15 to infinity while ignoring the interband transitions . although the general sum rule is certainly preserved , the optical integral @xmath17 is not a conserved quantity since both @xmath16 @xcite and @xmath11 @xcite depend on temperature @xmath18 , and also on details of interactions @xcite . this dependence of @xmath19 on @xmath18 and other parameters of the system under study has been termed the `` sum rule violation '' . it was actively studied experimentally , especially in cuprates , where pronounced anomalies were observed both in c - axis and in - plane conductivity both in normal and superconducting states @xcite . the finite cut - off effects were extensively studied in several theoretical papers on the @xmath18 dependence of the optical integral @xcite . in refs . @xcite , the effect of the cut - off was considered in the context of electrons coupled to phonons . in a simple drude model , @xmath20 and the sum rule can only be `` violated '' due to the presence of @xmath16 . integrating over @xmath21 and expanding for @xmath22 , one can see that @xmath23 for infinite cut - off , @xmath24 and @xmath25 , but for a finite cut - off @xmath16 contains the term proportional to @xmath26 . if @xmath27 changes with @xmath18 , then one obtains a sum rule `` violation '' even if @xmath28 is @xmath18 independent @xcite . other aspects of cut - off dependence were discussed recently in detail in ref . @xcite . in the present study we neglect the cut - off effects in optical integral from the outset . our goal is to study @xmath19 dependence on @xmath18 and a number of interaction parameters , determining the electronic properties of strongly correlated systems , such as cuprates . in this context we shall discuss the problem of a possible `` violation '' of the optical sum rule in the normal ( non superconducting ) state of strongly correlated electronic systems , using our recently proposed dmft+@xmath0 approach @xcite , as applied to dynamic conductivity in two typical models of such systems : the `` hot spot '' model of the pseudogap state @xcite and disordered anderson hubbard model @xcite . our aim is both to check the consistency of dmft+@xmath0 approach as applied to calculations of optical conductivity , as well as to demonstrate rather important dependences of the optical integral @xmath19 not only on @xmath18 , but also on such important characteristics as pseudogap width , disorder and correlation strength , making ( single band ) sum rule `` violation '' rather ubiquitous for any strongly correlated system , even neglecting the cut - off effects . characteristic feature of the general sum rule as expressed by eqs . ( [ 1 ] ) , ( [ dist_fun ] ) is that the integral @xmath19 over frequency in the l.h.s . is calculated via two - particle property ( dynamic conductivity , determined by two - particle green s function , in general , with appropriate vertex corrections ) , while the r.h.s . is determined by the single - particle characteristics , such as bare dispersion and occupation number ( [ dist_fun ] ) ( determined by a single - particle green s function ) . thus , checking the validity of this sum rule , we are in fact thoroughly checking the consistency of any theoretical approach , used in our model calculations . our generalized dynamical mean field theory ( dmft+@xmath0 ) approach @xcite , supplying the standard dynamical mean field theory ( dmft ) @xcite with an additional `` external '' self - energy @xmath0 ( due to any kind of interaction outside the scope of dmft , which is exact only in infinite dimensions ) , provides an effective method to calculate both single- and two - particle properties @xcite . the consistency check of this new approach is obviously of great interest by itself . we shall also see , that it gives a kind of a new insight in the problem of sum - rule `` violation '' . pseudogap phenomena in strongly correlated systems have essential spatial length scale dependence @xcite . to merge pseudogap physics and strong electron correlations we have generalized the dynamical - mean field theory @xcite by inclusion of the dependence on correlation length of pseudogap fluctuations via additional ( momentum dependent ) self - energy @xmath29 . this self - energy @xmath29 describes non - local dynamical correlations induced either by short - ranged collective sdw - like antiferromagnetic spin or cdw - like charge fluctuations @xcite . to calculate @xmath29 in two - dimensional `` hot spots '' model @xcite for an electron moving in the random field of pseudogap fluctuations ( considered to be static and gaussian ) with dominant scattering momentum transfers of the order of characteristic vector @xmath30 ( @xmath31 is the lattice spacing ) , we used @xcite the recursion procedure proposed in refs . @xcite , which is controlled by two main physical characteristics of the pseudogap state : @xmath32 ( pseudogap amplitude ) , which characterizes the energy scale of the pseudogap , and @xmath33 the inverse correlation length of short range sdw ( cdw ) fluctuations . both parameters @xmath32 and @xmath34 , determining pseudogap behavior , can in principle be calculated from the microscopic model at hand @xcite . weakly doped one - band hubbard model with repulsive coulomb interaction @xmath35 on a square lattice with nearest and next nearest neighbour hopping was numerically investigated within this generalized dmft+@xmath0 self - consistent approach , as described in detail in refs . @xcite . briefly , the dmft+@xmath0 self - consistent loop looks like as follows . first we guess some initial local ( dmft ) electron self - energy @xmath36 . second we compute the @xmath37-dependent `` external '' self - energy @xmath29 which is in general case a functional of @xmath36 . then neglecting interference effects between the self - energies ( which in fact is the major assumption of our approach ) we can set up and solve the lattice problem of dmft @xcite . finally we define effective anderson single impurity problem which is to be solved by any `` impurity solver '' ( we mostly use numerical renormalization group - nrg ) to close dmft+@xmath0 equations . the additive form of self - energy is in fact an advantage of our approach @xcite . it allows one to preserve the set of self - consistent equations of standard dmft @xcite . however there are two distinctions from conventional dmft . during each dmft iteration we recalculate corresponding * p-dependent self - energy @xmath38)$ ] via an approximate scheme , taking into account interactions with collective modes or order parameter fluctuations , and the local green function @xmath39 is `` dressed '' by @xmath29 at each step . when input and output green s functions ( or self - energies ) converge to each other ( with prescribed accuracy ) we consider the obtained solution to be selfconsistent . physically it corresponds to the account of some `` external '' ( e.g. pseudogap ) fluctuations , characterized by an important length scale @xmath34 , into fermionic `` bath '' surrounding the effective anderson impurity of the usual dmft . both cases of strongly correlated metals and doped mott insulators were considered @xcite . energy dispersions , quasiparticle damping , spectral functions and arpes spectra calculated within dmft+@xmath0 , all show a pseudogap effects close to the fermi level of quasiparticle band . in ref . @xcite this dmft+@xmath0 procedure was generalized to calculate two - particle properties , such as dynamic conductivity , using previously developed recursion procedure for vertex corrections due to pseudogap fluctuations @xcite , producing typical pseudogap anomalies of optical conductivity and dependence of these anomalies on correlation strength @xmath35 . below we use the approach of ref . @xcite to investigate the sum - rule in `` hot spots '' model . to calculate optical integral @xmath19 we have just used the conductivity data of ref . @xcite ( extended to a wider frequency range needed to calculate @xmath19 ) , while the r.h.s of ( [ 1 ] ) was recalculated , using recursion relations for @xmath29 and the whole self - consistency dmft+@xmath0 loop . all calculations have been done for a tight - binding `` bare '' spectrum on the square lattice , with the nearest neighbor transfer integral @xmath40 and next nearest neighbor transfer integral @xmath41 . in fig . [ fig1 ] we present our typical data for the real part of conductivity ( @xmath42 , @xmath43 ev , band filling @xmath44 , temperature @xmath45 ) for different values of hubbard interaction @xmath46 and fixed pseudogap amplitude @xmath47 ( correlation length @xmath48 ) . it is obvious from these data , that optical integral @xmath19 is different for all of these curves , actually its value drops with the growth of @xmath35 ( along with damping of pseudogap anomalies @xcite ) . however , the single band optical sum - rule ( [ 1 ] ) is satisfied within our numerical accuracy , as seen from table i. the small `` deficiency '' in the values of @xmath19 in table i is naturally due to a finite frequency integration interval over conductivity data of fig . [ fig1 ] . , @xmath43 ev , @xmath45 ) in dmft+@xmath49 approximation @xmath35 dependence . band filling @xmath44 , pseudogap amplitude @xmath47 , correlation length @xmath48 . conductivity is given in units of @xmath50 . ] [ tab1 ] .single - band optical sum rule check in the `` hot - spots '' model . @xmath35 - dependence . optical integral in units of @xmath51 . [ cols="^,^,^",options="header " , ] again , to study the details of this sum - rule `` violation '' , i.e. the dependence of @xmath19 on the parameters of anderson - hubbard model , we performed detailed calculations of its dependences on the temperature @xmath18 , disorder amplitude @xmath32 and correlation strength @xmath35 . some of the results are presented in figs . [ fig7 ] [ fig9 ] . in fig . [ fig7 ] we show the dependence of normalized optical integral on hubbard @xmath35 , for different degrees of disorder ( both for strongly disordered metal and correlated anderson insulator ) . it is seen that in all cases the growth of correlation strength leads to rather sharp drop of @xmath19 in metallic state , which becames mush slower in mott insulator . ( 1,2 strongly disordered metal , 3 correlated anderson insulator ) . ] in fig . [ fig8 ] we present similar dependences on disorder strength @xmath32 . in metallic state optical integral generally drops with the growth of disorder , while an opposite behavior is observed if we start from mott insulator ( both obtained with the growth of @xmath35 from metallic state and under diminishing @xmath35 in hysteresis region of the phase diagram @xcite ) . note the absence of any significant changes in the immediate vicinity of critical disorder @xmath52 , corresponding to anderson metal insulator transition . at the same time it should be noted that the most significant growth of the optical integral takes place as the system transforms into disorder induced metallic state , obtained from mott insulator , as observed in ref.@xcite . . lines 1,2,3 correlated metal , transforming into anderson insulator . line 4 mott insulator state obtained with the growth of @xmath35 from correlated metal , line 5 mott insulator obtained with diminishing @xmath35 in hysteresis region of the phase diagram . ] in fig . [ fig9 ] we show the temperature dependence of the normalized optical integral , for different degrees of disorder . in anderson hubbard model it appears to be significantly stronger , than in `` hot spots '' model ( see above ) , and decreases with the growth of disorder . moreover , while in relatively weakly correlated state it is qualitatively the same optical integral diminishes with the growth of @xmath18 , it actually grows in disordered mott insulator , as seen from line 3 at the insert in fig . , line 3 here corresponds to disordered mott insulator . ] again , as in the case of the pseudogap `` hot spots '' model , these results for anderson hubbard model clearly demonstrate the value of the optical integral is not universal and depends on all the major parameters of the model and , so the single band optical sum rule is strongly `` violated '' . based on dmft+@xmath0 approach , we have studied the single band optical sum rule for two typical strongly correlated systems , which are outside the scope of the standard dmft : ( i ) the `` hot spots '' model of the pseudogap state , which takes into account important nonlocal correlations due to afm(cdw ) short - range order fluctuations and ( ii ) anderson - hubbard model , which includes strong disorder effects , leading to disorder induced metal - insulator ( anderson ) transition , alongside with mott transition . we have explicitly demonstrated that the single band optical sum rule ( [ 1 ] ) is satisfied for both models , confirming the self - consistency of dmft+@xmath0 approach for calculation of two - particle properties . however , the optical integral @xmath53 , entering the single band sum rule ( [ 1 ] ) is non universal and depends on the parameters of the model under consideration . most of the previous studies addressed its ( relatively weak ) temperature dependence . here we have analyzed dependences on essential parameters of our models , showing that these may lead to rather strong `` violations '' of the optical sum rule . as most of the parameters under discussion may be varied in different kinds of experiments , these dependences should be taken into account in the analysis of optical experiments on strongly correlated systems . we thank thomas pruschke for providing us the nrg code . this work is supported by rfbr grants 08 - 02 - 00021 , 08 - 02 - 00712 , ras programs `` quantum macrophysics '' and `` strongly correlated electrons in semiconductors , metals , superconductors and magnetic materials '' . in is also supported the grant of the president of russia mk.2242.2007.2 and russian science support foundation . a. f. santander - syro , r. p. s. m. lobo , n. bontemps , z. konstantinovic , z. z. li and h. raffy , europhys . lett . 62 * , 568 ( 2003 ) ; a. f. santander - syro , r. p. s. m. lobo , n. bontemps , w. lopera , d. girata , z. konstantinovic , z. z. li and h. raffy , phys . rev . b * 70 * , 134504 ( 2004 ) .	we discuss the problem of a possible `` violation '' of the optical sum rule in the normal ( non superconducting ) state of strongly correlated electronic systems , using our recently proposed dmft+@xmath0 approach , applied to two typical models : the `` hot spot '' model of the pseudogap state and disordered anderson hubbard model . we explicitly demonstrate that the general kubo single band sum rule is satisfied for both models . however , the optical integral itself is in general dependent on temperature and characteristic parameters , such as pseudogap width , correlation strength and disorder scattering , leading to effective `` violation '' of the optical sum rule , which may be observed in the experiments .
in the currently favored cold dark matter cosmology , cosmic structures are built up of dark matter halos . the formation of halos is hierarchical , in that small halos form first and subsequently merge to form bigger ones . after mergers , smaller halos become the subhalos of the more massive host halo . galaxies are thought to form in the centers of halos @xcite , and most of them become satellites when their host halos merge with a more massive ones . after mergers , the motions of subhalos / satellites are mainly dominated by the gravitational potential of the hosts , and in principle they can be well traced using numerical simulations ( e.g. , * ? ? ? * ) or analytical models ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? it was found from @xmath0-body simulations ( e.g. , * ? ? ? * ) that dark matter halos are not spherical , but rather they are tri - axial . the non - spherical shapes are related to the formation history of halos , which happens preferentially along filaments . as the assembly history of a halo should be imprinted in the phase - space distribution of its satellite galaxies , observational attempts have been made to infer halo shapes using satellite distributions . though the task is challenging , progress has been made using the distribution of stellar velocity @xcite , satellite tidal streams @xcite , and gravitational lensing ( e.g. , * ? ? ? * ; * ? ? ? * ) . the measurements of halo shapes from satellite kinematics or weak lensing rely on an estimate of the host potential . in fact , more useful insight can also be gained from the pure spatial distribution of satellites . most studies have focused on how satellites are distributed with respect to the shape of the central galaxy , known as galaxy alignment . the observational study of the alignment of galaxies has a long history ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? based on large galaxy surveys such as 2dfgrs and sloan digital sky survey , there is general agreement that the distribution of satellite galaxies is typically along the major axis of the central galaxy . moreover the alignment signal depends on galaxy color . @xcite found a stronger alignment signal for red central galaxies or red satellites . such an effect is also seen at high redshift @xcite . a rough idea to explain the observed galaxy alignment is that if satellite galaxies follow the distribution of dark matter , and the central galaxy also shares a shape similar to the host halo , then the non - spherical nature of dark matter halos naturally produces an alignment effect . in fact , many theoretical works follow this idea @xcite . however , most studies over - predict the alignment signal when the central galaxy is assumed to follow the same shape as the whole host halo . furthermore , most studies are unable to reproduce the alignment dependence on galaxy color unless a dependence of central galaxy alignment with the host halo is assumed @xcite . the main difficulty faced by theoretical studies of galaxy alignment is how to assign the shape of the central galaxy . the most natural way is to use hydro - simulations including physics governing galaxy formation , such as gas cooling , star formation and feedback . unfortunately , current simulations are typically unable to produce a galaxy population which matches observational data ( see , however , vogelsberger et al . 2013 and references therein ) . in this paper , we revisit the problem of galaxy alignment using an @xmath0-body simulation which allows good statistics with a large number of massive halos . since the simulation does not include models for galaxy formation , we instead use the subhalos as tracers of satellite galaxies . for central galaxies , we follow previous studies and assume that the shape of the central galaxy follows the shape of its host halo . in our study , the halo shape is measured at different iso - density surfaces using a method different from previous studies . in an upcoming paper , we will present the results using smoothed particle hydrodynamics(sph ) simulations performed with gadget-2 ( paper ii , in preparation ) . in addition to the overall alignment signal , we investigate the dependence of alignment on the accretion and formation history of subhalos . since red satellites have stronger alignment with central galaxies ( e.g , * ? ? ? * ) , and in general red satellites are accreted at earlier times , it is natural to ask whether the stronger alignment of red satellites is already set before their accretion into the host halo or it is an evolution effect that red satellites follow more closely the shape of dark matter halo after accretion . to study this question , we study the alignment of subhalos as a function of their formation and accretion time . to probe if the color dependence is imprinted in the large - scale environment or is due to an evolution effect , we also study the alignment of neighboring halos which are within one to a few virial radii from the host halos . the paper is organized as follows . in section [ cha : axis_define ] , we briefly describe the simulation and how we determine the shapes of dark matter halos . in section [ cha : angular_distri ] and [ cha : mass_dependence ] , we show the alignment of subhalos and its dependence on the subhalo mass . in section [ cha : time_dependence ] we investigate if the alignment signal is dependent on the accretion or formation time of the subhalo , and present the results of alignment of neighboring halos . we summarize and briefly discuss our results in section [ cha : concl ] . the cosmological simulation used in this paper was performed using the massive parallel code gadget-2 @xcite and evolved from redshift @xmath1 to the present epoch in a cubic box of @xmath2 with @xmath3 dark matter particles , assuming a flat @xmath4 `` concordance '' cosmology with @xmath5 , @xmath6 , @xmath7 . a plummer softening length of @xmath8 was adopted and each dark matter particle has a mass of about @xmath9 . the simulation has been widely used in previous studies ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) and the readers are referred to these papers for more details of the simulation . in total 62 snapshots between @xmath10 and @xmath11 have been used to construct the merger history of halos and subhalos . all halos were found using the standard friends - of - friends algorithm with a linking length of 0.2 times the mean particle separation , while subhalos were found using the subfind routine @xcite . we select host halos as those with a mass larger than @xmath12 ( to mimic the data of yang et al . in our simulation there are about 2000 host halos in total , at redshift @xmath11 . then we re - select host halos with @xmath13 of inner axes less than @xmath14 . here @xmath13 is the error on determining the axis given by following equation @xmath15 where @xmath0 is the number of particles used and @xmath16 is the relevant axis ratio : @xmath17 for the major axis , @xmath18 for the minor axis , and max@xmath19 for the intermediate axis ( here @xmath20,@xmath21,@xmath22 are the lengths of the axes with @xmath23)(cf . * ? ? ? there are 1604 halos that survive our two cuts . in table . [ tab:1 ] we list the number of host halos in different mass bins . unless otherwise specified , the following analysis will use the same halo catalogue as described in table . [ tab:1 ] . there are various methods to determine the shape of halos and they differ in details @xcite . in the most widely used method , the axes of a dark matter halo are calculated by its overall inertia tensor defined as @xmath24 where @xmath25 is the distance of particle @xmath26 from the halo center in dimension @xmath27 and the summation is over all particles in a halo . in the three - dimensional(3d ) configuration space @xmath28 . the three eigenvectors of @xmath29 define the orientation of the three axes of a halo , and the eigenvalues are related to the rms along the corresponding eigenvectors . the halo shape determined in this way should be closely correlated with the large scale environment of the halo , for example , external tidal field or large nearby filaments @xcite . an improved version of the above method uses the reduced inertia tensor defined as @xmath30 which can alleviate the contamination of substructures @xcite . here @xmath31 is the distance of the @xmath26th particle from the halo center . in this work , when we calculate the axes for the whole halo , we use this reduced inertia tensor . since dark matter halos form hierarchically , the inertia tensors of the whole halo might not precisely describe its internal orientation or that of the central galaxy residing in it . therefore , following @xcite we use different iso - density surfaces to determine the axes of a halo . the local density at a particle position is calculated by using the smoothed - particle - hydrodynamics ( sph ) method . we select those particles with local density within a narrow bin as an input list for the group finder and apply the fof method again to find subhalos using a smaller linking length . to exclude particles from subhalos and eliminate their influence on the shape determination of the host halo , only those particles within the largest sub - group are retained . finally the iso - density surfaces of a halo are built up and then the axes of them can be calculated by fitting with a tri - axial ellipsoid . following @xcite , we determine the axes for five different iso - density surfaces , for which the @xmath26th bin has local density of @xmath32 ( @xmath33 ) times the cosmic critical density . the mean radius of each elliptical surface is approximately 0.6 , 0.4 , 0.25 , 0.12 , and 0.06 of the virial radius respectively . the orientation of the five iso - density surfaces represents the orientation of different parts of a halo , from the most inner part to the outer part . for convenience , in the following content we will call the axis of the most outer part of the halo as `` _ _ outer axis__''(@xmath26=1 ) , the axis of the most inner part of the halo as `` _ _ inner axis__''(@xmath26=5 ) and the axis of intermediate part as `` _ _ intermediate axis__''(@xmath26=3 ) . intuitively , the shape of the outer iso - density ellipsoids should be closely correlated to that of the whole halo . in figure [ axis_diverse ] we plot the alignment between the axes of the iso - density ellipsoid and that of the halo inertia tensor . it is not surprising to find that the outer axes align strongly with the axes derived from the inertia tensor of the whole halo , and that the strength of alignment for the inner axes becomes weaker ( but still significant ) . interestingly but not unusual , it again commendably demonstrates the hierarchical structure of dark matter halos . the results also show that there is non - negligible mis - alignment between the halo major ` _ _ inner axis _ _ ' and major ` _ _ outer axis _ _ ' , which may imply a mild alignment signal between bcg orientation and satellite distribution , as presented in the next section . the alignment between the ` inner ' and ` outer ' axes is stronger with increasing halo mass , with average alignment angles of @xmath34 , for host halos with @xmath35 , @xmath36 , and @xmath37 , respectively . although determining axes will suffer from the limited number of particles especially for small halos , our strategy which excludes host halos with less accuracy ( @xmath38 ) can alleviate this problem and making the above results more reliable . in this section , we present the spatial distribution of subhalos and the neighboring halos from our simulation . for each host halo , we place a central galaxy at its center whose shape is set to follow the shape of the host dark matter halo ( either inner , intermediate or outer axis ) . the neighboring halos are defined as individual halos which are less massive than the host halo and reside within the range of @xmath39 from the center of the host halo . to obtain more robust results , we limit subhalos to being more massive than @xmath40 ( containing more than 30 particles ) and neighboring halos to being more massive than @xmath41 ( containing more than 100 particles ) . table [ tab:1 ] lists the number of subhalos and neighboring halos in each mass bin . .number of host halos in different mass bins , and subhalos and neighboring halos with different masses of their host or of themselves . [ cols="^,^,^,^",options="header " , ] [ tab:1 ] host halos with inner axes less accurate than @xmath14 and their subhalos and neighboring halos are excluded from this table . note that we set a minimum number of 30 and 100 particles for subhalos and neighboring halos , so @xmath42 means about @xmath43 for subhalos and @xmath44 for neighboring halos . the alignment signal is described as the distribution of angular separation between the axis of the host halo and the connecting line between the centers of the host halo and each subhalo / neighboring halo . the distribution is described as , @xmath45 where @xmath46 is the count of subhalos ( or neighboring halos ) in bin @xmath47 , while @xmath48 is the average count of random samples in the same bin . so @xmath49 means the absence of any alignment , while @xmath50 at small @xmath47 implies a distribution with a preferred alignment along the major axis . to compare with the observational results of yang et al . ( 2006 ) , here we performed the analysis in two - dimensional space . we first calculate the axes in 3d space , and then project them onto the @xmath51 plane . the angular position is then obtained for every subhalo / neighboring halo using its projected position in the same plane . to generate the random sample , for each host halo containing @xmath0 subhalos ) , we produce @xmath0 subhalos in 3d space with a spherical random distribution . in principle , the produced random sample should share the same radial distribution of the real subhalos in each host halo . however , as our calculation is only dependent on the angular separation , our results are not affected by this requirement . in figure [ ang_comp ] we show the alignment along the major axes of central galaxies . the left , middle and right panels show the results assuming that the central galaxies follow the shape of their host halos defined at the outer , intermediate and inner axes . in each panel the dotted and dashed lines are for subhalos and neighboring halos , respectively . for comparison with other results , we also plot the observational work of @xcite , and the theoretical one by @xcite as circles and triangles in each panel . figure [ ang_comp ] shows that both subhalos and neighboring halos are aligned with the major axes of central galaxies . the signal is stronger than the data ( solid points ) if the central galaxy follows the shape of the dark matter halo at the outer or intermediate axes ( left and middle panels ) . our results are consistent with those of kang et al . ( 2007 ) , who used satellite galaxies from their semi - analytic models(sams ) and have found strong alignment if central galaxy follows the shape of the whole dark matter halo . @xcite pointed out that some observational effects , such as the flux - limit , redshift - space distortions and the galaxy group finder , could cause a shallowing of the alignment signal , but these would not be enough to reconcile the simulations with observations . they further proposed that the observed alignment signal could be reproduced if the spin of the central galaxy aligns with that of dark matter halo . our result shows that there may be another way to remove such a discrepancy with observations . as showed in the right panel of figure [ ang_comp ] , the predicted alignment signal is close to the data if the shape of the central galaxy follows the shape defined at the inner halo region . this is also consistent with the results of @xcite in which they also found only a weak misalignment if using the inner halo to define the shape of central galaxy , although they use a different method to measure the halo shape . it is also found from figure [ ang_comp ] that the alignment of neighboring halos is similar to that of the subhalos . there seems to be a slight evolution effect that subhalos alignment is stronger if the central galaxy follows the outer shape of dark matter halo ( left panel ) . however , the predicted alignment signal is too strong in this case , thus it can not account for the observed color dependence of satellite alignment . if the central galaxy follows the shape of the inner axis , and blue satellite galaxies are recently accreted from neighboring halos , then their alignment should be identical to that of the red satellites ( subhalos in our case ) . to explain the observed color dependence , it will require that either blue ( red ) satellites are not random sample of neighboring halos ( subhalos ) . we will later investigate if the color dependence arises from the assembly bias of satellite galaxies . the right panel of figure [ ang_comp ] shows that if the central galaxy follows the inner shape of the dark matter halo , then the predicted alignment signal is more or less consistent with the data . however , it is not implied that this is the only way to reproduce the observed alignment signal . it indicates that some degree of misalignment between the central galaxy and the overall shape of dark matter halo has to be assumed . for example , @xcite and @xcite both have found that if the minor axis of central galaxies follows the angular momentum of the dark matter halo , then the alignment of satellites is also closer to the observational data . the observed alignment effect can also be reproduced if the central galaxy follows the overall shape of the dark matter halo , but with some distribution . figure [ ang_comp_re ] shows the results when the angle between the major axes of the central galaxy and the host halo follows a gaussian distribution with a mean of @xmath52 and a standard deviation of @xmath53 . it is seen that the observed alignment signal is well reproduced . some previous works find results consistent with this : @xcite found a misalignment between the angular momentum vector and the minor axis , with a mean value of about @xmath53 ; @xcite found that there is a median angle of @xmath53 between the angular momentum vectors of inner ( @xmath54 ) part and the whole halo . in fact , there is observational evidence for a mis - alignment angle of about @xmath55 between the major axis of brightest central galaxies ( bcg ) and that of their dark matter halos inferred from the distribution of satellite galaxies@xcite . in addition , a larger mis - alignment angle of about @xmath56 was also suggested from gravitational shear - intrinsic ellipticity correlation of luminous red galaxies @xcite . so far , there is no definitive way to assign a shape to the central galaxy . the most promising way should be the one which can reproduce not only the overall alignment of satellites , but their dependence on galaxy properties . studies based on @xmath0-body simulations and semi - analytical models for galaxy formation @xcite have made some progress to achieve this goal . however , these studies still face problems . first , the predicted total alignment signal is still stronger than observations . second , the color dependence of satellites in the models themselves are not correctly reproduced ( however , see * ? ? ? third , it is difficult to interpret the contamination of interlopers in observations . in our simulations we have no galaxies , only subhalos . thus we are unable to explore the dependence of alignment on galaxy properties . however , as galaxy properties are closely related to the halo ( subhalo ) mass and accretion time @xcite , we could learn the shape of the central galaxy from the dependence of alignment on the mass of host halos and subhalos . in particular , we want to study if the alignment is dependent on the accretion / formation time of subhalos and will present such results in the following sections . observationally , it was found that the alignment of satellites is stronger in massive halos , and it is even stronger for red satellites in massive red central galaxies ( c.f . such a mass dependence should be imprinted in the alignment of subhalos / neighboring halos . in this section , we investigate the dependence of alignment on the host halo mass and the mass of subhalos / neighboring halos themselves . different from the previous section , here we give the results in 3d space since it gives a stronger alignment signal in real space . figure [ ang_host ] shows the effects of host halo mass on the alignment of subhalos and neighboring halos . note that the error - bars are for the @xmath57 standard deviation of the mean throughout the paper . it is found that more massive host halos have a stronger alignment signal . though subhalos / neighboring halos preferentially align along the `` _ _ outer axis _ _ '' , it seems that dependence on halo mass is stronger for the alignment along the `` _ _ inner axis _ _ '' . this reflects the dependence of central - to - outer alignment strength with halo mass . given the fact that more massive bcgs have redder colors , the results for `` _ _ inner axis _ _ '' are in agreement with observational results of stronger alignment for red central galaxies than blue central galaxies @xcite , as well as the results from a semi - analytical model ( e.g. , * ? ? ? * ) . in figure [ ang_self ] we show the alignment of subhalos / neighboring halos with dependence on their own mass . contrary to the dependence on the host halo mass , the dependence is quite weak on the mass of subhalos / neighboring halos themselves . only those most massive subhalos / neighboring halos show a little stronger alignment signal along the `` _ _ outer axis _ _ '' of their host halo than less massive ones . we also find that if the mass of subhalos at accretion is used instead , the results are also very similar . assuming that massive subhalos or neighboring halos harbor redder satellites and the major axis of the central galaxy can be represented by the `` inner axis '' of host halos , the results of alignment along the `` inner axis '' ( left panels in figure [ ang_self ] ) can not explain the observational trend that red satellites are better aligned than the blue counterparts @xcite . our results above clearly show that the neighboring halos are preferentially distributed along the major outer axes of the host halos . this is easy to understand as the neighboring halos define the local large - scale environment of the host halo which dominate the direction of mass accretion . also found is that more massive host halos are stronger aligned with the neighboring halos ( lower right panel of figure 4 ) and the most massive neighboring halos have stronger alignment with the host halos ( lower right panel of figure 5 ) . these results are consistent with the picture that more massive host halos are connected with stronger filaments where more massive subhalos are embedded and accreted along them . intuitively one would speculate that galaxies which formed earlier ( such that their host halos formed earlier ) should have a redder color , therefore there could be helpful clues for the dependence of alignment signal on galaxy color from the formation histories of halos . on the other hand , being accreted is an important event for a subhalo . one would expect that subhalos accreted earlier will spend more time within host halos and suffer longer from tidal stripping as well as dynamical friction and thus may lose memory of initial accretion positions(as @xcite indicate , these initial accretion positions are more likely to be close to the major axis of the host halo ) , consequently attenuating the alignment signal . however , it is also possible that earlier accreted satellites will experience more gravitational effects from the host , thus following the shape of the host more closely and predicting stronger alignment @xcite . these two effects will compete with each other , and finally determine the alignment of satellite galaxies . in this section , we focus on the relation between subhalos ( or neighboring halos ) angular alignment and their assembly history . here we define two key time points for the formation and evolution of subhalos . a subhalo was once an independent halo before it is accreted into a virialized halo nowadays . we define the time when the progenitor of a subhalo was accreted into a large halo as `` _ accretion time _ '' . as subhalos often cross the virial radius of the host halos for a few times ( e.g. * ? ? ? * ) , we use the first time it cross the virial radius as the accretion time . before being accreted , the progenitor halo evolves as an individual halo , and its mass grows almost monotonically with time @xcite . the time when this progenitor halo has assembled half of its mass at the accretion time is defined as its `` _ formation time _ '' . for neighboring halos , there is no accretion time to apply since it has not been accreted by the central halo . therefore we define their `` _ formation time _ '' as the time when they accrete half of their mass at @xmath11 and consider the effect of their `` _ formation time _ '' only . first , we show the distribution of accretion and formation redshifts of all subhalos / neighboring halos in our sample in figure [ time_distri ] . the accretion redshifts mostly range from @xmath11 to @xmath58 . one can find that most subhalos are accreted to their host halos at redshift @xmath59 . this is consistent with previous work @xcite . note that here we only analyze the accretion redshifts for those subhalos which finally survive at @xmath11 in their host halos . in some previous works ( e.g. , * ? ? ? * ) , the statistic of accretion time based on merger trees including un - resolved subhalos are different , which have more accretion events at higher redshift . also in our simulation there are some subhalos which can not be traced back to their individual halo progenitors due to the resolution of the simulation . such kind of subhalo accounts for approximately @xmath60 of all subhalos . the formation time distribution of neighboring halos is consistent with the prediction for low mass halos ( e.g. , * ? ? ? figure [ accretion ] shows the angular distribution of subhalos with different accreted redshifts . we choose the 20 accreted earliest , the 20 mid - period from the whole samples and plot their angular distribution along the major axes of their host halos . note that the use of 20% is arbitrary . we have also tried using choosing criterion of 25% , 33% and 40% ( not shown in the plots ) . as might be expected , it turns out that the difference between early accreted sub - halos and late accreted ones becomes less distinct when the samples include more intermediate accretion redshifts . the upper panel of figure [ accretion ] shows that the alignment signals along the inner axes of host halos are almost identical ( within the 1@xmath61 errors ) for the three samples with different formation times . the lower panel shows that later accreted subhalos have significantly stronger alignment with the outer axes of the host halos . from models of galaxy formation ( e.g. , * * ; * ? ? ? * ) , it is expected that early accreted subhalos should host red satellite galaxies . the observed stronger alignment of red satellites implies that early accreted subhalos should have strong alignment . indeed such a dependence is seen from the analysis of @xcite . however , such an expectation is not seen , and is in fact contrary to what we see from figure [ accretion ] in our work . to understand this puzzle , we have to note that our results are based on subhalos only , while observations or the analysis of @xcite were based on galaxies . to address the dependence of alignment on accretion time of subhalos , we plot the radial distance of subhalos with different accretion time in figure [ radius_distri ] . the solid line shows that early accreted subhalos stay in the inner host halo , but the late accreted subhalos stay near the virial radius of the host . it is known that subhalos orbiting the inner host halo will suffer from strong tidal forces , and thus are more likely to be disrupted . this effect is more efficient for subhalos with more ` radial ' orbits , as they have more chance to pass the host center , thus being more exposed to the strong tidal force and being disrupted more efficiently ( e.g. , * ? ? ? thus only those with more ` circular ' orbits will likely survive . those subhalos with more ` circular ' orbits will naturally produce a more isotropic distribution . for recently accreted subhalos , the effects of tidal forces are still not strong enough to disrupt them as most of them are still orbiting at the outer host halo . although the analysis of @xcite made use of an @xmath0-body simulation , they do include model galaxies given by the semi - analytical model of @xcite . the ` galaxies ' are traced by subhalos and unresolved subhalos ( those disrupted ones ) . thus the main difference between @xcite and ours is that they include the so - called ` orphan ' galaxies @xcite which are not associated with any resolved subhalos , but their positions can be traced using the most - bound particles from the previous resolved subhalos . these orphan galaxies can also be resolved in hydrodynamical simulation with gas physics included as the condensation of baryons boost the density distribution in subhalos , making it more resistant to tidal disruptions ( e.g. , * ? ? ? the ` orphan ' galaxies are mainly from early accreted subhalos , and most of them are staying in the inner host halos , contributing significantly to the total galaxy populations @xcite . on one hand , these orphan galaxies will include those subhalos with more ` radial ' orbits , leading to a more aligned distribution . on the other hand , orphan galaxies have orbited in the inner halo for longer time , and they will follow the shape of halo more closely . under the assumption that central galaxies are also shaped dominantly by the inner host halo , it will predict that red satellites align more strong with central galaxies than the blue ones which are dominated by recent accreted subhalos at the outer skirts of host halos . this will be fully consistent with the observations . now we investigate the dependence of subhalos alignment on their formation time in figure [ formation ] . similar to the results shown in figure [ accretion ] the dependence of alignment of subhalos ( upper panels ) is identical or slightly stronger in late formed subhalos . this is due to the link between formation and accretion time such that more early formed halos are more likely to be accreted at earlier times . from the lower panel , it is found that for neighboring halos , there is almost no dependence on their formation times . therefore , the influence of formation time is similar to that of accretion time . in the bottom panels of figure [ formation ] , the same method is used as above to explore the influence of formation time on neighboring halos . there is no obvious deviation between two lines in these two panels . the formation time does not influence the distribution of neighboring halos , different from the case for sub - halos . the alignment signal should come from the coherent mass distribution between halos and large scale structures , nothing to do with the formation histories of neighboring halos . finally in this section , we show the effects of host halo formation time on the alignment of subhalos / neighboring halos . it is seen from figure [ formationhost ] that the formation time of host halos have weak effects on the alignment of subhalos / neighboring halos . from the press - schechter(ps ) theory @xcite , and extended ps theory ( @xcite ; see also @xcite ) , it is known that the formation time of a halo depends on its mass such that more massive halos form on average at later epochs . we should then expect that the alignment of subhalos should be stronger in later - formed halos , similar to the dependence seen in figure [ ang_host ] . however , it should be also noticed that for given halo mass , the formation time has a broad distribution . thus for a given formation time , there is a wide distribution of halo masses which dilutes the dependence of alignment on halo mass . recent observations have found that satellite galaxies are not randomly distributed , but rather they align with the major axis of the central galaxy ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? this intriguing result has generated great interest , with many studies investigating the origin of this phenomenon . the common conclusion from previous studies ( e.g. , * ? ? ? * ; * ? ? ? * ) is that the alignment arises from the non - spherical nature of dark matter halos in the cold dark matter cosmology ( e.g. , * ? ? ? * ) . however , there is no model which can accurately predict the observed alignment signal and its dependence on galaxy properties . the most difficult part of theoretical modeling is in assigning the shape of central galaxies , and most models do not agree on this aspect . in this paper , we re - visit the alignment problem using a cosmological @xmath0-body simulation . compared to previous studies , we focus on the origin of the alignment with its dependence on the formation / accretion of subhalos . we investigate if this dependence is from the assembly bias or an evolution effect . our results are summarized as follows , * we use a new method to characterize the tri - axial halo shape following @xcite . unlike the most widely used inertia tensor method which depends on the mass distribution within a given radius and is often contaminated by subhalos , the new method is able to determine the tri - axes of the halo at given local mass over - density , and excludes the effects on subhalos on the shape determination . we find that the measured halo shapes at different radii are well aligned . the mean alignment angle between the inner and outer part of halo is about @xmath62 for host halos with @xmath37 , @xmath63 and @xmath64 respectively . the alignment between the inner and outer shapes is increasing with halo mass . * we study the alignment of both subhalos and neighboring halos around selected host halos . both subhalos and neighboring halos are found to align preferentially with the outer axes of host halos . consistent with previous results ( e.g. , * ? ? ? * ) , we find that the alignment of subhalos is stronger than the data if the outer axis of host halo is used for the shape of central galaxy . better agreement with the data is achieved if the central galaxy follows the shape of host halo determined at the inner region . we also find that if the alignment between the central galaxy and the outer axis follows a gaussian distribution with a mean of @xmath52 and a deviation of @xmath53 , the predicted alignment also agree with the data . * the alignments of subhalos and neighboring halos depend on the mass of the host halos , such that more massive host halos have stronger alignment . this is due to the reasons that more massive halos are more flattened ( embedded in filaments ) and more massive halos have better alignment between its inner and outer axes . this is consistent with the observations that satellites in massive red central galaxies are more strongly aligned . in fig [ ca ] , we show the dependence of halo flattening on the mass . it is seen that higher mass halos are more flattened ( with lower c / a ) . also found is that the inner region of halo is more flattened than the outer part . these results are consistent with previous studies ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? we also note that resolution ( lower mass halos have fewer particles ) has no effect on these results as the distribution shows consistent tendency up to the low - mass end in either case . in fig [ ca ] , we include halos with their axes poorly detected . if these halos are excluded , the line of inner axes goes down at the low mass end . there is weak ( if any ) dependence of alignment on the mass of subhalos or neighboring halos themselves . * we study the alignment of subhalos with dependence on their formation time . it is found that there is no dependence along the inner major axes of host halos , and a strong dependence along the outer axes of the host halos such that the early accreted subhalos have lower alignment than the recent accreted ones . this is not consistent with the results of @xcite , and is also inconsistent with observational evidence that red satellite galaxies ( accreted more early ) have stronger alignment with the central galaxy than blue satellites . the main contribution of this paper is that we find that the mis - alignment between the inner and outer axes of the dark matter halos can account for the observed alignment of satellite galaxies including the mass dependence . we confirm the results from most studies that the shape of the central galaxy can not follow the shape of the whole dark matter halo , otherwise the predicted alignment signal is too strong . however , better agreement with observations can be obtained if the central galaxy follows the shape of the dark matter halo defined at the inner region ( measured at an overdensity of @xmath65 ) . in this case , the dependence of alignment on halo mass is also reproduced . finally , we discuss whether the stronger alignment of red satellites due to the assembly bias or an evolution effect after accretion . as our simulation do not include models for galaxy formation , we use the subhalo population to address this question . our conclusion is that the main contribution to the strong alignment of red satellites is from an evolution effect . if red satellites reside in subhalos that form at earlier times or having higher mass at accretion , figure [ ang_self ] and figure [ formation ] have shown that those subhalos do not have stronger alignment . also the alignment of neighboring halos with higher mass or higher formation redshift are also identical . these results indicate that the stronger alignment of red satellites is not set at the time of accretion . on the contrary , their strong alignment should come from evolution effects after accretion . actually , figure [ ang_comp ] further shows that the alignment of neighboring halos is lower than the subhalo , implying that subhalos acquire stronger alignment after accretion . however , our results from figure [ accretion ] seems not to support the evolution scenario . it shows that early accreted subhalos have lower alignment , indicating a negative evolution effect . it is also inconsistent with the results of @xcite who have found that early accreted satellites have stronger alignment . we note that there are differences between our analysis and theirs . in our work , we do not have galaxies , but only subhalos . also we do not include those disrupted subhalos ( ` orphan ' galaxies in the model of @xcite ) . thus in our work , those subhalos accreted at earlier times may have been disrupted , and this effect is more efficient for subhalos with more ` radial ' orbits as they will come to the host center and suffer strong tidal disruption . in addition , cautions should be taken here that the alignment on ` inner ' axes are determined less accurately because of limited resolution especially for low - mass halos and also for iso - density surfaces with small ellipticity ( the latter one is also the case when determining the major axes observationally ) . further investigation using simulations with higher resolution should be helpful . we simply conclude that to accurately predict the alignment of satellite galaxies found in observations and study its dependence on galaxy properties , we should use hydrodynamical simulation with gas physics and star formation , which can self - consistently predict the shape of central galaxies and the distribution of satellite galaxies around the central galaxy . although current simulations with star formation still have difficulties to achieve better agreement with the data , we will show in an upcoming paper that the alignment of satellites and their color dependence can be better reproduced . we thank the anonymous referee for useful suggestions to improve the manuscript . and we thank yipeng jing , xiaohu yang , cheng li and jiaxin han for helpful discussion and comments . wpl acknowledges supports by the nsfc projects ( no.10873027 , 11121062 , 11233005 , u1331201 ) . xk is supported by the nsfc ( no . 11073055 , 11333008 ) , national basic research program of china ( 2013cb834900 ) and the bairen program of the chinese academy of sciences . the work is also supported by the strategic priority research program the emergence of cosmological structures of the chinese academy of sciencesgrant no . the simulations were run in shanghai supercomputer center . part of the computing were carried on the shao super computing platform.we acknowledge support from the mpg - cas through the partnership programme between the mpia group lead by a. macci and the pmo group lead by x. kang	observations have shown that the spatial distribution of satellite galaxies is not random , but aligned with the major axes of central galaxies . this alignment is dependent on galaxy properties , such that red satellites are more strongly aligned than blue satellites . theoretical work conducted to interpret this phenomenon has found that it is due to the non - spherical nature of dark matter halos . however , most studies over - predict the alignment signal under the assumption that the central galaxy shape follows the shape of the host halo . it is also not clear whether the color dependence of alignment is due to an assembly bias or an evolution effect . in this paper we study these problems using a cosmological @xmath0-body simulation . subhalos are used to trace the positions of satellite galaxies . it is found that the shapes of dark matter halos are mis - aligned at different radii . if the central galaxy shares the same shape as the inner host halo , then the alignment effect is weaker and agrees with observational data . however , it predicts almost no dependence of alignment on the color of satellite galaxies , though the late accreted subhalos show stronger alignment with the outer layer of the host halo than their early accreted counterparts . we find that this is due to the limitation of pure n - body simulations that satellites where satellite galaxies without associated subhalos ( ` orphan galaxies ' ) are not resolved . these orphan ( mostly red ) satellites often reside in the inner region of host halos and should follow the shape of the host halo in the inner region .
complex oxides are characterized by a strong coupling between charge , spin and lattice distortions , giving rise to a rich variety of electronic and magnetic behavior which is at the origin of the multifunctional properties exhibited by this class of materials . in particular , much effort is being made towards the control of these properties by means of electric and magnetic fields . while intrinsic multiferroic materials already exhibit a coupling between magnetic and electric order parameters,@xcite the magnitude of the coupling in these compounds is weak , and a new class of composite materials is being developed that exploit the different properties of the constituent compounds to tailor systems with larger susceptibilities and more varied response functions than those of single component systems . in this context , the cross coupling in ferroelectric / magnetic heterostructures and composite materials is being intensively studied . the coupling mechanism can be classified as strain - driven or charge driven : in the first case , the piezoelectric effect of ferroelectrics is used to modify the magnetic properties of ferro- or ferrimagnets via magnetoelastic coupling;@xcite in the second case , the coupling originates from modifications in the electronic properties of the magnetic system driven by charge modulation induced by the polarization state of the ferroelectric.@xcite a particular case of strain - mediated magnetoelectric coupling occurs in epitaxial heterostructures composed of thin magnetic films grown on batio@xmath0 single crystals , including la@xmath1sr@xmath2mno@xmath0,@xcite srruo3,@xcite cofe@xmath3o@xmath4,@xcite fe,@xcite and fe@xmath0o@xmath4,@xcite where the changes in strain occur as a function of temperature via the phase transitions of batio@xmath0 or via applied electric fields . however , to date , the magnetoelastic coupling has not been addressed explicitly in detail . here , we present analytical expressions for the magnetoelastic contributions to the magnetic anisotropy induced by elastic coupling between the different phases of the batio@xmath0 substrate to the epitaxial magnetic film . this coupling explains the magnetic behavior exhibited by these systems as a function of temperature in terms of the changes in the magnetoelastic anisotropies . batio@xmath0 has a simple perovskite structure at high temperatures and undergoes several crystal phase transitions as a function of temperature.@xcite it is cubic in the temperature range from @xmath51733 k to @xmath5393 k ( @xmath6 at 393 k ) , tetragonal at temperatures down to @xmath5278 k ( @xmath7 and @xmath8 at 293 k ) , orthorhombic down to @xmath5183 k ( @xmath9 , @xmath10 , and @xmath11 at @xmath12 k ) , and rhombohedral at lower temperatures ( @xmath13 , @xmath14 at 105 k ) . for temperatures below @xmath5393 k , batio@xmath0 is ferroelectric . the non - cubic lattices correspond to slight distortions of the cubic phase , with the direction of the distortion linked to that of the electric polarization , * p:@xcite in the tetragonal phase the electric polarization points along the @xmath15 axis ; in the orthorhombic phase , the spontaneous polarization points along the @xmath16 directions of the cubic cell , leading to a slight elongation of the unit cell along that direction . although the symmetry group is orthorhombic , with lattice parameters @xmath17 , @xmath18 , and @xmath19 when p * points along [ 110 ] , it is often convenient to consider the monoclinic cell with @xmath20 and @xmath21 , giving @xmath22 , @xmath23 , @xmath24 at @xmath12 k. for the rhombohedral phase , the polarization points along the @xmath25 directions of the cubic cell , corresponding to an elongation of the cubic cell along * p*. in the following , we are interested in determining the effect of the in - plane strain on the anisotropy constants of ferromagnetic ( or ferrimagnetic ) films due to magnetoelastic coupling . we assume that the strain propagates uniformly throughout the film . when relaxation occurs , as is often the case in real systems , the expressions below correspond to upper bounds to the magnetolastic anisotropy induced in the magnetic system . if the residual strain is known , the effective anisotropies can be estimated . to calculate the magnetoelastic anisotropy contributions arising from strain , we minimize the sum of the elastic ( @xmath26 ) and magnetoelastic ( @xmath27 ) energies for a given strain state of the system . we consider the case of a cubic magnetic system:@xcite @xmath28 where @xmath29 are the direction cosines of the magnetization , @xmath30 are the components of the strain tensor , @xmath31 are the magnetoelastic coupling coefficients , @xmath32 are the elastic constants and @xmath33 are the cartesian coordinate indexes . the magnetoelastic constants are related to the magnetostriction coefficients by:@xcite @xmath34 for a general in - plane distortion in a cubic lattice , it is straightforward to calculate the in - plane strain tensor components , since by definition the latter correspond to the relative deviations from the equilibrium lattice constant ; with the aid of fig . [ fig : strain ] , we find : @xmath35/2,\end{aligned}\ ] ] where @xmath36 , @xmath37 are the equilibrium in - plane lattice parameters of the magnetic film and @xmath38 , @xmath15 , @xmath39 and @xmath40 are the parameters characterizing the planar lattice distortion . here , we discuss the case of a ( 001 ) ferromagnetic film , where @xmath41 , but the calculations can be easily modified for other cases . the out of plane strain components are free to settle to the values that minimize the total energy ( zero out of plane stress ) . the magnetoelastic anisotropy is then calculated by finding the minimum of @xmath42 with respect to @xmath43 , @xmath44 and @xmath45 : @xmath46 inserting ( [ eq : ezz]-[eq : eyz ] ) into the expression for @xmath47 gives , retaining only the terms that depend on the magnetization : @xmath48 + \\ & & + b_2\epsilon_{xy}\alpha_x\alpha_y + o(b^2/c)\end{aligned}\ ] ] where the last term is strain independent , corresponding to a sum of terms of the order of @xmath49 and @xmath50 with @xmath51 and @xmath52 dependencies . these are much smaller than the first terms and will be ignored in the following discussion . they arise from a shear strain exerted by the magnetization on the lattice ( which favors specific directions that depend on the relative sign of the magnetoelastic coupling coefficients ) and are usually neglected to first order of approximation . we note that the magnetoelastic terms are uniaxial and lower in symmetry than the magnetocrystalline anisotropy of the unstrained cubic magnetic system . a more convenient expression is obtained by using spherical coordinates , where the polar axis is set along the out of plane direction : @xmath53\sin^2\theta -\\ & & - \frac{b_1}{2}(1 + 2c_{12}/c_{11})(\epsilon_{xx}+\epsilon_{yy})\cos^2\theta\end{aligned}\ ] ] where the first two terms correspond to in - plane magnetoelastic anisotropies , and the last term is the perpendicular magnetoelastic anisotropy contribution . an in - plane uniaxial anisotropy arises for uniaxial in - plane strains ( @xmath54 ) or for shear strains , although the anisotropy direction is distinct for these two cases : in the first case , it is oriented along the main crystal axes , while for shear strains , the anisotropy is directed along the in - plane diagonal directions . a perpendicular anisotropy arises when @xmath55 . the relative magnitude of @xmath56 with respect to the magnetostatic energy term ( shape anisotropy , which favors in - plane magnetization ) , which is on the order of @xmath57 , where @xmath58 is the saturation magnetization , determines whether a state of in - plane or perpendicular magnetization is favored . for the particular case of a biaxial strain ( @xmath59 , @xmath60 ) , the magnetoelastic anisotropy energy reduces to a perpendicular magnetic anisotropy term only , with @xmath61 , as expected.@xcite we consider next the effect of square , rectangular , oblique and rhomboidal distortions on the magnetoelastic anisotropy . these planar domains arise in the cubic , tetragonal , orthorhombic and rhombohedral phases of batio@xmath0 , and are responsible for the large magnetic anisotropy modulations observed in ferromagnetic / batio@xmath0(001 ) heterostructures . we consider the magnetoelastic anisotropy constants , @xmath62 , @xmath63 , @xmath64 , expressed as : @xmath65\sin^2\theta + k_\mathrm{p}\cos^2\theta.\ ] ] the direction of easy magnetization depends on the particular @xmath66 values ; positive @xmath64 favors in - plane magnetization . the magnetoelastic anisotropy for the cubic phase of batio@xmath0(@xmath67 , @xmath68 ) leads to zero in - plane anisotropy . for the tetragonal phase ( @xmath67 ) , two types of domains are possible , @xmath15-domains ( square ) and @xmath38-domains ( rectangular , which may co - exist in two orthogonal directions ) . a multidomain state can arise when cooling the batio@xmath0 from the cubic to the tetragonal phase after film growth at elevated temperatures , for example . for this phase , no shear strains are present and @xmath69 . the orthorhombic phase may give rise to ( two ) rectangular ( @xmath70 , @xmath71 ) and ( two ) oblique ( @xmath72 , @xmath73 ) domains . finally , the rhombohedral phase ( @xmath72 , @xmath68 ) is characterized by ( two ) oblique domains . it also follows that for the oblique domains of the orthorhombic and rhombohedral phases of batio@xmath0 , @xmath74 . in general , it may be difficult to control the ferroelectric state of batio@xmath0 across the different phase transitions . for example , the growth of oxide magnetic films is often carried out at elevated temperatures , where batio@xmath0 is cubic and , as the temperature is lowered to room temperature , the batio@xmath0 tends to break into a ferroelectric multidomain state , giving rise to @xmath38 and @xmath15 domains in a batio@xmath0(001 ) single crystal . similar processes operate at the other phase transitions , leading to the presence of different types of planar domains at the interface with the magnetic film . in such cases , an effective magnetoelastic anisotropy corresponds to an average over the relative number and size of domains present ; in the case where an equal number of the two possible oblique domains are present in the rhombohedral phase , no effective in - plane anisotropy would result . however , experimental estimates of the relative population of the different domains is difficult . in the following we will consider single domain states ; for similar , but orthogonal , domains , the in - plane anisotropy constants will be identical , but with opposite signs , while the perpendicular anisotropies are unaffected . another difficulty relates to the fact that for systems with large lattice mismatch , such as fe@xmath0o@xmath4 and cofe@xmath3o@xmath4 grown on batio@xmath0 , the strain is strongly reduced for films thicker than the coherent critical thickness through the onset of misfit dislocations;@xcite in the case of fe@xmath0o@xmath4/batio@xmath0(100 ) , the experimental evidence points to full relaxation at the temperature of growth.@xcite here , we do not take explicitly into account the effect of dislocations to the magnetoelastic energy , but instead account for strain relaxation by adjusting the lattice parameter of the magnetic film so that the correct strain value is obtained at the measurement temperature , as will be discussed in more detail below . we consider three magnetic systems that have been studied experimentally and which are of particular interest : cofe@xmath3o@xmath4,@xcite fe@xmath0o@xmath4 , @xcite and fe films@xcite grown on batio@xmath0 . in table [ tab : constants ] , we list the relevant materials parameters for these magnetic systems . [ cols="<,<,<,<,<,<,<,<,<",options="header " , ] we consider first the case of fe@xmath0o@xmath4/batio@xmath0(100 ) epitaxial films , where experimentally it is found that the fe@xmath0o@xmath4 has a residual strain at room temperature with a sign that is opposite that expected from the lattice mismatch with batio@xmath0 ; this has been explained as due to strain relaxation at the temperature of growth , followed by an in - plane dilation due to the abrupt increase in the cell parameters of batio@xmath0 at the cubic to tetragonal phase transition.@xcite therefore , the magnetoelastic anisotropy contribution to the magnetic energy is very different from that expected for a fully strained fe@xmath0o@xmath4 film : while the values for the magnetostriction coefficients of fe@xmath0o@xmath4 favor a state of in - plane magnetization , experimentally a non - zero perpendicular remanence and magnetic anisotropy are observed.@xcite in fig . [ fig : fe3o4 ] we present our estimate for the variation of the magnetoelastic anisotropy constants as a function of temperature , where we took into account the temperature variation of the lattice constants of batio@xmath0 and fe@xmath0o@xmath4.@xcite the strain relaxation was taken into account , in a first approximation , by shifting the lattice constant of magnetite such that the out of plane strain at room temperature coincides with the experimental value of 0.62% for tetragonal domains.@xcite this assumes that the misfit dislocation density remains constant throughout the temperature range and across the phase transitions of batio@xmath0 . one indication that this approach is a simplification over the real processes occurring in the magnetite film is the fact that , as observed in fig . [ fig : fe3o4 ] , the strain in the cubic phase does not drop to zero as the temperature approaches the growth temperature , indicating that a strong temperature dependence of the lattice constant , or misfit dislocation density , may occur in the high temperature range . the temperature variation of the magnetic anisotropies is instructive , however , and helps explain the variations in the magnetic properties observed experimentally . focusing on the out of plane anisotropy constant @xmath64 first , we see that negative out of plane strains favor perpendicular anisotropy , which are large for the rectangular domains of the tetragonal phase and for the oblique domains of the orthorhombic and rhombohedral phases ; they are smaller by a factor of about 2 for the square domains of the tetragonal phase and the rectangular domains of the orthorhombic phase . depending on the particular domain path followed by the batio@xmath0 substrate , it is clear that large changes in the magnetic anisotropies may follow . for example , the experimental results of vaz et at.@xcite can be explained if , starting from a rectangular domain at room temperature , the batio@xmath0 surface reverts to a rectangular domain in the orthorhombic phase , corresponding to a drop in the perpendicular magnetic anisotropy , followed by an increase in @xmath64 in the rhombohedral phase . the in - plane anisotropies are also significant , but they tend to be smaller than the perpendicular anisotropies . the presence of multidomains in batio@xmath0 also tends to average out the in - plane effective anisotropy . o@xmath4/batio@xmath0(001 ) . square symbols correspond to anisotropy in planar square domains , diamonds to oblique domains , and rectangles to rectangular domains ; full symbols are for in - plane anisotropies , empty symbols for the perpendicular anisotropy constants . ] it must be emphasized that a negative @xmath64 is not sufficient to yield perpendicularly magnetized films , since the in - plane magnetostatic ( shape ) anisotropy must also be overcome , and the latter is on the order of @xmath75 erg/@xmath76 for magnetite , assuming bulk magnetization ( thin magnetite films are known to exhibit a slightly reduced magnetization , which would lead to a smaller magnetostatic energy).@xcite this value would imply that the fe@xmath0o@xmath4/batio@xmath0 films would remain in - plane magnetized , but this assumes a state of uniform strain ; in real films , one may expect larger interface strains that could give rise to larger magnetoelastic anisotropies than those predicted in our simple model calculations . we consider here the case when the cofe@xmath3o@xmath4 film is fully strained to the batio@xmath0 substrate , a situation which is expected to occur at thicknesses below the coherent growth thickness . large changes in magnetization as a function of temperature have been reported for partially relaxed 50 - 150 nm cofe@xmath3o@xmath4/batio@xmath0(100 ) epitaxial films.@xcite cobalt ferrite ( cofe@xmath3o@xmath4 ) is characterized by very large magnetostriction coefficients and has been a material of choice for strain - driven composite multiferroics.@xcite it is ferrimagnetic with a curie temperature of 790 k and is predicted to be nearly fully spin polarized.@xcite the calculated magnetoelastic anisotropies across the different phases of batio@xmath0 using the expressions derived above are shown in fig . [ fig : cofe2o4 ] , where we took into account the temperature variation of the lattice parameter of cofe@xmath3o@xmath4.@xcite the out of plane anisotropies for the fully strained film are very large , of the order of @xmath77 erg/@xmath76 , and favor in - plane magnetization , so we have omitted these from the figure . in the cubic phase , the biaxial strain contributes only to second order to the magnetic anisotropy,@xcite and the in - plane magnetic anisotropy will be dominated by the magnetocrystalline anisotropy , of the order of @xmath78 erg/@xmath76 ( favoring the in - plane @xmath79 directions ) . in the tetragonal phase , the rectangular domains induce a very large uniaxial anisotropy favoring the direction of in - plane compression ( @xmath80 axis in fig . [ fig : strain ] ) . in the orthorhombic phase , the distortion induced by the rectangular domains is smaller , and the magnetic anisotropy decreases slightly . the oblique distortions of the orthorhombic and rhombohedral phases induce relatively small magnetic anisotropies due to the much smaller value of @xmath81 , and the magnetic anisotropy may be dominated by the magnetocrystalline anisotropy , assuming that a value comparable to the tabulated room temperature value still holds . o@xmath4/batio@xmath0(100 ) . diamonds correspond to anisotropy in oblique domains , rectangles to rectangular domains ; the dashed line corresponds to the room temperature magnetocrystalline anisotropy of bulk co@xmath82fe@xmath83o@xmath4 ( which favors the @xmath84 directions ) . ] magnetization versus temperature measurements of epitaxial cofe@xmath3o@xmath4/batio@xmath0(001 ) films have been presented by chopdekar and suzuki@xcite for a 56 nm film under an applied magnetic field @xmath85 koe , along the nominal [ 100 ] and [ 010 ] in - plane directions . even at this very large magnetic field , large jumps in the magnetization are observed at the tetragonal to orthorhombic and orthorhombic to rhombohedral phase transitions of batio@xmath0 . as the temperature is lowered from 350 k to 150 k , the magnetization for @xmath86 $ ] direction increases slightly at 290 k , and then decreases more drastically at 190 k ; when @xmath87 $ ] direction , the magnetization behaves in the opposite fashion , starting and ending at values that are close to those of the magnetization when the magnetic field is applied along the [ 010 ] direction . these changes in the magnetization can be understood by assuming predominantly square domains in the tetragonal phase of batio@xmath0 , which give similar anisotropies for both the [ 100 ] and [ 010 ] directions ; the presence of a small number of rectangular domains , or a small imbalance in the two possible rectangular domains , would give rise to a small uniaxial magnetic anisotropy favoring , say , the [ 010 ] direction ; these domains would either remain rectangular in the orthorhombic phase or revert to oblique domains . in the orthorhombic phase , the square domains change to rectangular and the increase in magnetic energy along the [ 100 ] direction ( leading to a smaller magnetization ) could result from rectangular domains that are predominantly elongated along the [ 100 ] direction ( the easy axis direction is along the direction of compression for cofe@xmath3o@xmath4 , which has @xmath88 ) . the transition to the rhombohedral phase would lead to similarly small magnetic anisotropies for both [ 100 ] and [ 010 ] directions , leading to an increase in magnetic energy for the [ 010 ] direction , and a decrease along the [ 100 ] direction . despite clear hysteresis in the ferroelectric transitions of batio@xmath0 , the observation of reproducible paths in the magnetic behavior suggests that also the ferroelectric domains of batio@xmath0 also follow similar switching paths with thermal cycling . as a final example , we consider the case of fe , where the relatively small lattice mismatch with batio@xmath0 of 1.4% should allow the growth of fully strained epitaxial films . further interest in this particular system stems from recent theoretical predictions of surface magnetoelectric effects induced by spin - polarized screening at the fe / batio@xmath0 interface.@xcite the calculated magnetoelastic anisotropies for the different phases of batio@xmath0 are shown in fig . [ fig : fe ] . due to the strong crystal field of metallic fe , the orbital moment is quenched , and the magnetic anisotropies are relatively small in this system . it can be seen that for fe / batio@xmath0(100 ) , the perpendicular magnetoelastic anisotropies favor perpendicular magnetization , but given the very large magnetostatic energy of fe ( @xmath89 erg/@xmath76 ) , an in - plane magnetized state is energetically more favorable . the magnetoelastic in - plane anisotropies are also relatively small due to the small strains involved , but are likely to strongly influence the switching behavior of the fe film . square symbols correspond to anisotropy in planar square domains , diamonds to oblique domains , and rectangles to rectangular domains . filled symbols are for in - plane anisotropies and empty symbols are for the perpendicular anisotropy constants . ] sahoo et al.@xcite have reported magnetic measurements on a 10 nm polycrystalline fe film deposited on a batio@xmath0(100 ) substrate at 373 k , where the magnetization under a small applied magnetic field ( 20 - 80 oe ) shows sudden jumps at the phase transitions of batio@xmath0 . although the fe film is likely to be fully relaxed , the relative changes in magnetic anisotropy can still be compared qualitatively with the results shown in fig . [ fig : fe ] . for the particular field direction , which is applied in the plane of the film , the magnetization is found to increase steeply with decreasing temperature at both @xmath5278 k and @xmath5183 k , indicating a decrease in the magnetic anisotropy along this direction . the results shown in fig . [ fig : fe ] suggest that the magnetic field is being applied along the @xmath38 direction of the rectangular domains of the tetragonal phase , which corresponds to a hard axis direction ( in fe this is the direction of compression , since @xmath90 ) and that the dominant planar domains in the orthorhombic phase of batio@xmath0are rectangular , compressed in this direction . this also agrees with the measured temperature variation of the perpendicular magnetization , where an increase at @xmath5278 k is followed by a decrease at @xmath5183 k with decreasing temperature , in agreement with the variation found for @xmath64 , shown in fig . [ fig : fe ] . explicit expressions for the magnetoelastic anisotropy constants have been derived for arbitrary in - plane distortions of the unit cell of cubic systems and applied to the particular case where those distortions are induced by elastic coupling to batio@xmath0 substrates . batio@xmath0 undergoes a series of phase transitions as a function of temperature and has been employed to study the effect of strain on the magnetic properties of epitaxial magnetic films and , in particular , aimed at designing new magnetoelectric composite systems . since the elastic properties of ferroelectrics , and batio@xmath0 in particular , can be modulated by electric fields via the piezoelectric effect , modulation of magnetism via electric fields is possible . we have considered three particular magnetic systems that have been studied experimentally ( fe@xmath0o@xmath4 , cofe@xmath3o@xmath4 , and fe ) , and explained the observed magnetic behavior in terms of the changes in magnetic anisotropies arising from strain induced in the different phases of batio@xmath0 . this work provides a quantitative account of the experimental observations and opens the way to a more calibrated targeting of the properties of heterosystems for optimum magnetoelectric coupling effects .	analytical expressions for the magnetoelastic anisotropy constants of cubic magnetic systems are derived for rectangular and oblique distortions originating from epitaxial growth on substrates with lower crystal symmetry . in particular , the temperature variation of the magnetic properties of magnetic films grown on barium titanate ( batio@xmath0 ) substrates are explained in terms of strain - induced magnetic anisotropies caused by the temperature dependent phase transitions of batio@xmath0 . our results quantify the experimental observations in ferromagnet / batio@xmath0-based structures , which have been proposed as templates for magnetoelectric composite heterostructures .
one of the most sensational discoveries of the past decade is that our universe is currently accelerating . this observation finds support in the luminosity measurements of high - redshift supernovae @xcite , measurements of degree - scale anisotropies in the cosmic microwave background @xcite and , indirectly , in the observations of gravitational clustering @xcite . although a cosmological constant appears to satisfy all current observations , the formidable fine - tuning difficulties which accompany a non - evolving @xmath7-term have prompted theorists to investigate alternative , evolving forms of dark energy . theoretical models of dark energy usually involve matter fields possessing unusual properties such as a negative equation of state . popular dark - energy models include scalar - field - based ` quintessence ' models , models based on quantum particle production , chaplygin gas etc . ( see @xcite for recent reviews ) . in this paper , we examine a radically different mechanism for the late time acceleration of the universe in which dark energy effectively emerges from the gravity sector and not the matter sector of the theory . we study a higher - dimensional cosmology in which the observable universe is a four - dimensional ` brane ' embedded in a five - dimensional bulk . models of this kind appeared after the seminal paper by randall & sundrum ( rs ) @xcite . subsequent intensive investigation showed that cosmology based on the rs model exhibits departure from standard general - relativistic behaviour only at very early cosmological times @xcite . the model which we study in this paper generalizes the rs model in that , besides the brane and bulk cosmological constants , it also includes the scalar curvature term in the action for the brane . in an interesting recent development , deffayet , dvali , and gabadadze ( ddg ) @xcite demonstrated that the presence of the scalar curvature term in the action for the brane can lead to a late - time acceleration of the universe even in the absence of any material form of dark energy @xcite . the main difference between our model and the ddg model is that , in addition to the scalar curvature term in the action for the brane , we also include the brane and bulk cosmological constants . ( the bulk and brane cosmological constants were set equal to zero in the ddg model . ) the presence of bulk and brane cosmological constants in our braneworld ( in addition to the brane curvature term ) leads to several qualitatively new features that distinguish our model from the ddg model as well as from the scalar - field - based ` quintessence ' models . an important property of our braneworld universe is that the ( effective ) equation of state of dark energy can be @xmath8 . ( the ddg braneworld model and quintessence models have @xmath9 . ) in addition , the acceleration of the universe in our braneworld scenario can be a _ transient _ phenomenon . we find that a class of braneworld models accelerate during the present epoch but revert back to matter - dominated expansion at late times . a _ transiently accelerating _ braneworld does not possess an event horizon and could reconcile a currently accelerating universe with the demands of string / m - theory . our paper is organised as follows . after describing the braneworld model in the next section , in sec . [ vacua ] we proceed to the case of vacuum braneworlds , that is , solutions with zero stress - energy tensor of matter . we derive a general expression for the effective cosmological constant in such models and describe some classes of its symmetric solutions , including _ empty _ static homogeneous and isotropic universes . we also present simple expressions for a static universe filled with matter . in sec . [ cosmology ] we study the cosmological evolution of the braneworld . we shall show that braneworld models have properties similar to the observed ( accelerating ) universe for a large region of parameter space . for instance , the model under consideration can mimic the evolution of a frw universe in which dark energy has pressure to energy - density ratio @xmath10 both larger _ as well as smaller _ than the critical value of @xmath11 . for a subclass of parameter values , braneworld dark energy and the acceleration of the universe are _ transient _ phenomena so that the universe , after the current period of acceleration , re - enters the matter - dominated regime . in sec . [ final ] we formulate our conclusions . in this paper , we consider the case where a braneworld is the timelike boundary of a five - dimensional purely gravitational lorentzian space ( bulk ) , which is equivalent to the case of a brane embedded in the bulk with @xmath12 symmetry of reflection with respect to the brane . the theory is described by the action @xcite @xmath13 + \int_{\rm brane } \left ( m^2 r - 2 \sigma \right ) + \int_{\rm brane } l \left ( h_{ab } , \phi \right ) \ , .\ ] ] here , @xmath14 is the scalar curvature of the metric @xmath15 in the five - dimensional bulk , and @xmath16 is the scalar curvature of the induced metric @xmath17 on the brane , where @xmath18 is the vector field of the inner unit normal to the brane , and the notation and conventions of @xcite are used . the quantity @xmath19 is the trace of the symmetric tensor of extrinsic curvature @xmath20 of the brane . the symbol @xmath21 denotes the lagrangian density of the four - dimensional matter fields @xmath22 whose dynamics is restricted to the brane so that they interact only with the induced metric @xmath23 . all integrations over the bulk and brane are taken with the natural volume elements @xmath24 and @xmath25 , respectively , where @xmath26 and @xmath27 are the determinants of the matrices of components of the corresponding metrics in a coordinate basis . the symbols @xmath28 and @xmath29 denote , respectively , the five - dimensional and four - dimensional planck masses , @xmath30 is the bulk cosmological constant , and @xmath31 is the brane cosmological constant , also called the brane tension . the parameter @xmath32 in the first term of action ( [ action ] ) reflects the possibility of different relative signs between the bulk and brane terms in the action . in the context of homogeneous friedmann cosmology that will be subject of this paper , the differential equations on the brane will not depend on the sign of @xmath33 . however , different signs of @xmath33 result in different ways of embedding of a brane with one and the same induced metric in the bulk space , which may be important for studying perturbations of the theory . the term containing the scalar curvature of the induced metric on the brane with the coupling @xmath34 in action ( [ action ] ) is often neglected in the literature . however , this term is qualitatively essential for describing the braneworld dynamics since it is inevitably generated as a quantum correction to the matter action in ( [ action ] ) in the spirit of an idea that goes back to sakharov @xcite ( see also @xcite ) . note that the effective action for the brane typically involves an infinite number of terms of higher order in curvature ( this was pointed out in @xcite for the case of braneworld theory ; a similar situation in the context of the ads / cft correspondence is described in @xcite ) . in this paper , we retain only the terms linear in curvature . the linear effects of the curvature term on the brane were studied in @xcite , where it was shown that it leads to four - dimensional law of gravity on sufficiently small scales . for some recent reviews of the results connected with the induced - gravity term in the brane action , one may look into @xcite . action ( [ action ] ) leads to the bulk being described by the usual einstein equation with cosmological constant : @xmath35 while the field equation on the brane is @xmath36 here , @xmath37 is the stress - energy tensor on the brane , and it stems from the last term in action ( [ action ] ) . in eq . ( [ brane ] ) we have underlined the term whose presence makes braneworld theory different from general relativity . ( it should perhaps be mentioned that the variation of action ( [ action ] ) which includes the gibbons hawking surface term @xmath38 leads to the israel junction conditions @xcite on the brane , as demonstrated in @xcite . ) from ( [ brane ] ) one can show that there exists a scale whose value determines the domain in which general relativity is approximately valid @xcite . indeed , we can write @xmath39 where @xmath40 and @xmath41 are the characteristic curvature radii of solution of eq . ( [ brane ] ) . then the ratio of the last term on the right - hand side of eq . ( [ brane ] ) to the first term on its left - hand side is of the order @xmath42 , where @xmath43 is a convenient definition to be used later . thus , for @xmath44 , the last term on the left - hand side of eq . ( [ brane ] ) can be neglected , and general relativity is recovered . in typical situations , e.g. , in cosmology , we often have @xmath45 , and general relativity becomes valid for @xmath46 . this fact was used in @xcite , where the cosmological model with @xmath47 and @xmath48 was considered . therefore , an important distinguishing feature of the braneworld models which we consider , is that they can show departure from standard frw behaviour at _ late times _ and on large scales . this property should be contrasted with the braneworld cosmology based on the randall sundrum model @xcite ( with @xmath49 ) , in which nonstandard behaviour is encountered at very early times @xcite . by contracting the gauss identity @xmath50 on the brane and using eq . ( [ bulk ] ) , one obtains the ` constraint ' equation @xmath51 which , together with ( [ brane ] ) , implies the following closed scalar equation on the brane : @xmath52 where @xmath53 . one method for obtaining solutions of the theory consists in first solving the scalar equation ( [ closed ] ) on the brane together with the stress - energy conservation equation , and then integrating the einstein equations in the bulk with the given data on the brane @xcite . we note that eq . ( [ closed ] ) describes the evolution on the brane in terms of its intrinsic quantities , and that all homogeneous and isotropic cosmological solutions for the brane can be obtained from this equation @xcite , as will also be done below . all such solutions are embeddable in the schwarzschild - ads five - dimensional bulk . however , one can not confine oneself to this single equation in the general case , e.g. , in studying perturbations of the cosmological solution , where one must also solve equations in the bulk imposing certain boundary and/or regularity conditions . from this general viewpoint , ( [ closed ] ) can not be regarded as complete : only those of its solutions are admissible which can be developed to regular solutions in the bulk . this illustrates the importance of the precise specification of boundary and/or regularity conditions in the bulk for a complete formulation of the brane - world theory @xcite . the gravitational equations in the bulk can be integrated by using , for example , gaussian normal coordinates , as described , e.g. , in @xcite . specifically , in the gaussian normal coordinates @xmath54 , where @xmath55 are the coordinates on the brane and @xmath56 is the fifth coordinate in the bulk , the metric in the bulk is written as @xmath57 introducing also the tensor of extrinsic curvature @xmath58 of every hypersurface @xmath59 in the bulk , one can obtain the following system of differential equations for the components @xmath60 and @xmath61 : @xmath62 @xmath63 where @xmath64 are the components of the ricci tensor of the metric @xmath60 induced on the hypersurface @xmath59 , @xmath65 is its scalar curvature , and @xmath66 is the trace of the tensor of extrinsic curvature . the second equality in ( [ s ] ) is true by virtue of the ` constraint ' equation ( [ constraint ] ) . equations ( [ s ] ) and ( [ metric ] ) together with the ` constraint ' equation ( [ constraint ] ) represent the @xmath67 splitting of the einstein equations in the gaussian normal coordinates . the initial conditions for these equations are defined on the brane through eq . ( [ brane ] ) . we emphasise once again that , to obtain a complete braneworld theory in the general case ( including a stability analysis ) , one must also specify additional conditions in the bulk such as the presence of other branes or certain regularity conditions . in our paper , we deal only with the homogeneous and isotropic cosmology on the brane , so this issue does not arise . in this sense , we are studying here the cosmological features common to the whole class of braneworld models described by action ( [ action ] ) with arbitrary boundary conditions in the bulk . we proceed to the cosmological implications of the braneworld theory under consideration . in this section , we discuss the situation pertaining to a vacuum brane , i.e. , when the matter stress - energy tensor @xmath68 . it is interesting that the brane approaches this condition during the course of cosmological evolution provided it expands forever and its matter density asymptotically declines to zero . in this case , eq . ( [ closed ] ) takes the form @xmath69 where @xmath70 it is important to note that the second term in eq . ( [ vacuum ] ) has _ precisely _ the form of one of the terms in the expression for the conformal anomaly , which describes the vacuum polarization at the one - loop level in curved space - time ( see , e.g. , @xcite ) . can not be obtained by the variation of a local four - dimensional lagrangian , the very same term is obtained via the variation of a local lagrangian in the five - dimensional braneworld theory under investigation . ] it therefore immediately follows that all _ symmetric spaces _ are solutions of eq . ( [ vacuum ] ) with appropriate @xmath71 , just as they are solutions of the einstein equations with one - loop quantum - gravitational corrections @xcite . symmetric spaces satisfy the condition @xmath72 , which implies that geometrical invariants such as @xmath73 , and @xmath16 are constants so that eq . ( [ vacuum ] ) becomes an algebraic equation . prominent members of this family include : \(i ) the homogeneous and isotropic de sitter space - time @xmath74 \ , , \ ] ] where @xmath75 , @xmath76 , @xmath77 . the four dimensional metric ( [ desitter ] ) has the property @xmath78 , and formed the basis for starobinsky s first inflationary model sustained by the quantum conformal anomaly @xcite . \(ii ) the homogeneous and anisotropic nariai metric @xcite @xmath79 where @xmath80 = constant , @xmath75 , @xmath81 , @xmath82 and for which @xmath83 . in fact , it is easy to show that any metric for which @xmath16 and @xmath84 are constants will automatically be a solution to eq . ( [ vacuum ] ) with an appropriate choice of @xmath71 . both de sitter space and the nariai metric belong to the class of space - times which satisfy the vacuum einstein equations with a cosmological constant @xmath85 such space - times also satisfy eq . ( [ vacuum ] ) if @xmath86 \ , .\ ] ] equation ( [ lambda ] ) expresses the resulting cosmological constant on the brane in terms of the coupling constants of the theory . for the frequently discussed special case @xmath49 , one obtains @xmath87 . the two signs in ( [ lambda ] ) correspond to the two different ways in which the lower - dimensional brane can form the boundary of the higher - dimensional bulk @xcite . in the case of a spherically symmetrical bulk , the ` @xmath88 ' sign corresponds to a brane as a boundary for which the _ inner _ normal to the brane points in the direction of decreasing bulk radial coordinate . the condition @xmath89 is the well - known fine - tuning condition of randall and sundrum @xcite and leads to the vanishing of the cosmological constant on an empty brane if we set @xmath90 in ( [ action ] ) . note that , under the randall sundrum condition , expression ( [ lambda ] ) with the sign opposite to the sign of the quantity @xmath91 also gives a zero value for the resulting cosmological constant on the brane , but the other sign usually leads to @xmath92 . we would like to draw the reader s attention to the fact that eq . ( [ lambda ] ) is meaningful only when the expression under the square root is nonnegative . when it is negative , solutions describing the corresponding empty universe simply do not exist . this leads to the following important conclusion : a universe which contains matter and satisfies @xmath93 _ can not expand forever_. for the special case @xmath94 , the expression for @xmath95 is given by @xmath96 in this case , both @xmath31 and @xmath71 must be negative in order that the corresponding empty universe exist , but the resulting cosmological constant on the brane can be of any sign . another interesting example is that of a static empty universe . the radius ( scale factor ) @xmath97 of such a universe is easily determined from ( [ vacuum ] ) to be @xmath98 where @xmath99 is the sign of the spatial curvature . one can see that the radius of the universe can be arbitrarily large . in the general case , the development of this solution to the five - dimensional bulk leads to a schwarzschild anti - de sitter metric . it was shown in @xcite that , for @xmath100 , this metric is purely anti - de sitter ( with zero schwarzschild mass ) if the constants of the theory satisfy the condition @xmath101 which implies negative brane tension @xmath31 . it should be pointed out that the static and empty braneworld solution described by ( [ static ] ) does not possess a general - relativistic analog , since , in general relativity , a static cosmological model ( the ` static einstein universe ' ) _ can not _ be empty ( see , for instance , @xcite ) . furthermore , from ( [ static ] ) we find that the static empty universe can be spatially open ( @xmath102 ) for example , in the case @xmath103 and @xmath104 , again a situation without an analog in general relativity . for static homogeneous and isotropic braneworlds filled with matter , eq . ( [ closed ] ) gives the following relation : @xmath105 = 3 \kappa \left [ m^2 ( \rho_{\rm tot } + 3 p_{\rm tot } ) - 3 m^6 \right ] \ , , \ ] ] where the total energy density @xmath106 and pressure @xmath107 include the contribution from the brane tension , i.e. , @xmath108 and @xmath109 corresponds to the sign of the spatial curvature . this relation reduces to ( [ static ] ) for @xmath110 . having obtained all these solutions on the brane , one can find the corresponding solutions in the bulk by integrating eqs . ( [ s ] ) and ( [ metric ] ) with the initial conditions on the brane given by eq . ( [ brane ] ) , as described in sec . [ basic ] . in doing this , one can consider various additional conditions in the bulk , for example , existence of other branes , or one can impose certain regularity conditions . it is worth noting that one and the same cosmological solution on the given brane can correspond to different global solutions in the bulk , for example , other branes may be present or absent , be static or evolving , close or far away from our brane , etc . in the most general case ( for instance in the absence of special symmetries on the brane ) integration on the brane needs to be performed in conjunction with dynamical integration in the bulk . all such situations must be separately studied and issues such as their stability to linearized perturbations must be examined on a case - by - case basis . in the homogeneous and isotropic cosmological setting , eq . ( [ closed ] ) can be integrated with respect to time with the result @xcite @xmath111 where @xmath112 is the integration constant that corresponds to the black - hole mass of the schwarzschild(anti)-de sitter solution in the bulk , @xmath113 is the hubble parameter on the brane , and @xmath114 corresponds to the sign of the spatial curvature on the brane . a generalization of this equation to the case of absence of @xmath12 symmetry was obtained in @xcite . for the frequently considered special case @xmath49 , this equation reduces to the familiar one @xcite @xmath115 equation ( [ cosmo ] ) can be solved with respect to the hubble parameter with the result @xcite : @xmath116 \ , , \ ] ] or , equivalently , @xmath117 ^ 2 \ , , \ ] ] where the length scale @xmath118 was introduced earlier in sec . [ basic ] . again , the @xmath119signs in ( [ solution ] ) and ( [ solution1 ] ) correspond to two different ways of bounding the schwarzschild(anti)-de sitter bulk space by the brane @xcite : the lower sign corresponds to the case where the inner normal to the brane ( which bounds the bulk ) points in the direction of decreasing bulk radial coordinate . alternatively , the two different signs in ( [ solution ] ) and ( [ solution1 ] ) could correspond to the two possible signs of the five - dimensional planck mass @xmath28 . henceforth we shall refer to models with the lower ` @xmath88 ' sign as brane1 and models with the upper ` @xmath120 ' sign as brane2 . in the ensuing discussion , we shall neglect the decaying term @xmath121 in eq . ( [ solution ] ) ( called dark radiation ) , assuming it to be negligibly small at present . similarly to @xcite , we also introduce the dimensionless cosmological parameters @xmath122 where the subscript ` 0 ' refers to current values of cosmological quantities . the system of cosmological equations with the energy density @xmath123 dominated by dust - like matter can now be written in the transparent form using eq . ( [ solution ] ) : * brane1 @xmath124 * brane2 @xmath125 where @xmath126 is the cosmological redshift . we have underlined the terms which cause our equations to differ from their general - relativistic counterparts . the constraint equations for the cosmological parameters are most conveniently derived using eq . ( [ solution1 ] ) . we obtain @xmath127 this equation can be transformed to a more convenient form . in doing this , we note that the theory makes sense only in the case @xmath128 . since @xmath129 , when taking the square root of both sides of this equation , we must consider two possibilities for the sign of the quantity @xmath130 under the square root in ( [ original ] ) . in the case @xmath131 taking the square root of both sides of ( [ original ] ) and rearranging terms , we get @xmath132 now taking the square of both sides of the last equation , we finally obtain the constraint : @xmath133 in the opposite case @xmath134 taking the square root of both sides of ( [ original ] ) and rearranging terms , we get @xmath135 taking the square of both sides of the last equation , we finally obtain the constraint : @xmath136 to summarise , in the case @xmath137 , the constraint equations in the brane1 and brane2 models read : @xmath138 @xmath139 in the opposite case @xmath140 , both constraint equations for brane1 and brane2 are given by eq . ( [ omega - r2 ] ) . again , we have underlined the terms which cause our equations to differ from their general - relativistic counterparts . it is easy to see that the general - relativistic limit of this theory is reached as @xmath141 ( i.e. , @xmath142 ) . in this case , setting @xmath143 , we find that ( [ hubble1 ] ) & ( [ hubble2 ] ) give the expression for the hubble parameter in the lcdm model : @xmath144 an example of the behaviour of the hubble parameter in the braneworld models is illustrated in fig . [ fig : hubble ] . it is convenient to rewrite eq . ( [ lambda - rs ] ) in terms of the dimensionless cosmological parameters @xmath145 from where it follows that the randall sundrum constraint @xmath89 has the simple form @xmath146 this constraint implies @xmath147 ( @xmath148 ) and ensures the vanishing of the effective cosmological constant in the future for the braneworld embedding described by the brane1 model for @xmath149 or for @xmath150 and @xmath151 , and by the brane2 model for @xmath150 and @xmath152 . the case where the randall sundrum constraint is imposed will be discussed in detail in sec . [ sec : rs ] below . it will be shown that , in the case of brane1 model , for the case @xmath153 , eq . ( [ omega - r1 ] ) would imply that @xmath154 , which apparently contradicts current observations of the large - scale distribution of matter . this is an important result since it suggests that for reasonable values of the matter density ( @xmath155 ) and @xmath156 , the brane1 model demands the presence of a nonvanishing effective cosmological constant . the value of the cosmological constant is given by eq . ( [ lambda ] ) and can be determined by using observational data to fix the constants of the theory . on the other hand , the brane2 model is in principle observationally compatible with the randall sundrum constraint , which can result in vanishing effective cosmological constant in the future thus making the current acceleration of the universe a transient phenomenon . typical values of the @xmath157 parameters ( [ omegas ] ) that we consider in this paper are of order @xmath158 . for such values , the fundamental constants of the theory have the following orders of magnitude : @xmath159 the smallness of the bare cosmological constants in the bulk and on the brane represents a fine - tuning similar to what is the case for the cosmological constant in the standard lcdm model . however , even with such small values of the bare cosmological constants in action ( [ action ] ) , the braneworld evolution exhibits some qualitatively new properties ( discussed in the following subsections ) when compared to the case where these constants are set to zero . this is the reason why we find it important to keep open the possibility of nonzero , albeit small , values of @xmath30 and @xmath31 . a possible procedure for testing the braneworld model against observations is presented in the appendix . observations of high - redshift type ia supernovae indicate that these objects are fainter than they would be in a standard cold dark matter cosmology ( scdm ) with @xmath160 @xcite . this observation is taken as support for a universe which is accelerating , fuelled by a form of energy with negative pressure ( dark energy ) . in standard frw cosmology the acceleration of the universe is described by the equation [ eq : acc ] = -_i ( _ i + 3p_i ) , where the summation is over all matter fields contributing to the dynamics of the universe . it is easy to show that a necessary ( but not sufficient ) condition for acceleration ( @xmath161 ) is that the strong energy condition is violated for _ at least one _ of the matter fields in ( [ eq : acc ] ) , so that @xmath162 . in the case of the popular @xmath95cdm model this requirement is clearly satisfied since @xmath163 in pressureless ( cold ) matter , while @xmath164 in the cosmological constant . the situation with respect to braneworld models is different since , as we have demonstrated in the previous section , braneworld evolution is distinct from frw evolution at late times . however it is easy to show that braneworld models can accelerate . we demonstrate this by noting that a completely general expression for the deceleration parameter @xmath165 is provided by @xmath166 where @xmath167 is given by ( [ hubble1])([omega - r2 ] ) and the derivative is with respect to @xmath126 . the current value of the deceleration parameter is easily calculated to be @xmath168 where the lower and upper signs correspond to brane1 and brane2 models , respectively . thus the present universe will accelerate ( @xmath169 ) for brane parameter values which satisfy 32 _ m ( 1 ) < 1 . it should be pointed out that the proposition to use the induced curvature term on a braneworld to account for the accelerated phase of our universe was first made in @xcite in the context of the braneworld model of @xcite , which is the special case of our brane2 model with @xmath170 . it was subsequently discussed and tested in @xcite . in the present paper , we allow both @xmath171 and @xmath172 , and the presence of these two free parameters makes the braneworld model more flexible and allows for qualitatively new behaviour which shall be discussed in detail in this section . the connection between large extra dimensions and dark energy has also been investigated in @xcite . observationally , a pivotal role in the case for an accelerating universe is played by the _ luminosity distance _ @xmath173 , since the flux of light received from a distant source varies inversely to the square of the luminosity distance , @xmath174 . this effect is quantitatively described by the magnitude luminosity relation : @xmath175 , where @xmath176 is the corrected apparent peak @xmath177 magnitude and @xmath178 is the absolute peak luminosity of the supernova . a supernova will therefore appear fainter in a universe which possesses a larger value of the luminosity distance to a given redshift . in a frw universe , the luminosity distance is determined by the hubble parameter and three - dimensional spatial curvature @xcite : @xmath179 where @xmath180 and @xmath181 is defined as follows : @xmath182 if @xmath100 ( @xmath183 ) , @xmath184 if @xmath102 ( @xmath185 ) , and @xmath186 if @xmath187 ( @xmath188 ) . inflationary models suggest that the universe is nearly flat with @xmath189 and we shall work under this assumption in the rest of this paper ( see also @xcite ) . in this case , eq . ( [ eq : age11d ] ) simplifies to @xmath190 where @xmath167 is determined by ( [ hubble1 ] ) for brane1 , and by ( [ hubble2 ] ) for brane2 . in fig . [ fig : lum ] we show the luminosity distances for the brane1 & brane2 models . also shown for comparison is the value of @xmath173 in a spatially - flat two - component frw universe with the hubble parameter @xmath191^{1/2 } , \label{hubble3}\ ] ] where @xmath192 describes dark energy with equation of state @xmath193 . three cosmological models will be of interest to us in connection with ( [ hubble3 ] ) : \(i ) scdm : the standard cold dark matter universe with @xmath160 and @xmath194 . \(ii ) lcdm : cold dark matter @xmath120 a cosmological constant with @xmath195 . \(iii ) phantom models : cold dark matter @xmath120 ` phantom energy ' satisfying @xmath3 @xcite . we find from fig . [ fig : lum ] that the luminosity distance in both braneworld models exceeds that in scdm . in fact , brane1 models have the unusual feature that their luminosity distance can even exceed that in lcdm ( for a fixed value of @xmath196 ) ! in fact it can easily be shown that @xmath197 where @xmath198 refers to the luminosity distance in the spatially flat coordinatization of de sitter space ( equivalently , the steady state universe ) . the second inequality presumes a fixed value of @xmath196 . brane2 models show complementary behaviour @xmath199 where the first inequality is valid for a fixed value of @xmath196 . in the case @xmath200 , the equations of the braneworld theory formally reduce to those of general relativity , and we have @xmath201 . one might add that the behaviour of brane1 is mimicked by frw models with @xmath202 , whereas brane2 resembles dark energy with @xmath203 @xcite . in fact , from fig . [ fig : lum ] we see that the luminosity distance in the brane1 model is quite close to what one gets from ` phantom energy ' described by ( [ hubble3 ] ) with @xmath204 . ( the parameters for this brane1 model are @xmath205 , @xmath206 , and @xmath207 . ) it should be pointed out that phantom energy models were introduced by caldwell @xcite , who made the important observation that dark - energy with @xmath3 appears to give a better fit to supernova observations than lcdm ( which has @xmath195 ) . unfortunately , phantom models have several bizarre properties , which may be the reason why cosmologists have been reluctant to take these models seriously despite their seemingly better agreement with observations ( see however @xcite ) . some strange properties of phantom energy with a very negative equation of state ( @xmath3 ) are summarised below ( see also @xcite ) : \(i ) a negative equation of state suggests that the effective velocity of sound in the medium @xmath208 can become larger than the velocity of light . \(ii ) the expansion factor of a universe dominated by phantom energy grows as @xmath209^{2 \big/ 3(1+w ) } \ , , ~~ w < -1 ~,\ ] ] where @xmath210 marks the epoch when the densities in matter and phantom energy are equal : @xmath211 . it immediately follows that the scale factor diverges in a _ finite _ amount of cosmic time @xmath212 substitution of @xmath213 and @xmath3 in ( [ hubble3 ] ) shows that the hubble parameter also diverges as @xmath214 , implying that an infinitely rapid expansion rate for the universe has been reached in the _ finite _ future . as the universe expands , the density of phantom energy ( @xmath3 ) _ grows _ instead of decreasing ( @xmath215 ) or remaining constant ( @xmath195 ) , @xmath216^{-2 } , \ ] ] reaching a singular value in a finite interval of time @xmath217 , @xmath214 . this behaviour should be contrasted with the density of ordinary matter which drops to zero : @xmath218 as @xmath219 . a universe dominated by phantom energy is thus doomed to _ expand towards a physical singularity _ which is reached in a finite amount of proper time . ( an exact expression for the time of occupance of the phantom singularity is given in @xcite , which also contains an interesting discussion of related issues . ) at this stage one must emphasize that , although the brane1 model has several features in common with phantom energy , ( which makes us believe that , like the latter , it too is likely to provide a good fit to supernova data ) , it is not necessarily afflicted with phantom s pathologies . indeed , in a broad range of parameters , both brane1 and brane2 are physically well motivated and remain _ well behaved during all times_. the present paper therefore adds an important new dimension to the current debate about the acceleration of the universe by showing that cosmological models with @xmath220 are possible to construct within the framework of the braneworld scenario and should be taken seriously . future studies will address the important quantitative issue of whether braneworld models of dark energy provide as good ( or better ) fit to high-@xmath126 supernova observations as lcdm ( see fig . [ fig : sn ] as an illustration ) . ( it is appropriate to mention here that the weak energy condition @xmath221 can also be violated in a class of scalar - tensor theories of gravity , as discussed in @xcite . ) we should add that the ` angular - size distance ' @xmath222 is related to the luminosity distance @xmath223 through @xmath224 , therefore much of the above analysis carries over when one discusses properties of the angular - size distance within the framework of braneworld models . some cosmological features of braneworld models are shown in figs . [ fig : age][fig : omega ] . in fig . [ fig : age ] , the age of the universe at a given cosmological redshift @xmath225 is shown for the two braneworld models and for lcdm & scdm . we find that the age of the universe in brane1 ( brane2 ) is larger ( smaller ) than in lcdm for identical values of the cosmological density parameter @xmath196 . this is a direct consequence of the fact that the hubble parameter in brane1 ( brane2 ) is smaller ( larger ) than in lcdm . both braneworld models are significantly older than scdm . considerable insight into the dynamics of the universe is provided by the cosmological deceleration parameter ( [ decel ] ) . our results , shown in fig . [ fig : decel ] , indicate that at late times the brane1 ( brane2 ) universe accelerates at a faster ( slower ) rate than lcdm ( with identical @xmath196 ) . curiously , the brane1 universe shows an earlier transition from deceleration to acceleration than any of the other models . ( for the given choice of parameters this transition takes place at @xmath226 for brane1 and at @xmath227 for lcdm . the brane2 model begins accelerating near the present epoch at @xmath228 . ) a related point of interest is that at @xmath229 the deceleration parameter in brane1 marginally exceeds that in scdm indicating that the brane1 model is decelerating at a faster rate than scdm ( @xmath230 ) . in conventional models of dark matter this behaviour can occur only if the equation of state of the dark component is stiffer than dust , implying @xmath231 in ( [ hubble3 ] ) , or if the universe is spatially closed . on the other hand , the current _ acceleration _ rate of brane1 in our example ( @xmath232 ) significantly exceeds that of lcdm ( @xmath233 ) with an identical value of @xmath234 in both models . within the framework of four - dimensional einstein gravity , this situation can only arise if the equation of state of dark energy is strongly negative : @xmath3 in ( [ hubble3 ] ) . the unusual high-@xmath126 behaviour of the deceleration parameter in brane1 can be better understood if we consider the cosmological density parameter @xmath235 ^ 2 \omega_{\rm m}(0 ) ( 1+z)^3 \ , , \label{eq : omega}\ ] ] where @xmath167 is given by ( [ hubble1])([omega - r2 ] ) for braneworld models . from fig . [ fig : omega ] we notice that , for @xmath236 , the value of @xmath237 in brane1 _ exceeds _ its value in scdm ( @xmath160 ) . this is precisely the redshift range during which @xmath238 . thus , the rapid deceleration of brane1 at high redshifts can be partly attributed to the larger value of the matter density @xmath237 at those redshifts , relative to scdm . having established partial similarity of brane1 with phantom models at low redshifts , we can investigate the analogy further and calculate the effective equation of state of dark energy @xmath239 } \ , , \ ] ] where @xmath237 is given by ( [ eq : omega ] ) . one notes that @xmath240 has a pole - like singularity at @xmath241 for brane1 , which arises because @xmath242 crosses the value of unity at @xmath241 ( see fig . [ fig : omega ] ) . this demonstrates that the notion of ` effective equation of state ' is of limited utility for this model . equations ( [ hubble1 ] ) , ( [ hubble2 ] ) , ( [ decel ] ) , and ( [ eq : dark ] ) also illustrate the important fact that dark energy in braneworld models , though similar to phantom energy in some respects , differ from it in others . for instance , in both braneworld models , @xmath243 at @xmath244 and @xmath245 as @xmath246 , whereas phantom energy has @xmath247 at _ all _ times . a useful quantity is the _ current value _ of the effective equation of state of dark energy in braneworld theories : @xmath248 where the lower and upper signs , as usual , correspond to brane1 and brane2 models , respectively . we easily see that @xmath249 for brane1 , whereas @xmath250 for brane2 . the braneworld models under consideration also admit the intriguing possibility that the current acceleration of the universe may not be a lasting feature . it may be recalled that most models of dark energy including the cosmological constant have the property that , once the universe begins to accelerate , it accelerates forever . as shown in a number of recent papers , an eternally accelerating universe is endowed with a cosmological event horizon which prevents the construction of a conventional s - matrix describing particle interactions within the framework of string or m - theory @xcite . in this section we show that , provided the randall sundrum constraint relation ( [ rs ] ) is satisfied , the acceleration of the universe can be a transient phenomenon in braneworld models . an anisotropic solution of bianchi v class with the same feature was described in @xcite . from eqs . ( [ hubble1 ] ) and ( [ hubble2 ] ) we obtain the following asymptotic expressions for the hubble parameter @xmath251 as @xmath213 , assuming that the universe expands forever : @xmath252 where the lower and upper signs correspond to brane1 and brane2 models , respectively . in applying the randall sundrum constraint ( [ rs ] ) , we first consider the case where @xmath253 . then @xmath254 and @xmath255 \ , .\ ] ] one can see that this expression vanishes for the lower sign . thus , for positive @xmath256 , it is the brane1 model that leads to vanishing effective cosmological constant in the future . however , in this case , the constraint equation ( [ omega - r1 ] ) with @xmath156 becomes @xmath257 and implies @xmath154 , which is hardly compatible with the observations . in the case of @xmath150 , we have @xmath258 and @xmath259 if @xmath260 , then this expression vanishes for the lower sign , which brings us back to the nonphysical brane1 models with @xmath154 . the case @xmath261 is compatible with the condition @xmath262 , which corresponds to brane1 ( [ hubble1 ] ) with the constraint equation ( [ omega - r2 ] ) . however , it can be shown that these conditions imply @xmath263 , which is also physically unacceptable . there remains the case of @xmath150 and @xmath264 . in this case , expression ( [ negative ] ) vanishes for the upper sign , which corresponds to brane2 models . the constraint equation ( [ omega - r2 ] ) with @xmath156 now reads @xmath265 and implies @xmath266 . therefore , brane2 with @xmath150 and @xmath264 , provides us with an interesting example of a physically meaningful cosmological model in which the current acceleration of the universe is a _ transient phenomenon_. an example of this behaviour as probed by the deceleration parameter is shown in fig . [ fig : decel_plot ] , which demonstrates that the current period of cosmic acceleration takes place between two matter - dominated epochs . we emphasize that these models require negative brane tension @xmath31 . since an observer in this model resides on a negative - tension brane one must ponder over the issue of whether such a braneworld will be perturbatively stable and hence physically viable . we consider this to be an open question for future investigations . remarks made at the end of sec . [ vacua ] are relevant , however , since one and the same cosmological solution on the ` visible ' ( negative tension ) brane can correspond to many different global conditions in the bulk , for instance , other ( ` hidden ' ) branes may be present or absent , static or evolving , close to or far away from our brane , etc . moreover , here the sign of the parameter @xmath32 in action ( [ action ] ) is expected to be of crucial importance , since , in particular , the bulk gravity will be localised around the _ negative _ tension ( visible ) brane in the case of @xmath267 . issues of perturbative stability involve the examination of all such distinct global solutions on a case - by - case basis . such a study , though interesting and important , clearly lies beyond the scope of the present paper . useful insight into the brane2 model is also provided by the effective equation of state of dark energy ( [ eq : dark ] ) . our results shown in fig . [ fig : state_future ] indicate that the past and future behaviour of dark energy in the braneworld universe can be very different . the past behaviour @xmath243 for @xmath244 arises because , in a spatially flat braneworld , the second most important contribution to braneworld expansion at high redshifts is caused by the @xmath268 term in ( [ hubble2 ] ) ; see also @xcite . the future behaviour @xmath269 as @xmath213 , on the other hand , reflects the decreasing importance of dark energy as the universe expands . the acceleration of the universe is therefore a transient phenomenon which ends once the universe settles back into the matter - dominated regime . finally , we should mention that a transiently accelerating regime also arises in a class of brane2 models which do not satisfy the randall - sundrum constraint ( [ rs ] ) . in these models the current epoch of acceleration is succeeded by an epoch during which the deceleration parameter grows without bound . this unusual ` future singularity ' is reached in a _ finite _ interval of expansion time and is characterised by the fact that both the matter density and the hubble parameter remain finite , while @xmath270 ( a feature that distinguishes it from the phantom singularities discussed in sec . a detailed discussion of the ` new ' singularities which occur in braneworld models can be found in @xcite . in this paper , we have considered cosmological implications of a braneworld model described by action ( [ action ] ) , which contains both bulk and brane curvature terms and cosmological constants . the curvature term for the brane arises naturally , as a quantum correction from the matter part of the brane action , and significantly changes the behaviour of the braneworld theory . for example , as is well known , braneworld cosmology without this term deviates from general relativity at _ large _ matter densities , i.e. , at early cosmological times ; see eq . ( [ cosmolim ] ) . in the model that we considered , on the other hand , early cosmological evolution remains virtually unaffected for a broad range of parameters , significant deviations from standard cosmology appearing only during _ later _ times . we restrict our attention to the important case where the brane forms the boundary of the five - dimensional bulk space . this is equivalent to endowing the bulk with the @xmath12 reflection isometry with respect to the brane . the presence of the curvature term in the action leads to two families of braneworld models . these two families ( called brane1 & brane2 ) differ in the manner in which the brane forms the boundary of the five - dimensional bulk . for example , if a spatially three - spherical brane is embedded into the three - spherically symmetric bulk , then one can discard either the interior part of the bulk , or its exterior ( asymptotically flat ) part . in this example , the models which discard the exterior part leaving the interior were classified as brane1 models , while the complementary models were called brane2 . similar classification arises in the case of spatially flat and open braneworld models . alternatively , the two different families of braneworld models can be regarded as corresponding to the two possible signs of the five - dimensional planck mass @xmath28 . braneworld models of dark energy considered in this paper have the interesting and unusual property that their luminosity distance @xmath223 can _ exceed _ that in lcdm . this is unusual since , within the general - relativistic framework , the luminosity distance has this property _ only if _ the equation of state of dark energy is strongly negative ( @xmath3 ) . caldwell @xcite has shown that dark energy with @xmath3 ( phantom energy ) may in fact fit supernova observations better than lcdm . however , phantom energy is beset with a host of undesirable properties which makes this model of dark energy unattractive . we show that braneworld models have all the advantages and none of the disadvantages of phantom models and therefore endow dark energy with exciting new possibilities . a recent analysis of braneworld models in @xcite has demonstrated that brane1 models ( which generically have @xmath271 ) provide an excellent fit to supernovae observations for higher values of the matter density ( @xmath272 ) , whereas lower values ( @xmath273 ) are preferred by brane2 models ( which generically have @xmath274 ) . a distinctive feature of the braneworld scenario discussed in this paper is that it allows for a universe which is _ transiently accelerating_. recent investigations indicate that an eternally accelerating universe , which possesses a cosmological event horizon , prevents the construction of a conventional s - matrix describing particle interactions within the framework of string or m - theory @xcite . we have demonstrated that braneworld models can enter into a regime of accelerated expansion at late times _ even if _ the brane tension and the bulk cosmological constant are tuned to satisfy the randall sundrum constraint on the brane . in this case , braneworld dark energy and the acceleration of the universe are _ transient _ phenomena . in this class of models , the universe , after the current period of acceleration , re - enters the matter - dominated regime . we have shown that viable models realising this behaviour are those of brane2 type . ( since these braneworlds have a negative brane tension the question of their stability is important and needs to be examined in detail . we shall return to this issue in a companion paper . ) thus braneworld models can give rise to a _ transiently accelerating phase _ thereby reconciling a dark energy dominated universe with the requirements of string / m - theory . ( a similar observation was previously made in @xcite in the context of an anisotropic cosmological solution of bianchi v class . ) the authors acknowledge support from the indo - ukrainian program of cooperation in science and technology . we also thank tarun deep saini for several useful discussions and for producing figure [ fig : sn ] . in this appendix , we present a possible procedure for testing the braneworld model against observations . ( i ) : : @xmath275 [ eq . ( [ eq : first ] ) ] . + ( a ) ; ; @xmath276 @xmath277 \{([hubble1 ] ) , ( [ omega - r1 ] ) } and \{([hubble2 ] ) , ( [ omega - r2 ] ) } are possible . ( b ) ; ; @xmath278 @xmath277 only \{([hubble1 ] ) , ( [ omega - r1 ] ) } is possible . ( ii ) : : @xmath279 [ eq . ( [ eq : second ] ) ] . + ( a ) ; ; @xmath276 @xmath277 only \{([hubble2 ] ) , ( [ omega - r2 ] ) } is possible . ( b ) ; ; @xmath278 @xmath277 only \{([hubble1 ] ) , ( [ omega - r2 ] ) } is possible . let us introduce the quantity @xmath280 that enters the above conditions ( a ) and ( b ) . now we note that , when using ( [ omega - r1 ] ) for determining @xmath256 , we have @xmath281 so that condition ( i ) is satisfied automatically for any sign of @xmath282 . similarly , when using ( [ omega - r2 ] ) for determining @xmath256 , we have @xmath283 so that condition ( ii ) is satisfied automatically for @xmath284 . this means that conditions ( i ) and ( ii ) need not be checked in the procedure . the case of the randall sundrum constraint with @xmath156 is much simpler , since it allows only brane2 model . note that the condition @xmath285 must be satisfied in this case . [ see the explanations above eq . ( [ eq : con - rs ] ) . ] the value of @xmath286 can be obtained from ( [ eq : con - rs ] ) : @xmath287 so that @xmath288 this inequality can be simplified by taking the square of both sides : @xmath289 finally , we get the inequality @xmath290 specify the values of @xmath291 , @xmath292 , @xmath293 , and @xmath196 ( initial @xmath157 s ) . in doing this , note that always @xmath129 by definition , @xmath294 from observations , and @xmath295 by physical consistency [ see eq . ( [ original ] ) ] . in addition , one can assume the universe to be spatially flat : @xmath156 . ( a ) : : to study the brane1 model ( [ hubble1 ] ) with constraint ( [ omega - r1 ] ) : + use ( [ omega - r1 ] ) to calculate @xmath256 , and ( [ hubble1 ] ) to calculate @xmath173 . ( b ) : : to study the brane1 model ( [ hubble1 ] ) with constraint ( [ omega - r2 ] ) : + in specifying the values for the initial @xmath157 s , always choose @xmath296 . use ( [ omega - r2 ] ) to calculate @xmath256 , and ( [ hubble1 ] ) to calculate @xmath173 . ( c ) : : to study the brane2 model ( [ hubble2 ] ) with constraint ( [ omega - r2 ] ) : + in specifying the values for the initial @xmath157 s , always choose @xmath297 . use ( [ omega - r2 ] ) to calculate @xmath256 , and ( [ hubble2 ] ) to calculate @xmath173 . ( d ) : : the case of _ disappearing dark energy _ discussed in sec . [ sec : rs ] and described by brane2 model is studied separately as follows : + set @xmath156 , specify the value of @xmath196 in the range @xmath298 , and choose @xmath292 in the range given by ( [ app : constraint ] ) . now use ( [ eq : con - rs ] ) to find @xmath291 , and then ( [ eq : omsigma ] ) to find @xmath256 . + finally , use ( [ hubble2 ] ) to calculate @xmath173 . a. riess , astron . j. * 116 * , 1009 ( 1998 ) [ astro - ph/9805201 ] ; s. j. perlmutter , astrophys . j. * 517 * , 565 ( 1999 ) [ astro - ph/9812133 ] ; j. l. tonry , _ cosmological results from high-@xmath126 supernovae _ , astro - ph/0305008 . p. bintruy , c. deffayet , and d. langlois , nucl . b * 565 * , 269 ( 2000 ) [ hep - th/9905012 ] ; c. cski , m. graesser , c. kolda , and j. terning , phys . lett . b * 462 * , 34 ( 1999 ) [ hep - ph/9906513 ] ; j. m. cline , c. grojean , and g. servant , phys . lett . * 83 * , 4245 ( 1999 ) [ hep - ph/9906523 ] ; p. bintruy , c. deffayet , u. ellwanger , and d. langlois , phys . b * 477 * , 285 ( 2000 ) [ hep - th/9910219 ] . r. maartens , d. wands , b. a. bassett , and i. p. c. heard , , 041301 ( 2000 ) [ hep - ph/9912464 ] ; e. j. copeland , a. r. liddle and j. e. lidsey , 023509 ( 2001 ) [ astro - ph/0006421 ] ; r. maartens , v. sahni , and t. d. saini , 063509 ( 2001 ) [ gr - qc/0011105 ] ; g. huey and j. lidsey , , 217 ( 2001 ) [ astro - ph/0104006 ] ; v. sahni , m. sami , and t. souradeep , 023518 ( 2002 ) [ gr - qc/0105121 ] . c. deffayet , g. dvali , and g. gabadadze , phys . d * 65 * , 044023 ( 2002 ) [ astro - ph/0105068 ] ; c. deffayet , s. j. landau , j. raux , m. zaldarriaga , and p. astier , phys . d * 66 * , 024019 ( 2002 ) [ astro - ph/0201164 ] . a. albrecht , c. p. burgess , f. ravndal , and c. skordis , phys . d * 65 * 123506 ( 2002 ) [ hep - th/0105261 ] ; a. albrecht , c. p. burgess , f. ravndal , and c. skordis , phys . d * 65 * 123507 ( 2002 ) [ astro - ph/0107573 ] . w. fischler , a. kashani - poor , r. mcnees , and s. paban , jhep * 0107 * , 3 ( 2001 ) [ hep - th/0104181 ] ; j. ellis , n. e. mavromatos , and d. v. nanopoulos , _ string theory and an accelerating universe _ , hep - th/0105206 ; j. m. cline , jhep * 0108 * , 35 ( 2001 ) [ hep - ph/0105251 ] ; x .- g . he , _ accelerating universe and event horizon _ , astro - ph/0105005 .	we explore a new class of braneworld models in which the scalar curvature of the ( induced ) brane metric contributes to the brane action . the scalar curvature term arises generically on account of one - loop effects induced by matter fields residing on the brane . spatially flat braneworld models can enter into a regime of accelerated expansion at late times . this is true even if the brane tension and the bulk cosmological constant are tuned to satisfy the randall sundrum constraint on the brane . braneworld models admit a wider range of possibilities for dark energy than standard lcdm . in these models the luminosity distance can be both smaller and larger than the luminosity distance in lcdm . whereas models with @xmath0 imply @xmath1 and have frequently been discussed in the literature , models with @xmath2 have traditionally been ignored , perhaps because within the general - relativistic framework , the luminosity distance has this property _ only if _ the equation of state of matter is strongly negative ( @xmath3 ) . within the conventional framework , ` phantom energy ' with @xmath3 is beset with a host of undesirable properties , which makes this model of dark energy unattractive . braneworld models , on the other hand , have the capacity to endow dark energy with exciting new possibilities ( including @xmath3 ) without suffering from the problems faced by phantom energy . for a subclass of parameter values , braneworld dark energy and the acceleration of the universe are _ transient _ phenomena . in these models , the universe , after the current period of acceleration , re - enters the matter - dominated regime so that the deceleration parameter @xmath4 when @xmath5 , where @xmath6 is the present epoch . such models could help reconcile an accelerating universe with the requirements of string / m - theory .
as is well known , an ising system quenched from high temperature to low temperature exhibits coarsening in its temporal evolution @xcite . this property holds in any dimension . the voter model @xcite , defined as a purely dynamical system , is identical to the ising model with glauber @xcite or heat bath dynamics in one dimension but its behaviour progressively departs from that of the latter when dimension increases . for instance the two dimensional voter model also exhibits properties similar to coarsening @xcite , though the way the system coarsens differs , in some respect , from that of the two dimensional ising model . measuring the persistence probability of the two models demonstrates more strikingly the difference in behaviour between them . while the persistence probability for the two dimensional ising model at zero temperature decays as @xmath1 with @xmath0 @xcite , it behaves as @xmath2 $ ] for the two dimensional voter model @xcite . in an attempt to understand the differences in behaviour between the dynamics of the two dimensional ising and voter models , we were naturally lead to introduce a whole class of models interpolating between them . the idea stems from the observation that the rules for updating a spin in the voter model resemble those used for the ising model , in the glauber or heat bath algorithms . this class of models is defined in any dimension @xmath3 , but the present study is restricted to two dimensions , where the models depend continuously on two parameters , which may be interpreted as two temperatures . the aim of the present work is to use the freedom allowed by the space of parameters in order to investigate the mechanisms by which coarsening and persistence continuously change when going from the ising model to the voter model . we will see that physical insight is provided by interpreting the two parameters defining the dynamics as two different temperatures , respectively associated to interfacial and to bulk noise . we shall also investigate whether the persistence exponent @xmath4 , which seems constant @xmath5 for the two dimensional ising model when temperature varies @xcite , is a universal exponent for the whole class of models , in the low temperature region of the plane of parameters . in section 2 we first focus our interest on the definition of this class of models . we then give a qualitative description of its dynamics , showing the existence of a critical line between a low and a high temperature region in the space of parameters defining the models ( section 3 ) . section 4 is devoted to the study of persistence in the low temperature region . we first measure the fraction of spins which never flipped up to time @xmath6 , along the special line joining the voter model to the zero temperature ising model . we then discuss the question of persistence at finite temperature ( section 4 ) . finally an appendix is devoted to the dual description of the models in terms of reaction diffusion processes . after this work was completed , we discovered that this class of models had been previously introduced in @xcite , which also contains a determination of the critical line by a finite size scaling analysis in the stationary state . for completeness , we kept the original wording of sections 2 and 3 , adding references to this work where appropriate . let us consider a two dimensional lattice of spins @xmath7 , evolving with the following dynamical rule . at each evolution step , the spin to be updated flips with the heat bath rule : the probability that the spin @xmath8 takes the value @xmath9 is @xmath10 , where the local field @xmath11 is the sum over neighbouring sites @xmath12 and @xmath13\right ) .\ ] ] the functions @xmath14 and @xmath15 are defined over integral values of @xmath16 . for a square lattice , @xmath16 takes the values 4 , 2 , 0 , @xmath17 , @xmath18 . we require that @xmath19 , in order to keep the up down symmetry , hence @xmath20 . note that this fixes @xmath21 . the dynamics therefore depends on two parameters @xmath22 or equivalently on two effective temperatures @xmath23 defining the coordinate system @xmath24 with @xmath25 , yields @xmath26 with @xmath27 . hereafter we shall interpret @xmath28 and @xmath29 as two temperatures , respectively associated to _ interfacial noise _ , and to _ bulk noise_. this should be understood in the following sense . consider the initial configuration where the system is divided by a flat interface into two halves , one half with all spins @xmath30 , and the other one with all spins @xmath31 . then , if @xmath32 , i.e. @xmath33 , this configuration will not evolve in time , neither in the bulk , nor on the interface , since all spins are surrounded by at least three spins of the same value . however , if @xmath34 , then as soon as @xmath35 , spins at the interface will flip , while those in the bulk will not . conversely , if @xmath36 , then spins in the bulk will flip if @xmath37 , while those on the interface will not . ( they will later do so because of the noise coming from the bulk . ) note that a configuration where the system is divided into two halves by a curved interface will always evolve , even if @xmath38 , since @xmath39 . each point in the parameter plane @xmath40 , or alternatively in the temperature plane @xmath41 , corresponds to a particular model . the class of models thus defined comprises as special cases the ising model , the voter and antivoter models , as well as the majority vote model , the description of which follows . the ising model with heat bath dynamics corresponds to choosing @xmath42 where @xmath43 is the usual inverse temperature . hence @xmath44 this corresponds to the line @xmath45 when @xmath43 varies . for example , at zero temperature , one has @xmath38 , hence @xmath33 . the dynamics is therefore only driven by the curvature of the interfaces between domains of equal values of the spin @xcite . while the ising model possesses a well defined energy for which the heath bath dynamics satisfies detailed balance , none of the models outside the ising line shares these properties @xcite , hence these models do not possess a usual equilibrium description , and their definition is purely dynamical . the dynamics of the voter model is defined as follows @xcite . at each time step , the spin to be updated is aligned with one of its neighbours , chosen at random . therefore @xmath46 i.e. @xmath47 this definition corresponds to a model with no bulk noise ( @xmath48 , or @xmath34 ) . the noisy voter model is defined as a simple generalization of ( [ defvot1 ] ) . the spin to be updated is now aligned with one of its neighbours , chosen at random , with a probability @xmath49 @xcite . in other terms @xmath50 hence @xmath51 this corresponds to the line @xmath52 when @xmath49 varies from 0 to 1 . one may also extend the definition ( [ noisyvoter ] ) to @xmath53 , by allowing negative values of the coordinates @xmath54 and @xmath55 , or equivalently letting @xmath56 and @xmath57 be less than @xmath58 . the model thus defined is known as the antivoter model @xcite . finally , for the majority vote model @xcite , spins are aligned with the local field ( i.e. with the majority of neighbours ) with some given probability . more precisely , if @xmath59 @xmath60 and @xmath39 . hence @xmath61 therefore the model corresponds to the line @xmath62 i.e. @xmath63 , when @xmath64 varies from 0 to 1 . = 65truemm = 65truemm = 65truemm = 65truemm figure [ fig_phase ] shows , in the @xmath41 or @xmath65 ) planes , the lines corresponding to the ising , voter and majority vote models . three other lines deserve attention . firstly , the @xmath48 line corresponds to models with no bulk noise ( @xmath34 ) , hence the dynamics is only driven by interfacial noise , defined above . secondly , the @xmath66 line corresponds to models with no interfacial noise ( @xmath36 ) , hence the dynamics is only driven by bulk noise . ( in both cases the effect due to the curvature of the interfaces is always present , as mentioned above . ) for these last models , the local spin aligns in the direction of the majority of its neighbours with probability one , if the local field @xmath67 , i.e. if there is no consensus amongst the neighbours . if there is consensus amongst them , i.e. if @xmath68 , the local spin aligns with its neighbours with a probability @xmath69 . finally the transition line between the low and high temperature regions is discussed below . see reference @xcite for similar definitions . the introduction of the temperatures @xmath28 and @xmath29 , together with their interpretations as given above , is original to the present work . a dual description of the dynamics of the models in terms of reaction diffusion processes is given in the appendix . in this section we investigate the dynamics of phase ordering for the two parameter class of models presented here . for @xmath56 and @xmath57 given , we let the system evolve , starting from a random initial configuration . this study demonstrates the existence of a critical line , in the @xmath40 or @xmath41 planes , between a low temperature region where clusters ( or domains ) grow indefinitely , and a high temperature region where one only observes fluctuations at a finite scale . it also provides a visual illustration of the role of the two temperatures @xmath28 and @xmath29 , respectively associated to interfacial and to bulk noise . two methods for updating the spins are at our disposal . with sequential updating , time is continuous and each spin evolve independently of others , at times chosen randomly according to a poissonian law . the normalization is chosen such that each spin evolves once in a mean unit of time . therefore , in a simulation of a lattice of @xmath70 spins , one has to choose randomly a spin and update it , @xmath70 times per time unit . instead of using sequential updating , we adopt a slightly different procedure for our simulations , namely a parallel updating of the spins . the lattice is divided into two odd and even sublattices , and , during a unit of time , the two sublattices are visited in turn , with a systematic update of their spins . this division is needed in order to avoid undesirable effects due to the simultaneous update of neighbouring sites . this process allows a functional parallelization on the computer . spins are represented as single bits , and it is possible to arrange the spins into computer words , such that all spins of a word can be updated simultaneously by global logical operations . this method is numerically efficient , allowing for long simulations on large lattices . in figures 27 , we have selected snapshots of the same fragment @xmath71 of a @xmath72 lattice with periodic boundary conditions , for three times @xmath73 , 64 and 512 , and six different points in the parameter space of the models . figures [ evi0 ] , [ evi97 ] depict , as reference views , the well known appearance of domain growth for the ising model , respectively at zero temperature , and at @xmath74 . the ising critical point corresponds to @xmath75 . for @xmath76 clusters grow as @xmath77 , while at @xmath78 , they only grow as @xmath79 , where the dynamical exponent @xmath80 . in the high temperature phase , clusters grow until their sizes reach the equilibrium correlation length . = = this distinction between a low temperature and a high temperature phase can be generalized to all models of the two dimensional parameter plane . the transition line between the low and high temperature regions , depicted in figure [ fig_phase ] , was located by a systematic investigation of the parameter plane . for instance , figures [ evv76 ] , [ evv75 ] , and [ evv72 ] correspond to an exploration of the line @xmath48 , with @xmath56 decreasing away from the zero temperature ising case ( @xmath38 ) . on this line no thermal fluctuations occur inside the clusters , so their interiors remain forever uniformly black or white . when @xmath56 decreases , the boundaries between domains take progressively a fractal appearance , in accord with the fact that interfacial noise increases along the line . this is particularly visible on figure [ evv76 ] , which illustrates the case of a model close to the voter model , in the low temperature region ( @xmath81 , @xmath48 ) , and on figure [ evv75 ] , for the voter model itself ( @xmath82 , @xmath48 ) . again one observes growing clusters , with a characteristic size behaving as @xmath77 . a similar behaviour is observed for all models such that @xmath83 . = = figure [ evv72 ] illustrates the case of a model located in the high temperature region , near the voter model ( @xmath84 , @xmath48 ) . in parallel to what is observed for the ising case above @xmath78 , clusters grow until their sizes reach the correlation length . the transition to the low temperature region along the line @xmath48 is therefore only induced by the interfacial noise . = we also display in figure [ evm93 ] the evolution observed for the majority vote model with ( @xmath85 ) . the transition takes place for @xmath86 , in agreement with the value ( @xmath87 ) given in @xcite . = finally exploring the line @xmath66 , we observe a transition between the low and high temperature regimes for @xmath88 , in agreement with the value ( @xmath89 ) given in @xcite . figure [ evr84 ] depicts snapshots of the evolution for the model with ( @xmath66 , @xmath90 ) . see @xcite for a determination of the critical line by a finite size scaling analysis in the stationary state . in this section we address the question of persistence in the low temperature region of the ( @xmath91 ) parameter plane depicted in figure [ fig_phase ] . we first present numerical measurements of the fraction @xmath92 of spins which never flipped up to time @xmath6 , for models with no bulk noise , i.e. for @xmath34 or @xmath48 . the case of persistence at finite temperature , i.e. in the present context for @xmath37 , will be discussed in section [ pert ] . we measured persistence in the low temperature region ( @xmath93 ) of the line @xmath48 . figure [ perss ] depicts the results of simulations for times up to 5000 , on a lattice @xmath94 . the fraction @xmath92 of spins which never flipped up to time @xmath6 behaves , in the early stage , as @xmath95 then one observes a crossover toward the power law @xmath96 with an exponent @xmath97 which seems independent of @xmath98 . this behaviour therefore interpolates between the two ` pure ' cases , namely the zero temperature ising model ( @xmath38 ) where ( [ perslaw2 ] ) is observed with the same exponent @xmath97 @xcite , and the voter model , for which ( [ perslaw1 ] ) holds @xcite . = 80truemm the origin of these two ` pure ' behaviours can be simply traced back to the two driving forces of the dynamics in absence of bulk noise , i.e. respectively to interfacial noise for ( [ perslaw1 ] ) , and to the curvature of interfaces for ( [ perslaw2 ] ) . indeed , consider first the case of the ising model at zero temperature ( @xmath33 ) , for which the dynamics is only driven by the curvature of interfaces . as a consequence , the only flipping mechanism for a spin is the slow sweeping mode of the interfaces , hence persistent spins deeply buried inside large clusters disappear slowly , according to ( [ perslaw2 ] ) . consider now the case of the voter model ( @xmath99 ) , starting from a configuration with a planar interface separating two regions of opposite spins . this interface thickens gradually , because of interfacial noise , taking ever more a fractal appearance . at large times it becomes difficult to speak any longer of an interface , the thickness of which should be considered as infinite . the decay of persistent spins is much faster and follows ( [ perslaw1 ] ) . this analysis allows the following interpretation of the results given above for persistence of a generic model along the @xmath48 line , and @xmath100 . in the early stages of coarsening , domains begin to grow and develop their interfaces . as a consequence , the fraction of persistent spins in the interface region is expected to decay more rapidly than if they were deep inside a cluster , following ( [ perslaw1 ] ) since this is the dominant process . the duration of this early stage is related to the thickness of the interfaces , which vanishes for the ising model where smooth interfaces are observed , and diverges for the voter model where a fractal appearance of the interfaces is observed . after this first stage , the fraction of persistent spins will decay according to ( [ perslaw2 ] ) . therefore , along the line @xmath48 ( @xmath34 ) , the resulting law for the decay of @xmath92 will be a weighted sum of ( [ perslaw1 ] ) and ( [ perslaw2 ] ) . the amplitude obtained by fitting the asymptotic algebraic tail ( [ perslaw2 ] ) is naturally interpreted as the fraction of persistent spins which are deep inside the clusters . this fraction vanishes when one approaches the voter model , with an apparent power law @xmath101 , with @xmath102 . this result is consistent with the fact that the voter model is critical , according to the criterion of section [ phase ] . note that , for any model such that @xmath100 , the prescription of suppressing the bulk noise by forbidding flips of a spin @xmath30 ( resp . @xmath31 ) surrounded by four spins with the same value @xmath103 ( resp . @xmath104 ) , leads to an algebraic decay of the persistence probability . this corresponds to a projection of the model onto the line @xmath66 . for a two dimensional ising system in presence of thermal noise ( i.e. in the present context , for @xmath69 ) , the number of spins which did not flip up to time @xmath6 decays exponentially . a more satisfactory quantity to investigate is the fraction @xmath92 of space which remained in the same phase , coming back to the original definition of persistence @xcite , where ` remaining in the same phase ' means remaining ` dry ' , i.e. unswept by an interface . the prescription given in @xcite to determine whether a given spin ` remained in the same phase ' , leads again to algebraic decay of @xmath92 , the fraction of ` persistent ' spins , when @xmath76 . further works @xcite seem to indicate that the persistence exponent @xmath4 for the two dimensional ising model is constant in the low temperature phase , and approximately equal to 0.22 . = 80truemm coming back to the class of models under investigation , one may therefore wonder what is the value of the persistence exponent outside the ising line , in the low temperature region . here we use the definition of @xcite , adopting the extensions given in @xcite , in order to measure persistence in the low temperature region of the parameter plane . we consider three copies of the system , a , b and c , submitted to the same noise ( @xmath91 ) . the initial condition for a is random , while it is ordered for b ( @xmath105 ) , and c ( @xmath106 ) where @xmath107 labels the sites . a spin @xmath108 of copy a is said to be persistent up to time @xmath6 , if it experienced thermal fluctuations only , i.e. if its history was either that of @xmath109 or that of @xmath110 . hence @xmath111 measures the fraction of persistent spins in one of the two phases @xmath112 @xcite . we also checked , using the determination of the interfaces proposed in @xcite , the equivalence between the definition of @xmath92 given by ( [ rtnew ] ) and that obtained by considering a spin as persistent if it was not swept by an interface . figure [ tous ] depicts the results . the exponent @xmath4 seems constant , with a value @xmath5 , for all models on the low temperature side of the critical line that we considered . it is nevertheless hard to conclude , on the basis of numerical measurements . the decrease of persistence at criticality is much faster . a possible explanation is that , as for the voter model ( see [ p21 ] ) , the thickness of interfaces diverges , leading to a behaviour similar to ( [ perslaw1 ] ) . the aim of the present work was to initiate the study of coarsening and persistence for a class of models interpolating between the voter and ising models , introduced previously in @xcite . in our opinion , the introduction of two temperatures @xmath28 and @xmath29 , respectively associated to interfacial and bulk noise , gives a useful interpretation of the phenomena observed when varying continuously the parameters defining this class of models . the present analysis seems to indicate that @xmath0 is the universal persistence exponent for all models in the low temperature phase . the situation on the critical line is less clear , and possibly related to the divergence of the interfaces between domains at criticality . in this respect it would be desirable to find a way of computing the equation of the critical line in the @xmath40 parameter plane . a more ambitious goal would be to compute @xmath4 for the two dimensional zero temperature ising model , in the dual reaction diffusion framework given in the appendix , using the field theoretical methods of @xcite . let us finally mention extensions to @xmath113 or to the potts model as possible further studies . we hope to address some of those points in the future . we wish to thank m howard and j - m luck for interesting discussions . in this appendix we show that the dynamics of this class of models possesses a dual description in terms of reaction diffusion processes for a set of random walkers moving backward in time . this description , beyond its intuitive value , is the starting point of an analysis of the temporal evolution of the system by means of diagrammatic expansions @xcite . we first review the case of the voter model @xcite . let us define the random variable @xmath114 , hereafter also called spin . the value of the spin @xmath115 at time @xmath6 and site @xmath116 can be traced back to the value of its ancestor @xmath117 at time @xmath118 , located on site @xmath119 . the idea is to follow the line of continuity of the value of the spin , backward in time , from ( @xmath120 ) to ( @xmath121 ) . indeed , one first follows the line of constant @xmath122 backward in time , from @xmath123 until the last updating of site @xmath116 is met . at this time , the spin had chosen the value of a neighbouring site . moving to this site one again follows the line of constant @xmath122 backward in time . proceeding until @xmath118 , one thus performs a random walk , backward in time , from ( @xmath124 ) to ( @xmath125 ) such that @xmath126 . it is easy to see that all spins at time @xmath6 possess an ancestor at time @xmath118 , but that conversely not all sites at time @xmath118 are ancestors of spins at time @xmath6 . indeed , starting from @xmath127 spins @xmath128 , located on sites @xmath129 , at time @xmath6 , the same reasoning leads to the consideration of @xmath127 random walkers starting from sites @xmath129 at time @xmath6 , and going backward in time . when two such walkers meet , they coalesce because the corresponding sites have a common ancestor from which they inherit the common value of their spin . = 50truemm hence the determination of the values of these @xmath127 spins is equivalent , for the voter model , to knowing the history of @xmath127 independent coalescing random walkers , starting from sites @xmath130 , and moving backward in time . an example is displayed in figure [ fig5 ] . the final stage , at time @xmath118 , contains @xmath131 walkers ( @xmath132 ) at sites @xmath133 . since @xmath134 , we conclude that the sum of the probabilities of all these processes starting from the @xmath127 points @xmath135 gives the correlation function @xmath136 . one can easily generalize the reasoning to the case of spins on different sites @xmath129 , at different times @xmath137 . for the voter model , the definition ( [ defvot1 ] ) can be rewritten as @xmath138 this means that for a two dimensional random walk the elementary process consisting of a jump from site @xmath116 to one of its neighbouring sites @xmath119 has probability @xmath139 . similarly ( [ noisyvoter ] ) leads to @xmath140 the interpretation of which is given below . the generalization to a generic model ( @xmath91 ) is now straightforward . by its very definition one has @xmath141 \\ \quad + \frac{1}{2}\left [ \tau _ { y_{1}}\tau _ { y_{2}}\left ( 1-\tau _ { y_{3}}\right ) \left ( 1-\tau _ { y_{4}}\right ) + \text{5 perm . terms}\right ] \\ \quad + ( 1-p_{1})\left [ \tau _ { y_{1}}\left ( 1-\tau _ { y_{2}}\right ) \left ( 1-\tau _ { y_{3}}\right ) \left ( 1-\tau _ { y_{4}}\right ) + \text{3 perm . terms}\right ] \\ \quad + ( 1-p_{2})\left ( 1-\tau _ { y_{1}}\right ) \left ( 1-\tau _ { y_{2}}\right ) \left ( 1-\tau _ { y_{3}}\right ) \left ( 1-\tau _ { y_{4}}\right ) \end{array}\ ] ] this expression can be rewritten as @xmath142 equation ( [ mdef ] ) is the generalization of ( [ vvdef ] ) and can be expressed as the sum of four elementary processes , namely : * the random walker disappears ( with a weight @xmath143 ) , * the random walker makes a step to a neighbouring site ( with a weight @xmath144 ) , * the random walker splits into two walkers located on different neighbouring sites ( with a weight @xmath145 ) , * the random walker splits into three walkers located on different neighbouring sites ( with a weight @xmath146 ) . hence the coalescing random walks considered above for the voter model have now also the possibility of annihilating , or of branching into two or three new walks . note that the weight @xmath147 vanishes on the line of the majority vote model , while the weight @xmath148 vanishes on the line of the noisy voter model . let us underline the fact that the weights corresponding to these elementary processes can be negative . therefore these drawings should be seen as diagrams , and not as the actual paths of real random walkers . they may indeed be used for actual diagrammatic computations . for example , near the voter model , one can perform a double series expansion of the correlation functions in the ` coupling constants ' @xmath143 and @xmath145 , in order to compute the rate of the exponential decay of persistence , in good agreement with numerical measurements @xcite .	we study the dynamics of a class of two dimensional stochastic processes , depending on two parameters , which may be interpreted as two different temperatures , respectively associated to interfacial and to bulk noise . special lines in the plane of parameters correspond to the ising model , voter model and majority vote model . the dynamics of this class of models may be described formally in terms of reaction diffusion processes for a set of coalescing , annihilating , and branching random walkers . we use the freedom allowed by the space of parameters to measure , by numerical simulations , the persistence probability of a generic model in the low temperature phase , where the system coarsens . this probability is found to decay at large times as a power law with a seemingly constant exponent @xmath0 . we also discuss the connection between persistence and the nature of the interfaces between domains . # 1#1 epsf [ coarsening and persistence in a class of stochastic processes ] _ submitted for publication to _
the stellar initial mass function ( imf ) is fundamentally important to understanding both stellar populations and galaxies . the milky way ( mw ) imf was originally characterized as a power - law mass - distribution , @xmath1 , with @xmath2 @xcite , and subsequently refined to flatten at lower masses ( @xmath3 ; @xcite ) . whether or not the mw imf describes stellar populations elsewhere in the universe can not yet be said through direct star counts . there have been some indirect observational hints of imf variations , and ample theoretical motivation for these , but no broadly convincing evidence has emerged ( cf . @xcite ) . this situation has recently changed , with a flurry of studies of early - type galaxies ( etgs ) turning up indirect evidence for systematic imf variations . these studies use models of stellar population synthesis ( sps ) to fit integrated - light data ( broad - band colors and spectroscopic features ) , and fall into two broad categories : `` pure '' sps , and `` hybrid '' sps+gravitating mass analyses . the pure analyses rely on spectral lines that are differentially sensitive to giant or dwarf stars . these include the tio feature at 6130 , the na i doublet near 8190 , the ca ii triplet near 8600 , the wing ford [ fe / h ] band at 9915 , and the ca i line at 10345 (e.g . , @xcite ; ( * ? ? ? * hereafter cvd12 ) ; @xcite ) . the hybrid analyses assume an imf and infer a stellar mass - to - light ratio using a more conventional sps approach based on colors , and age- and metallicity - sensitive spectral lines . estimates of are also derived using dynamics or gravitational lensing . comparison of the independent results then yields a revised imf ( e.g. , @xcite ) . one may characterize a revised imf through its relative to a mw - disk imf , @xmath4 ( the `` mismatch parameter '' ) , where for reference we adopt the chabrier imf . remarkably , almost all the above studies found `` heavy '' imfs ( @xmath5 ) for the most massive etgs . less massive etgs , and spiral galaxies , appear to have `` normal / light '' imfs ( @xmath6 ; e.g. , @xcite ) and the bulge components of spirals may also have a similar mass dependence to etgs ( @xcite ; ( * ? ? ? * f+10 ) ; ( * ? ? ? * d+12c ) ) . these imf findings are both potentially revolutionary , and highly controversial , and demand further investigation . in particular , with hybrid analyses there are lingering questions about degeneracies associated with the distribution of non baryonic dark matter ( dm ) . the time is also ripe to inventory the results to date , and see if the apparent emerging consensus holds up under quantitative , systematic comparison which could provide pressing motivation for understanding the physical origins of the trends . comparisons were made for some hybrid analyses @xcite , but nor for the pure sps work . in this letter we carry out such an inventory , while presenting our own novel results for a large sample of etgs for reference , following the dynamical@xmath7sps analyses of ( * ? ? ? * hereafter t+12 ) . we focus on the trends in with central stellar velocity dispersion , , and discuss some implications for the central dm fraction . is widely considered as crucially connected to galaxy evolution , and unlike , is relatively independent of the bandpass , and of . we describe our data and analysis methods in section [ sec : data ] . we present our results in section [ sec : results ] and make literature comparisons in section [ sec : lit ] . we summarize the conclusions and outlook in section [ sec : conclusions ] . we apply a combination of sps and stellar dynamical models to a sample of @xmath8 giant etgs , in the redshift range of @xmath9@xmath10 , drawn from the spider project ( @xcite ) . our data include optical@xmath7near - infrared photometry [ from the sloan digital sky survey ( sdss ) and the ukirt infrared deep sky survey - large area survey ] , high - quality measurements of galactic structural parameters ( effective radius and srsic index @xmath11 ) , and sdss central - aperture velocity dispersions @xmath12 . the sample galaxies are defined as bulge dominated systems with passive spectra while late - type systems are efficiently removed through the sdss classification parameters based on the spectral type and the fraction of light which is better described by a @xcite profile ( see t+12 for further details ) . the structural parameters are measured using 2dphot ( @xcite ) and are found to be significantly different from the sdss estimates ( @xcite ) . the sample is @xmath13 complete at a stellar mass of @xmath14 , which corresponds to @xmath15 . the sps - based values were derived by fitting @xcite models to the multi - band photometry , assuming a chabrier imf . these results have been shown to be consistent with independent literature ( e.g. , mpa masses in @xcite ) , while possible systematics in stellar mass ( ) estimates are discussed in @xcite and t+12 . despite the well known age - metallicity degeneracies in photometric data , these conspire to keep the stellar mass - to - light ratio well constrained , with scatter of @xmath16 dex . the agreement is also excellent when our colour derived masses are compared with spectroscopic estimates . the largest systematical uncertainty in our analysis comes from our ignorance on the imf shape , which can produce variations of the stellar m / l of a factor as large as @xmath17 . for this reason imf is a key issue in stellar population analysis and the central topic of this paper . our dynamical - mass estimates use spherical isotropic jeans equations fitted to the @xmath12 data . we also extend the t+12 analysis using two - component mass models : a srsic - based stellar distribution following the @xmath18-band light , and a standard dm profile . for the latter we adopt a series of plausible assumptions ( cf . @xcite ) , as the data do not allow us to constrain both components simultaneously . our dm models are based on the ( * ? ? ? * hereafter nfw ) profile , with an adjustable degree of baryon - induced adiabatic contraction ( ac ) . for the virial mass and concentration ( @xmath19 , @xmath20 ) , we adopt mean trends for a wmap5 cosmology @xcite , for the relation we used ( * ? ? ? * m+10 hereafter ) . each galaxy s mass model then has one free parameter , , plus optional ac ( * ? ? ? * g+04 hereafter ) , providing our no - ac - nfw and ac - nfw base - models . we explore the sensitivity of our results to these assumptions by doing the analyses with the following alternatives : a ) ac recipes of varying strengths ( ( * ? ? ? * b+86 ) ; ( * ? ? ? * a+10 ) ) ; b ) with a fixed value for the entire sample : @xmath21 , @xmath22 , or @xmath23 ; c ) wmap3-based @xmath20@xmath19 relation @xcite ; d ) no dm is present ; e ) mild kinematic anisotropy , with @xmath24 or @xmath25 ; f ) @xmath20@xmath19 relation altered to mimic a warm dark matter ( wdm ) cosmology , assuming different particle masses ( * ? ? ? * s+12a hereafter ) . to study the mean trends of with velocity dispersion , we construct `` average '' galaxies by dividing our sample into 10 -bins , for which we compute median values of @xmath26 , @xmath11 , , and . we show these values , and their 2575 percentile scatter , in figure [ fig : fig1 ] . for each -bin and a given dm model , we solve the radial jeans equation for the value which matches the observed . our final analysis - products , for each galaxy bin and mass model , will be the sps - determined @xmath27 , the dynamically - determined , the inferred , and the inferred central dm - fraction , @xmath28 . for homogeneity , we convert to ( the value at ) , using the best - fitting relation in @xcite ( also done for the literature results later ) . our main imf results are shown in figure [ fig : spider ] . the two thick black curves correspond to our standard no - ac - nfw ( solid ) and ac - nfw ( long - dashed ) models , and the suite of alternative models are also plotted , as labeled . it is clear that the overall normalization of is degenerate with the adopted halo model , as dm can be traded against stellar mass . this degeneracy is most severe when allowing for uncertainties in the halo response to baryons ( more so than with the virial mass assumptions ) . however , for a given flavor of halo model , there is always a strong correlation between and ( the trend is weaker ) . we quantify this correlation for each model with a log log fit , reporting the best - fit parameters in table [ tab : tab1 ] ( also plotted in figure [ fig : spider ] for the two reference models . ] ) ; the typical relation is @xmath29 . .best - fit parameters for the relation @xmath30 , for model suite ( see main text ) . [ cols="<,^,^ " , ] other models of potential interest are low - density dm cores @xcite , and alternatives to dm ( e.g. , @xcite ) . these would imply similar slopes to our limiting no - dm model . the no - dm assumption produces the uppermost curve in fig . [ fig : spider ] , and is qualitatively consistent with a slightly _ expanded _ nfw model . the same results , using the same dataset , but somewhat different mass modeling assumptions , have been anticipated in @xcite . one could alter these slopes with additional model - tuning ; e.g. , with strong - ac at high- and halo expansion at low- , one could completely flatten out the trend . on the other hand , decreasing the ac at the high and increasing the strength of the ac toward lower would make the trend even steeper . we currently have no a priori motivation for either direction for the ac variations , thus we can not argue in favor of a imf universality for the former case nor for a very strong imf non universality in the latter one . we next examine the overall imf normalization , with the no - ac - nfw and ac - nfw cases bracketing the most plausible range of models . for reference , we show predictions for several standard imfs ( salpeter , kroupa , chabrier ) . we note that at a fixed imf , age , and metallicity , the and values are uniquely predicted , but the reverse is not true . a given result can imply multiple imf solutions , particularly if one allows for mass - functions more complicated than a pure power - law . if we adopt a no - ac - nfw model , a mw - like imf is implied for low - galaxies , and a salpeter imf for high- . for ac - nfw , the low- galaxies have sub - mw imfs , and the high- ones have imfs intermediate to kroupa and salpeter . in all cases , extremely bottom - heavy imfs ( assumed single power - law , @xmath31 ) are ruled out , on average . even if one assumed _ no _ dm , such imfs would violate the dynamical constraints on the overall mass - to - light ratio . although resolving the remaining imf dm degeneracy will require more extensive analysis , we carry out a simple exercise to provide initial clues , inspired by @xcite . as in that paper , we select the galaxies with mean central stellar surface - densities @xmath32 2500 @xmath33 pc@xmath34 , and analyze them the same as the full sample . the rationale is that such galaxies are the most star - dominated , and the least sensitive to dm uncertainties . the results are shown by the gray curves in figure [ fig : spider ] , where as expected , the model curves are closer together . the implied normalization is fairly low similar to the full - sample ac results . we may then apply this imf result to the full sample if we assume no additional systematic @xmath35 correlation ( cf . @xcite ) . although this paper is primarily concerned with the imf , we briefly examine some implications for the central dm content . our main mass models are , by construction , fully consistent with current expectations for @xmath36cdm halo profiles , while also agreeing with the observations for a plausible imf range ( @xmath37 0.52.0 ) . we show the implied within 1 in figure [ fig : fdm ] . we find fairly universal values of @xmath38 and @xmath39 for the no - ac - nfw and ac - nfw models , respectively . these results are not altered appreciably in the alternative models explored above . note also that the high-@xmath35 test above prefers the ac - nfw model . the figure also shows that if we adopted a constant imf , then we would infer a strong increase of with . such behavior has been invoked as the driver for the `` tilt '' of the etg fundamental plane ( see @xcite and references therein ) , but now with renewed comparison to realistic dm halo models , we find that the tilt is driven in part by the imf . put differently , the observed relation is too steep to explain through standard dm models , and requires an additional factor . , vs. , for the spider sample with different modeling assumptions . the solid and dashed black curves are our default no - ac - nfw and ac - nfw models . the gray curves are for fixed chabrier imf , with isothermal and constant- models , bracketing the range of non - parametric models ( solid and dotted curves , respectively ; see t+12 ) . with standard @xmath36cdm halos and a variable imf , is constant or mildly decreasing with ; with a fixed imf , is strongly increasing ( cf . @xcite figure 16 ) . , scaledwidth=49.0% ] we now compare our spider - based results with an inventory of other literature results for etgs in figure [ fig : lit ] and late - type galaxies in figure [ fig : lit2 ] . we discuss the results for etgs in figure [ fig : lit ] , starting with those studies that used similar hybrid approaches , comparing sps - mass estimates to total mass using dynamics or lensing . rather than exhaustively comparing all such results , we will focus on the studies that explicitly derived for large samples of low-@xmath40 etgs . the study most closely related to ours is @xcite , who analyzed sdss data for etgs ( colors and , with sps@xmath7jeans modeling ) . their sample was much larger but without the srsic models and near - infrared photometry from spider . their results for a no - ac - nfw model are shown as a dotted orange curve in panel ( a ) , which is reassuringly similar to our no - ac - nfw result , with the @xmath0 30% residual difference in illustrating the level of systematic uncertainties for a fixed dataset and method . although a conclusive answer on the origin of this discrepancy is not available , we have found that shallower light profiles ( as may be equivalent to the combination of the @xmath41 and @xmath42 profiles in @xcite ) produce larger . finally , the solid line shows their refined result from a multi - parameter fit : they find dm halos that are slightly expanded , and consequently a slightly heavier imf . in panel ( b ) we show results from the atlas@xmath43 survey of nearby etgs , using both spectroscopically - based sps models and detailed two - dimensional jeans dynamical analyses ( * ? ? ? * where we show their no - ac - nfw results ) . as with spider , the atlas@xmath43 project found that the overall normalization was degenerate to the dm assumptions , but the trend with was robust . direct comparison in figure [ fig : lit ] reveals that these results are consistent with spider within the errors , which is also the case for an earlier study of etgs with sauron @xcite and for an analysis of high-@xmath40 galaxies @xcite . panel ( c ) shows results from @xcite and @xcite , who carried out spectroscopic sps and detailed orbit - modeling of etgs found in clusters ( including coma ) , using a variety of mass models ( constant- , no - ac - nfw , and cored halos ) . we see no systematic difference between the results obtained with different mass models . a systematic offset in between the two studies is found , but overall , the results are consistent with our no - ac - nfw results , both in amplitude and in slope . panel ( d ) shows gravitational lensing results , primarily from the slacs survey of etgs @xcite , using no - ac - nfw models as well as color - based sps models and stellar - dynamics constraints . their normalization agrees well overall with ours , but their slope is somewhat steeper . our results , along with the four hybrid studies from the literature , all suggest that the imf of etgs varies from mw - like at low to salpeter - like at high , modulo some lingering uncertainties from the dm imf degeneracy . in addition to the large - sample studies , we examine a few key single - galaxy results . these include extended kinematics data with nfw - based modeling @xcite , as shown in panel ( c ) , and the `` jackpot '' double lens @xcite , in panel ( d ) , and all look consistent with the trend from spider . our culminating comparisons are with a completely different set of results , based purely on modeling of imf - sensitive spectral lines ( see section [ sec : intro ] ) . most of these lines are susceptible to degeneracies with elemental abundances ( e.g. , sodium or calcium ) , and we consider only those studies that have directly accounted for such effects . we first show in panel ( e ) the results from cvd12 . they fitted spectral features across a wide wavelength range , focusing on the imf indicators na i , ca ii , and the wing ford band , while adopting a broken - power - law imf form and fitting for the relevant elemental abundances . their inferred values turn out to agree well with both the normalization and the trend versus from the spider results ( for the no - ac - nfw models in particular ) . recalling that the high values from hybrid studies could be due to either a bottom - heavy ( extra dwarfs ) _ or _ a top - heavy ( extra remnants ) imf , the cvd12 results agree with only the first solution . note that the apertures probed here are different : @xmath44 for cvd12 , and 0.30.7 for spider ( decreasing with ) . the close agreement of the results on average thus implies that the imf does not vary spatially on these scales , or that the ac model is the correct solution , and decreases with galactocentric radius , which is plausible ( e.g. , @xcite ; see also fig . 13 of cvd12 ) . panel ( f ) shows results from @xcite , who studied a large sample of coma - cluster etgs . they used the same sps models as cvd12 to fit near - infrared spectroscopic line - indices , analyzing the wing ford band and the ca i line separately . the former line was susceptible to uncertainties in the na abundance , but not the latter . we have converted their results to inferred by straightforward interpolation between the chabrier , salpeter , and @xmath45 models in their figure 10 . the final results are somewhat noisy and uncertain , but imply an overall salpeter - like normalization , and no obvious trend with . this agrees with all the aforementioned results at high- , but not at low- . it is possible that these coma - cluster galaxies are genuinely different . as environmental classification is part of the spider dataset ( @xcite ; @xcite ; @xcite ) , we have investigated the impact of environment on our results and found very little effect on . but we also have very few cluster galaxies in our sample and a definitive comparison with those coma - cluster results is not possible . @xcite analyzed nad , na i , and tio@xmath46 lines in sdss spectra of etgs with @xmath47 km s@xmath48 , comparing to the same sps models as cvd12 . they inferred a salpeter - like imf at low , which is consistent with spider . at high , they inferred @xmath49 , which would imply @xmath50 and violate the total mass constraints both from spider and from the lens galaxy that these authors also studied . this conflict suggests either further work is needed on the line - index modeling , or the imf shape deviates from a simple power - law . an ideal comparison with our spider results would be the work of @xcite , who analyzed na i and tio line - strengths from the spider parent dataset . although they did not provide values that we could compare to our results , their illustrative trends versus ( at fixed metallicity and age ) are qualitatively similar . they also demonstrated that the inferences in their approach could depend strongly on the detailed shape assumed for the imf . it would also be fascinating to see whether or not bulges / disks follow the same trends as etgs . unfortunately , although direct inferences from star counts are possible in the milky way and nearby galaxies , the literature in the field is not sufficient to investigate with accuracy imf variations with mass or , if any . however , it appears that a general consensus is arising within the community , which points to kroupa / chabrier type . in general , imf looks similar in the field , dense massive clusters and diffuse low density star - forming regions , with some deviations observed in a handful of other regions ( see @xcite and @xcite for a review of the main results ) . analysis of masses above @xmath51 have been performed in nearby galaxies ( for instance the irregular lmc , the dwarf smc and the spiral m33 ) , ruling out strong imf variations . similarly , starburst galaxies and their embedded young massive clusters imply no imf variation and no influence of the local environment . such analyses provide only limited information on a restricted sample of galaxies . however , in the recent years , similarly to the etgs case , studies of variations with galaxy mass have been accumulating . for example , @xcite found that bulges of spirals showed anti - correlations between and ca ii line - strengths , similar to etgs , but they could not determine if this was an imf effect . we have attempted to investigate further , using inferences for bulges and disks from the literature ( e.g. , @xcite ; f+10 ; ( * ? ? ? * b+12 ) ; cvd12 ; d+12c ) . this comparison is shown in figure [ fig : lit2 ] . using bulges from @xcite we find that correlates with stellar mass , giving hints of an etg - like correlation , but we also find the same for the disks , which would be peculiar . note that initial results on dynamical masses of nearby spiral disks suggest @xmath52 ( @xcite ) . we show the results for a sample of five massive spiral galaxies from @xcite , who found a ) stellar mass carrying out a photometric sps and b ) an independent imf estimate , using strong gravitational lensing and gas kinematics . the apparent scatter for both bulges and disks is enormous , suggesting more work is needed to address the systematic errors , and to understand any additional trends with detailed morphology . however , the imf normalization is higher in the bulges than in the disks . a similar analysis has been performed in b+12 ( updating the results in @xcite ) where gravitational lensing , gas rotation curve and stellar kinematics for the lens galaxy sdss j2141 - 0001 are used . they found a which is fully consistent with an intermediate - normalization imf ( in between a salpeter and a chabrier imf ) , but taking into account the expected cold gas fraction ( which had not been included in the fitting procedure ) they find lower , agreeing with a chabrier / kroupa imf . a bottom - light imf is also found by f+10 , who analyzed the strong lensing features of the einstein cross ( q2237 + 0305 ) , and @xcite using spectral lines in the nuclear bulge of m31 . we have analyzed the dynamics and stellar populations of a large sample of etgs from the spider project , and found compelling evidence for heavier imfs in the central regions of higher - galaxies . the imf mismatch relative to chabrier is @xmath53 0.51.1 at @xmath54 125 km s@xmath48 , and @xmath0 1.21.8 at at @xmath0 250 km s@xmath48 . the trend is robust to a wide set of modeling assumptions , and accounts for much of the tilt in the fundamental plane . the distribution of dm is degenerate with the overall imf normalization and difficult to constrain , and therefore we have assumed that any halo contraction is invariant with . some ways to break the degeneracy between imf and halo contraction are to a ) analyze cases with extended velocity dispersion or x - ray emission profiles ( e.g. , @xcite ) , or b ) incorporate complementary data from strong / weak gravitational lensing ( e.g. @xcite ) . however , we have argued that the only way to preserve the imf universality is to allow for halo contraction at high- and halo _ expansion _ at low- . we have performed the first general inventory of imf results from a variety of studies in the literature , using both pure and hybrid techniques , and found that these generally agree well with our spider results . there is remarkably widespread agreement on a salpeter - like imf for massive etgs ( @xmath55 200 km s@xmath48 ) . at lower , the data are still somewhat limited and the different results have not yet converged , but most of the studies point to mw - like imfs . agreement on a super - salpeter imf in the most massive galaxies ( @xmath55 275 km s@xmath48 ) is also not yet universal . these results appear to be fully compatible with underlying @xmath36cdm halos . however , more detailed conclusions about halo contraction or expansion are still elusive , and the data on galaxy centers do not clearly rule out alternative dm models once the variable imf is accounted for . more work is clearly needed to understand the systematics in the different analyses ; to build up better statistics on a wide range of galaxy types , environments , and redshifts ; and to determine which parameters correlate best with imf variations ( e.g. , metallicity or starburst intensity ; cvd12 ; @xcite ) . it may also be particularly helpful to venture beyond the centers of galaxies , using data from a wide baseline in radius to help break the imf dm degeneracies . it appears we are nearing convergence on determining _ what _ the basic components of galaxies are ( distributions of stars and dm ) . the next challenge will be to understand _ why _ these arrive at their distributions . what drives the power - spectrum in cloud fragmentation and star formation ? how do baryonic processes interact with and re - shape their surrounding dm halos ? we thank the referees for their constructive comments . we thank matt auger , michele cappellari , reinaldo de carvalho , aaron dutton , francesco la barbera , and tommaso treu for helpful discussions . we also thank charlie conroy for helpful discussions and for providing us with his updated results in tabular form . ct was supported by the swiss national science foundation and the forschungskredit at the university of zurich . ajr was supported by the national science foundation grant ast-0909237 .	given a flurry of recent claims for systematic variations in the stellar initial mass function ( imf ) , we carry out the first inventory of the observational evidence using different approaches . this includes literature results , as well as our own new findings from combined stellar - populations synthesis ( sps ) and jeans dynamical analyses of data on @xmath0 4500 early - type galaxies ( etgs ) from the spider project . we focus on the mass - to - light ratio mismatch relative to the milky way imf , , correlated against the central stellar velocity dispersion , . we find a strong correlation between and , for a wide set of dark matter ( dm ) model profiles . these results are robust if a uniform halo response to baryons is adopted across the sample . the overall normalization of , and the detailed dm profile , are less certain , but the data are consistent with standard cold - dm halos , and a central dm fraction that is roughly constant with . for a variety of related studies in the literature , using sps , dynamics , and gravitational lensing , similar results are found . studies based solely on spectroscopic line diagnostics agree on a salpeter - like imf at high , but differ at low . overall , we find that multiple independent lines of evidence appear to be converging on a systematic variation in the imf , such that high- etgs have an excess of low - mass stars relative to spirals and low- etgs . robust verification of super - salpeter imfs in the highest- galaxies will require additional scrutiny of scatter and systematic uncertainties . the implications for the distribution of dm are still inconclusive .
low - dimensional superconductors with layered structure have been extensively studied and still attract much interest due to their exotic superconducting properties and mechanism when compared to conventional bcs superconductors . the examples include high @xmath1 cuprates,@xcite sr@xmath0ruo@xmath2,@xcite na@xmath3coo@xmath4h@xmath0o,@xcite and iron - based superconductors.@xcite the discovery of lnofepn ( ln = rare earth elements , pn = p , as ) in particular revitalises the study of layered compounds with mixed anions , paving a way to materials with novel physical properties . for example , ln@xmath5o@xmath0tm@xmath0och@xmath0 ( tm = transition metals , ch = s , se ) show strong electron - electron interactions and mott insulating state on the two dimensional ( 2d ) frustrated antiferromagnetic ( afm ) checkerboard spin - lattice.@xcite@xmath6@xcite very recently , bulk superconductivity was found in bis@xmath0-type layered compounds with mixed anions : bi@xmath2o@xmath2s@xmath7 and ln(o , f)bis@xmath0.@xcite@xmath8@xcite experimental and theoretical studies indicate that these materials exhibit multiband behaviors with dominant electron carriers originating from the bi 6@xmath9 and 6@xmath10 bands in the normal state.usui@xmath6@xcite on the other hand , compounds with mixed anions exhibit remarkable flexibility of structure . different two - dimensional ( 2d ) building blocks , such as [ lno]@xmath11 , [ aef]@xmath11 ( ae = ca , sr , ba ) , [ ti@xmath12opn@xmath0]@xmath14 , [ fepn]@xmath6 , and [ tm@xmath0och@xmath0]@xmath14 , can sometimes be integrated to form new materials.@xcite@xmath15matsuishi@xmath6@xcite individual building blocks often keep their structural and electronic properties after being combined together.kabbour in this work , we report the discovery of a new bis@xmath0-based layered compound srfbis@xmath0 . it contains nacl - type bis@xmath0 layer and shows semiconducting behavior with relatively large thermopower . theoretical calculation indicates that this compound is very similar to lnobis@xmath0 . srfbis@xmath0 polycrystals were synthesized by a two - step solids state reaction . first , bi@xmath0s@xmath7 was prereacted by reacting bi needles ( purity 99.99% , alfa aesar ) with sulfur flakes ( purity 99.99% , aldrich ) in an evacuated quartz tube at 600 @xmath16 for 10 h. then bi@xmath0s@xmath17 was mixed with stoichiometric srf@xmath0 ( purity 99% , alfa aesar ) and srs ( purity 99.9% , alfa aesar ) and intimately ground together using an agate pestle and mortar . the ground powder was pressed into 10 mm diameter pellets . we used a maximum pressure of 5 tons . the pressed pellet was loaded in an alumina crucible and then sealed in quartz tubes with ar under the pressure of 0.15 atmosphere . the quartz tubes were heated up to 600 @xmath18 in 10 h and kept at 600 @xmath19 for another 10 h. phase identity and purity were confirmed by powder x - ray diffraction carried out by a rigaku miniflex x - ray machine with cu k@xmath20 radiation ( @xmath21 = 1.5418 ) . structural refinement of powder srfbis@xmath0sample was carried out by using rietica software.@xcite synchrotron x - ray experiment was conducted at 300 k on x17a beamline of the national synchrotron light source ( nsls ) at brookhaven national laboratory ( bnl ) . the setup utilized x - ray beam 0.5 mm @xmath22 0.5 mm in size and @xmath23 = 0.1839 ( @xmath24 = 67.4959 kev ) , conditioned by two - axis focusing with one - bounce sagittally - bent laue crystal monochromator , and perkin - elmer image plate detector mounted perpendicular to the primary beam path . finely pulverized sample packed in cylindrical polyimide capillary 1 mm in diameter was placed 204 mm away from the detector . multiple scans were performed to a total exposure time of 120 @xmath25 . the 2d diffraction data were integrated and converted to intensity versus 2@xmath26 using the software fit2d.hammersley the intensity data were corrected and normalized and converted to atomic pair distribution function ( pdf ) , @xmath27 , using the program pdfgetx2.@xcite the average stoichiometry of srfbis@xmath0 polycrystal was determined by examination of multiple points using an energy - dispersive x - ray spectroscopy ( edx ) in a jeol jsm-6500 scanning electron microscope . the sample pellets were cut into rectangular bar and the surface is polished by sandpaper . thin pt wires were attached using silver epoxy for four probe resistivity measurements . electrical and thermal transport measurements were carried out in quantum design physical property measurement system ( ppms-9 ) . first principle electronic structure calculation were performed using experimental crystallographic parameters within the full - potential linearized augmented plane wave ( lapw ) method @xcite implemented in wien2k package.@xcite the general gradient approximation ( gga ) of perdew _ et al_.,@xcite was used for exchange - correlation potential . the lapw sphere radius were set to 2.5 bohr for all atoms , and the converged basis corresponding to @xmath28 with additional local orbital were used where @xmath29 is the minimum lapw sphere radius and @xmath30 is the plane wave cutoff . 1(a ) shows the powder xrd pattern of srfbis@xmath0 measured by rigaku miniflex . almost all of reflections can be indexed using the p4/nmm space group . the unidentified peaks belong to the second phase of bi@xmath0s@xmath7 . using two - phase le bail fitting , the refined lattice parameters of srfbis@xmath0 are @xmath31 = 4.084(2 ) and @xmath32 = 13.798(2 ) . when compared to laobis@xmath0 , the a - axial lattice parameter is larger and the c - axial one is slightly smaller.@xcite the pdf structural analysis was carried out using the program pdfgui.@xcite the srfbis@xmath33 data are explained well within the model having p4/nmm symmetry with @xmath34 = 4.079(2 ) and @xmath32 = 13.814(5 ) (@xmath35 = 0.138 , @xmath36=0.024 ) . it is consistent with the fitting results obtained from miniflex . the final fit is shown in fig . 1(b ) , and the results are summarized in table 1 . in addition to the principal phase , the sample is found to have @xmath37 16(1 ) wt% of bi@xmath0s@xmath7 impurity with pnma symmetry , which is also observed in fig . structure of srfbis@xmath0 is similar to laobis@xmath0 , which is built up by stacking the rock - salt - type bis@xmath0 layer and fluorite - type srf layer alternatively along the c axis as shown fig . the edx spectrum of polycrystal confirms the presence of sr , f , bi , and s. the average atomic ratios determined from edx are sr : f : bi : s = 1.00(4 ) : 1.00(9 ) : 1.03(5 ) : 1.88(4 ) when setting the content of sr as 1 . it confirms the formula of obtained compound is srfbis@xmath0 . and the fitted result using two - phase le bail fitting . crosses are experimental data , red line is fitted spectra , green line is the difference between experimental data and fitted spectra , vertical lines are calculated bragg - positions for srfbis@xmath0 ( upper ) and bi@xmath0s@xmath7 ( lower ) , respectively . ( b ) synchrotron pdf refinement of data taken at room temperature . ( c ) crystal structure of srfbis@xmath0 . the biggest white , big purple , medium orange , and small green balls represent sr , bi , s , and f ions , respectively . ] cccccc & & + & & + & & + & & + & & + & & + & & + & & + & & + atom & site & x & y & z & @xmath38 ( @xmath39)@xmath40 + sr & 2c & 1/4 & 1/4 & 0.1025(2 ) & 0.0069(4 ) + f & 2a & 3/4 & 1/4 & 0 & 0.033(2 ) + bi & 2c & 1/4 & 1/4 & 0.6286(5 ) & 0.0183(3 ) + s1 & 2c & 1/4 & 1/4 & 0.379(3 ) & 0.060(2 ) + s2 & 2c & 1/4 & 1/4 & 0.811(2 ) & 0.019(1 ) + [ 1 ] of the srfbis@xmath0 at @xmath41 = 0 ( closed blue square)and 90 koe(open red circle ) . inset shows the fitted result using thermal activation model for @xmath42 at zero field where the red line is the fitting curve . ] as shown in fig . 2 , the resistivity @xmath43 of srfbis@xmath0 polycrystalline shows a semiconducting behavior in the measured temperature region ( 1.9 - 300 k ) . it should be noted that bi@xmath0s@xmath7 polycrystal shows metallic behavior becasue of sulfur deficiency.@xcite the impurity may have some minor influence on the absolute value of resistivity , but the semiconducting behavior should be intrinsic . neglecting the grain boundary contribution , the room - temperature resistivity @xmath44 k@xmath45 is about 0.5 @xmath46 cm . using the thermal activation model @xmath47 ( @xmath48 is a prefactor , @xmath49 thermal activated energy and @xmath50 the boltzmann s constant ) to fit the @xmath43 at high temperature ( 75 k - 300 k ) ( inset of fig . 2 ) , we obtain @xmath49 = 31.8(3 ) mev . the semiconductor behavior is consistent with theoretical calculation result shown below . on the other hand , theoretical calculations have indicated that undoped laobis@xmath0 is also a semiconductor , which is partially consistent with the experimental result.@xcite@xmath15@xcite transport measurement indicates that laobis@xmath0 shows semiconducting behavior at @xmath51 200 k , but exhibit an upturn of resistivity at higher temperature . the origin of the upturn is unclear . therefore , the replacement of lao by srf should not change the band structure and thus physical properties too much , especially at low temperature , similar to the relation between srffeas and laofeas.@xcite@xmath15@xcite the slight differences between laobis@xmath0 and srfbis@xmath0 , such as larger a - axial and smaller c - axial lattice parameters , could result in changing of physical properties at higher temperature . note that the semiconducting @xmath43 in laobis@xmath0 and srfbis@xmath0 are different from those in parent compounds of iron pnictide superconductors . the latter show metallic behaviors at high temperature and semiconducting - like upturn in resistivity curve related to the spin density wave ( sdw ) transition . there is no significant magnetoresistance in srfbis@xmath0 up to 90 koe magnetic field . under zero magnetic field within a temperature range from 2 to 340 k. ] the temperature dependences of the thermal conductivity @xmath52 and thermoelectric power ( tep ) @xmath53 for srfbis@xmath0 in zero field between 2 and 350 k are shown in fig . the electronic thermal conductivity @xmath54 estimated from the wiedemann - franz law using a value for the lorenz number of 2.44@xmath2210@xmath55 w @xmath56/k@xmath39 was less than 5@xmath2210@xmath57 of @xmath52 . therefore , lattice thermal conductivity dominates @xmath58 which exhibits a peak at around 60 k ( fig . the peak in @xmath52 commonly arises since different phonon scattering processes usually dominate in different temperature ranges . umklapp scattering dominates at high temperatures , while boundary and point - defect scattering dominate at low and intermediate temperatures , respectively.@xcite on the other hand , the @xmath52 of srfbis@xmath0 shows similar behavior to bi@xmath2o@xmath2s@xmath7 but with different peak position and absolute value.@xcite for tep @xmath53 of srfbis@xmath0 , there is a reversal in sign at about 11 k , i.e , hole - like carrier changes into electron - like carrier which is dominant at room temperature . according to two band model , @xmath59.@xcite if we assume that @xmath60 and @xmath61 are temperature independent , it suggests that electron and hole conductivities change dramatically with temperature : at low temperature , @xmath62 whereas @xmath63 above 11 k. hole - like carrier may originate from defect induced p - type doping . with increasing temperature , electron - like carrier due to intrinsic band excitation increase significantly , finally leading to @xmath64 and a sign change in @xmath53 . similar behavior was observed in laoznp and p - type si.@xcite even though the @xmath53 in srfbis@xmath0 is significant and not much smaller than in classics thermoelectric materials , @xcite its low electrical conductivity makes its figure of merit zt ( zt = @xmath65s@xmath39t/@xmath66 ) extremely small . . ] first principle calculations ( fig . 4 ) confirm that srfbis@xmath0 is a semiconductor with a direct band gap of 0.8 ev located at @xmath67 point . this is similar to laobis@xmath0 where the energy gap was found to be 0.82 ev.wan the calculation confirms the results of transport measurement . similar to laobis@xmath0,@xcite@xmath15@xcite both s @xmath68 and bi @xmath69 states are located around the fermi level ( -2.0 to 2.0 ev ) in srfbis@xmath0 . thus there is a strong hybridization between s @xmath68 and bi @xmath69 states . the absence of dispersion along @xmath70 line suggests quasi two dimensional character of the band structure in srfbis@xmath0 ( fig . 4(b ) ) . in laobis@xmath0 , f doping results in metallic states and superconductivity at low temperature . main influence of f substitution is a carrier doping that shifts the fermi level and has only minor effect on the lowest conduction band . due to similarity between srfbis@xmath0 and laobis@xmath0 , new superconductors could be obtained by chemical substitution . in summary , we report a discovery of a new layered fluorosulfide srfbis@xmath0 . it contains nacl - type bis@xmath0 layer similar to bi@xmath2o@xmath71s@xmath7 and ln(o , f)bis@xmath0 superconductors . srfbis@xmath0 polycrystals shows semiconducting behavior between 2 k and 300 k. we observe rather small thermal conductivity and large tep with sign reversal at low temperature . theoretical calculation confirms the semiconducting behavior and indicates similar dos and band structure to undoped laobis@xmath0 . because of the similarity between srfbis@xmath0 and the parent compound of bis@xmath0-based superconductors , it is of interest to investigate the doping effects on physical properties of srfbis@xmath0 . it could pave a way to new members in this emerging family of bis@xmath0-based superconductors . we thank john warren for help with sem measurements . work at brookhaven is supported by the u.s . doe under contract no . de - ac02 - 98ch10886 and in part by the center for emergent superconductivity , an energy frontier research center funded by the u.s . doe , office for basic energy science ( h. l. and c. p. ) . this work benefited from usage of x17a beamline of the national synchrotron light source at brookhaven national laboratory . we gratefully acknowledge zhong zhong and jonathan hanson for their help with the x17a experiment setup . p. blaha , k. schwarz , g. k. h. madsen , d. kvasnicka and j. luitz , wien2k , an augmented plane wave + local orbitals program for calculating crystal properties ( karlheinz schwarz , techn . universitat wien , austria ) , 2001 . isbn 3 - 9501031 - 1 - 2	we have synthesized a new layered bis@xmath0-based compound srfbis@xmath0 . this compound has similar structure to bis@xmath0 . it is built up by stacking up srf layers and nacl - type bis@xmath0 layers alternatively along the c axis . electric transport measurement indicates that srfbis@xmath0 is a semiconductor . thermal transport measurement shows that srfbis@xmath0 has a small thermal conductivity and large seebeck coefficient . first principle calculations are in agreement with experimental results and show that srfbis@xmath0 is very similar to laobis@xmath0 which becomes superconductor with f doping . therefore , srfbis@xmath0 may be a parent compound of new superconductors .
throughout the milky way s history , a non - negligible fraction of star formation has apparently occurred in starburst - like events ( see majewski , 1993 or freeman & bland - hawthorn , 2002 , for a review ) . on the other hand , observations suggest that present - day star formation in the disk of our galaxy takes place in stellar groupings rather than in isolation . these stellar aggregates form from molecular clouds , and in the disk of the milky way they appear in two types : bound and unbound . unbound , short lived stellar groupings are called associations ; bound , long lived stellar groups are known as open clusters . open clusters can also be formed out of the remains of rich stellar associations ( kroupa et al . , 2001 ) . as most of the field stars appear to have been formed in the so - called clustered mode ( i.e. , in clusters or associations ) , not in the dispersed mode ( i.e. , in isolation ) , it is a natural choice to consider these stellar clumpings as the _ de facto _ units of star formation in the disk of our galaxy ( clarke et al . , 2000 ) . the idea of open clusters being fundamental units of star formation in the galactic disk is , however , controversial ( meyer et al . , 2000 ) . it has been argued that bound open clusters can not contribute significantly to the field star population of the galactic disk because they are rare and long lived ( e.g. roberts , 1957 ) . in contrast , most young embedded clusters are thought to evolve into unbound stellar associations , which comprise the majority of stars that populate the milky way disk ( lada & lada , 1991 ) . nevertheless , if star clusters are the elementary units of star formation they can , in principle , be used to derive the star formation history , recent and old . unfortunately , stellar associations evolve and dissolve in a time - scale of @xmath3 50 myr ( brown , 2002 ) , therefore they can not be used to study the star formation history of the galactic disk . on the other hand , open clusters are comparatively long lived objects that may serve as excellent probes into the structure and evolution of the galactic disk . the problem of deriving the star formation history of the milky way has been considered by a number of authors using different techniques ( twarog , 1980 ; scalo , 1987 ; barry , 1988 ; gomez et al . , 1990 ; marsakov et al . , 1990 ; noh & scalo , 1990 ; meusinger , 1991 ; soderblom et al . , 1991 ; micela et al . , 1993 ; daz - pinto et al . , 1994 ; rocha - pinto & maciel , 1997 ; chereul et al . , 1998 ; isern et al . , 1999 ; lachaume et al . , 1999 ; rocha - pinto et al . , 2000a ; hernandez et al . , 2000a ; bertelli & nasi , 2001 ; gizis et al . , 2002 ; just , 2002 , 2003 ) . the majority of these studies suggest that the disk of the milky way galaxy has not experienced a smooth and constant star formation history but a bursty one with several episodes of enhanced star formation . although most of these papers are restricted to the study of the star formation history in the solar neighbourhood , the star sample birth sites are in fact distributed over a larger range of distances because of orbital diffusion , and so they can provide an estimate of the global star formation rate and their conclusions can be extrapolated to the entire milky way disk . in this paper we revisit the topic of the recent star formation history of the milky way disk by using data from the _ open cluster database _ ( mermilliod , 2003 ) , the _ new catalogue of optically visible open clusters and candidates _ ( dias et al . , 2003a , b ) as well as _ hipparcos _ data ( esa , 1997 ) to construct the open cluster age distribution . the age distribution diagram is then interpreted using numerical results from an extensive library of @xmath0-body calculations carried out during the last ten years . this method permits the reconstruction of the star formation history of the milky way disk over the last 2 gyr with a time resolution of 0.05 gyr using very few a priori assumptions . our main objective is to understand the recent star formation history of the entire galactic disk although only the solar circle ( @xmath1 1.75 kpc ) can be studied due to the incompleteness of the available open cluster sample ( see section 2 ) . this paper is organized as follows : in section 2 , we present the properties of the galactic open cluster sample . the raw histogram of open cluster number vs. age is also included in this section as well as a discussion of the potential limitations of our approach . in section 3 , we summarize some relevant results on the evolution and dissolution of realistic @xmath0-body star cluster models . this section also includes a comparison between published open cluster ages and number of stars with the results of the numerical models . in section 4 , we present our method as well as results for the galactic disk . a closer view to the last 0.2 gyr and the derived open cluster initial mass function are considered in section 5 . the star formation history is presented and discussed in section 6 ; corrections are also discussed here . a detailed comparison with results from other authors is carried out in section 7 . an alternative interpretation of our results in the context of dynamical merger histories is included in section 8 . open questions and conclusions are summarized in section 9 . for many years , the lyng catalogue ( 1985 , 1987 ) has been the classical reference source for open cluster data . an exhaustive analysis of the properties of the milky way open cluster system using this catalogue was carried out by janes et al . unfortunately , the lyng catalogue is not the most updated source of data in this field . open cluster database _ ( webda ) created and maintained by j .- c . mermilliod ( 1988 , 1992a , 1992b , 1993 , 1995 , 1996 , 2003 ) includes all the data already covered in lyng s catalogue and many more . the latest update of the open cluster database ( october 2003 ) includes 1731 open clusters with ages for 616 objects ( 36% ) . in this database we have found 581 clusters with age @xmath4 2 gyr . slightly less complete is the _ new catalogue of optically visible open clusters and candidates _ ( ncovocc ) . the first version of this catalogue was published in dias et al . the october 2003 version ( dias et al . , 2003b ) , includes 1637 objects , 603 ( 37% ) with published ages . this new catalogue updates the previous catalogues of lyng ( 1987 ) and of mermilliod ( 1995 ) included in the webda database . there are however non - negligible differences between the two databases as we can see from figs . [ mermilliod ] and [ dias ] . ncovocc includes 569 clusters of age @xmath4 2 gyr . on the other hand , salaris et al . ( 2004 ) have recently re - determined ages for a sample of 71 of the oldest open clusters ( all of them older than 0.67 gyr ) using a morphological age indicator , metallicities , and _ hipparcos _ parallaxes . in general , the new ages are older than those from webda and ncovocc . in order to study the impact of different age determination techniques on our results we have modified the ncovocc data with the new age determinations from salaris et al . the new subsample consists of 559 open clusters younger than 2 gyr . the age - distance diagram for this new sample appears in fig . [ salaris ] . in this paper we use the three subsamples presented above to construct an age distribution diagram . as explained below , we choose 2 gyr in order to obtain statistically relevant results . our samples contain objects from dense , very young clusters to old and almost dissolved moving groups . the age distribution diagram for the three open cluster samples considered is shown in fig . although a detailed analysis of the different features present in this histogram is postponed to section 4 , the figure suggests that several epochs of enhanced star formation may have taken place during the last 2 gyr . the age - distance diagrams for the three samples appear in figs . [ mermilliod ] , [ dias ] , and [ salaris ] . these figures indicate that the open cluster data used in this paper provide an incomplete sample ; unfortunately , it is difficult to estimate the degree of incompleteness of this sample due to the irregular distribution of galactic open clusters . the sample appears to be rather incomplete for open clusters more distant than @xmath3 3.5 kpc , therefore our conclusions can only be rigorously applied to the solar circle : an annulus of galactocentric distance @xmath5 kpc , where @xmath6 is the solar galactocentric distance ( @xmath3 8.5 kpc , e.g. , zombeck , 1990 ) . on the other hand , the hipparcos catalogue contains numerous selection biases which are difficult to characterize . it is , however , reasonable to assume that the sample is fairly complete ( 166 clusters ) for objects in the solar neighbourhood ( distances to the sun of 1 kpc or less ) although selection effects are likely to be very important for clusters located in the outskirts of the galactic disk as well as in the region behind the galactic center and for small clusters with galactocentric distance @xmath7 kpc as they are projected against the dense stellar fields of the southern milky way . on the other hand , intermediate age and old open cluster samples are incomplete as clusters with higher @xmath8 are easier to detect than clusters with lower @xmath8 . in this paper , we are interested in studying only the recent star formation history of the milky way ( @xmath4 2 gyr ) , therefore any effects arising from lack of completeness of the sample are likely to be less important than those from age errors . besides , open cluster samples are biased toward objects where the total luminosity is dominated by ultra - luminous o and b stars ( young clusters ) or red ( or blue ) giants ( old clusters ) as these systems are easier to identify . however , if o and b stars were the only stars the cluster would not be stable since the massive stars are destined to become supernovae , returning their mass to the interstellar medium . hence , the existence of low mass stars is critical for the survival of open clusters . old , dynamically depleted open clusters are very difficult to identify although they may exist in large numbers ( de la fuente marcos , 1998 ) . faint , population depleted open clusters are very difficult to identify against the field stars . therefore we are working with an incomplete , volume - limited sample that is also flux - limited . on the other hand , and as for any other physical parameter , open cluster age determinations are affected by errors . in a large sample , these errors are very likely to be non - homogeneous as different methods have been used by different authors to calculate the ages . the disparity in open cluster ages ( when considering different authors , see webda for multiple examples ) partly reflects the critical dependence of cluster isochrone fitting on the adopted reddening ( often large for open clusters ) , even small reddening errors can create significant errors in the derived age and metallicity . however , determining effective temperatures and metallicities for cluster turn off stars directly through echelle spectroscopy is free from systematic errors and subject only to uncertainties in the model atmospheres . some ages for clusters in the samples considered have been determined using the first technique , but some others have been found using the second one or even other indirect methods . it is relatively difficult , specially in papers older than about 5 years , to find published estimations of the errors associated with open cluster age determinations . in order to evaluate the impact of age errors in our analysis we have compiled errors for 52 objects within the studied age range from published sources : ngc 6134 and ngc 3680 ( bruntt et al . , 1999 ) ; ngc 2141 ( carraro et al . , 2001 ) ; ngc 2158 ( grocholski & sarajedini , 2002 ) ; ngc 2112 ( carraro et al . , 2002 ) ; berkeley 104 , berkeley 60 , king 15 , ngc 381 , berkeley 64 , king 6 , ngc 1348 , berkeley 23 , ngc 2259 , ngc 2304 ( ann et al . , 2002 ) ; ngc 1663 ( baume et al . , 2003 ) ; ngc 3990 ( prisinzano et al . , 2004 ) ; pismis 19 ( carraro & munari , 2004 ) ; ic 166 , ngc 752 , king 5 , ngc 1254 , ngc 1278 , ngc 1817 , ngc 2158 , ngc 2194 , ngc 2192 , ngc 2236 , ngc 2266 , berkeley 30 , ngc 2324 , ngc 2354 , ngc 2355 , ngc 2360 , haffner 6 , melotte 71 , pismis 2 , ngc 2660 , ngc 2849 , ngc 3680 , ngc 4815 , ngc 5822 , ic 4651 , ic 4756 , berkeley 42 , ngc 6802 , ngc 6827 , ngc 7044 , ngc 2477 , ngc 7789 , ngc 2204 , hyades , and praesepe ( salaris et al . , 2004 ) . [ ocerror ] shows the uncertainty as a function of the age for this open cluster sample . the average age error is about 22% and the plot suggests that younger ages are affected by larger errors ( for a recent discussion on the problems associated with age determinations in young open clusters see , e.g. , piskunov et al . , 2004 ) . if this error sample can be considered as representative of the error range for the entire sample , the youngest cluster ( age @xmath9 0.5 gyr ) errors are very likely in the range 50 - 150 myr with for older cluster errors in the range 150 - 250 myr . in order to test the reliability of the features found in the open cluster age distribution we have generated synthetic age distribution diagrams from the original data . each simulation was composed of the following steps : \1 ) the original sample from mermilliod ( 2003 ) is the input \2 ) a randomly distributed error within a given percentage of the input value is added / subtracted to each value in the original sample . \3 ) the simulated open cluster sample is binned at 0.050 gyr intervals . in this way the real age of the open clusters in the sample is shifted randomly according to a given average error ( 15 , 25 , and 35 % ) . \4 ) the process is repeated 175,000 times per error value and we calculate the average number of objects per age bin . the final , averaged , sample is plotted in fig . [ error ] . we have carried out three sets with 175,000 simulations to study this effect . these simulations appear to suggest that for age errors of about 15% or lower , most of the features in the age distribution remain relatively unaltered . for errors in age determination larger than about 25% only the youngest features are recovered . fig . [ ay ] is nothing more than the actual age distribution , a raw histogram of open cluster numbers vs. age , of relatively young , nearby open clusters . in order to interpret this distribution additional tools are needed . experimental stellar dynamics may help us in this analysis by matching cluster dynamics and cluster data to provide a better view of the star formation history of the milky way disk . this section summarizes some relevant results on the evolution and dissolution of realistic @xmath0-body star cluster models . some technical details on the modeling process are also discussed as well as the initial mass functions used in the calculations . after an open cluster is formed and its massive stars have expelled the gas it will evolve steadily , losing stars on a relatively long time - scale . distant encounters among cluster stars together with the galactic tidal field cause the slow and gradual evaporation of the cluster . close encounters and stellar evolution can speed up the process , specially for small , sparsely populated clusters . primordial binaries are also very important ( although less than stellar evolution ) as they dominate completely the dynamical evolution of poor clusters . violent , very fast cluster disruption can be induced during encounters with giant molecular clouds ( terlevich , 1983 , 1985 , 1987 ) . the disintegration time - scale of open clusters depends on both population and galactocentric distance ; but , for a given galactocentric distance , the initial population ( @xmath10 ) as well as the initial mass function are the main physical parameters to consider in order to analyze the long term evolution of the cluster . the numerical results discussed here come from an extensive set of simulations ( de la fuente marcos , 1995 , 1996a , 1996b , 1997a , 1997b ; de la fuente marcos & de la fuente marcos , 2000 , 2002 ) calculated with the standard @xmath0-body code nbody5 ( aarseth , 1985 , 1994 , 1999 , 2003 ) for clusters located in the solar neighbourhood . these calculations include the effects of stellar evolution , the galactic tidal field , primordial binaries , and realistic initial mass functions ( hereafter imf ) . nbody5 consists of a fourth - order predictor - corrector integration scheme with individual time steps . it utilizes an ahmad - cohen ( 1973 ) neighbour scheme to facilitate calculation of the gravitational forces , and handles close encounters via two- , three- , four- , and chain regularization techniques ( kustaanheimo & stiefel , 1965 ; aarseth & zare , 1974 ; mikkola , 1985 ; mikkola & aarseth , 1993 ) . five different imfs were used to generate stellar masses . two standard power law models ( salpeter , 1995 ; taff , 1974 ) and three modern imfs due to the miller & scalo ( miller & scalo , 1979 ; eggleton et al . , 1989 ) , kroupa ( kroupa & tout , 1992 ; kroupa et al . , 1990 , 1991 , 1993 ) and scalo ( scalo , 1986 ) . the imfs used in our calculations differ mainly in the number of low - mass and high - mass stars generated . the salpeter imf is given reasonably well by the approximation @xmath11 where @xmath12 is the number of stars in the mass interval @xmath13 to @xmath14 . the taff imf is given by @xmath15 with @xmath16 for clusters with @xmath17 and @xmath18 for clusters with @xmath19 . for generating masses with the miller & scalo imf we use the function ( eggleton et al . , 1989 ) @xmath20 where x is uniformly distributed in the interval @xmath21 $ ] . for x @xmath3 0.70.999 , this gives a power - law expression with @xmath22 and for x @xmath23 1 it gives an approximately constant value . for x @xmath24 0.999 , the imf slope is increased to 5 . the kroupa imf is given by ( kroupa et al . , 1993 ) @xmath25 we use the mass - generating function @xmath26 as a convenient representation of these relations , where x is uniformly distributed in the interval @xmath21 $ ] and the @xmath27-parameters are given by @xmath28 , @xmath29 , @xmath30 and @xmath31 . no masses less than 0.08 @xmath32 are generated by this formula . it is based on scalo s ( 1986 ) initial mass function . for the scalo imf we use the mass - generating function @xmath33 where x is uniformly distributed in the interval @xmath21 $ ] . it is similar to kroupa imf except for @xmath34 0.16 ; scalo imf gives a smaller number of stars with those masses . the kroupa imf is the most recent imf we have used , it generates a greater number of low - mass stars and a smaller number of heavy stars . on the contrary , the salpeter imf gives a smaller number of low - mass stars and a greater number of massive stars . among these , the taff imf gives a smaller number of massive stars than the salpeter one but greater than the miller & scalo one . the scalo imf is intermediate between miller & scalo and kroupa imfs . for low - mass stars , the miller & scalo imf gives a greater number than the salpeter and taff ones but smaller than scalo and kroupa imfs . in addition , three different star densities ( uniform , @xmath35 1/@xmath36 , plummer ) have been used to generate the initial positions of the stars in the cluster model . since the initial membership of a star cluster is the main parameter with regard to cluster life - time for a given galactocentric distance , stellar evolution ( through the imf ) is the only process able to induce a differential behaviour as a function of @xmath0 . averaging cluster life - times for the entire library of models as a function of @xmath0 to fit a power law to the data we obtain @xmath37 ( in gyr ) with a correlation coefficient of @xmath38 ( fig , [ our ] ) . this slope is very similar to the slopes found by baumgardt ( 2001 ) and baumgardt & makino ( 2003 ) although the details of our simulations are rather different . the above evolutionary track is only strictly valid for cluster models with @xmath39 100 - 150 members . unfortunately , for smaller clusters the characteristic life - time is not very well defined as the energy released in a single supernova event may be larger that the binding energy of the entire cluster ( de la fuente marcos , 1993 ) . for clusters with @xmath40 , life - times are in the range 5 - 40 myr and for @xmath0 = 75 , @xmath41 80 . in general , the fluctuation range for the life - time of small-@xmath0 cluster models is very large and depends strongly on the initial mass function , the spatial distribution , and the binary fraction . for richer clusters , simulations indicate that , in the solar neighbourhood and neglecting interactions with giant molecular clouds and the effect of the cluster gas on the early stages of the evolution of the cluster , an open cluster has to include about 200 - 400 stars ( at least ) in order to survive for about 0.5 gyr , 400 - 700 to last 0.7 gyr , 700 - 1000 to be detectable after 0.9 gyr and 1000 - 2000 to survive for about 1.3 gyr . if a cluster is still observable after 2 gyr , its initial population was at least 3000 stars . in terms of cluster masses and for an average stellar mass of 0.4 @xmath32 , the oldest open clusters considered in this paper ( 2 gyr ) were born with masses in excess of 1.2 @xmath42 and very likely around @xmath43 . the astrophysical interpretation of the solid line in fig . [ our ] is rather clear : it is the open cluster _ death line_. open clusters well below the line are young and they evolve towards the line as they lose stars into the field . as it evolves , an average open cluster spends most of its life - time close to the _ death line _ moving parallel to it , in the population vs. age diagram . open clusters become part of the stellar field when they cross the line . it is , however , very difficult to find open cluster memberships and masses among the published literature . star counts are in many cases unreliable and proper motions are not always available to confirm membership . there is however an observational parameter that may help in open cluster membership determinations : the integrated magnitude of the cluster . the integrated absolute visual magnitude of a star cluster can be written as : @xmath44 where the absolute visual magnitude of the sun is 4.82 ( e.g. zombeck , 1990 ) and the luminosity of the cluster could be , in principle , estimated using the mass - luminosity relationship @xmath45 , with @xmath46 ( popper , 1980 ) for stars in the main sequence . unfortunately , the main contribution to the intrinsic brightness of a relatively young cluster is not coming from the stars on the main sequence but from red ( or blue ) subgiants or giants . taking this fact into account the luminosity of the cluster can be written as : @xmath47 where @xmath48 and @xmath49 are the number of stars in the main sequence and outside the main sequence , respectively . their relative contribution to the integrated luminosity of the cluster is obtained multiplying by the corresponding average luminosity . this average luminosity can be obtained from the average mass and the mass - luminosity relationship for the contribution from the main sequence . for subgiants or giants it is also possible to obtain an average luminosity using data for standard stars ( e.g. zombeck , 1990 ) . in our calculations we will consider an average luminosity for stars outside the main sequence of 565 @xmath50 . combining eqs . [ ivm ] , [ oclu ] , @xmath51 , and considering the average stellar mass @xmath13 we obtain : @xmath52 the fraction of stars outside the main sequence can be estimated assuming an imf and taking into account that an estimation of the life - time in the main sequence for a given star of mass @xmath13 and luminosity @xmath53 is @xmath54 . the relationship given by eq . [ npop ] can be calibrated using a few clusters with well established memberships ( e.g. hyades , pleiades ) . for the age range studied in this work the fraction of stars outside the main sequence is in the range 0.5 - 10% . the final value of @xmath0 is affected by our choice for this fraction but the main contribution to the uncertainty in @xmath0 is coming from the observational determination of the integrated magnitude . for galactic open clusters this uncertainty could be @xmath55 mag ( battinelli et al . , 1994 ; lata et al . , 2002 ) . this translates into 20% errors in @xmath0 . we have applied this approach to estimate the current population of a small sample of open clusters ( hyades , pleiades , ic 4756 , ngc 2204 , ngc 2506 , ngc 7789 , ngc 1245 , ngc 381 , ngc 2324 , and ngc 2192 ) with age , error in age , and integrated absolute visual magnitude available from published sources ( webda ; battinelli et al . , 1994 ; gray , 1965 ; pandey et al . , 1989 ; lata et al . , 2002 ; sagar et al . , 1983 ; spassova & beav , 1985 ) . the results are also plotted in fig . [ our ] , the dashed line represents the power law fit to the open cluster data ( grey circles ) , @xmath56 with a correlation coefficient of @xmath57 . only one open cluster younger than 500 myr has been included because all of them are well below the limiting line . the power law fit includes all the plotted points . if the youngest cluster is not considered then the correlation coefficient is essentially 1 . in any case , [ our ] appears to confirm that observational data and results from @xmath0-body simulations are fully consistent within the error limits . in this section we provide a detailed description of our open cluster age distribution method . as stated in section 2 , our results are mainly sensitive to the errors in age determination but also to the degree of completeness of the open cluster sample . these two effects have been analyzed in section 2 and are discussed in detail below . on the other hand , any errors in theoretical stellar models also propagate into the results obtained because the published open cluster ages always make reference to theoretical stellar models , sometimes through direct isochrone comparison , others through the use of morphological features found in the cluster color - magnitude diagram . the open cluster catalogues offer high - quality data for a relatively large number of objects in the milky way disk , both in the solar neighbourhood and beyond . in particular , the age data can be used to construct an age distribution for the milky way open clusters , fig . [ ay ] . assuming that the sample considered is representative of the entire galactic disk , the star formation rate can be derived from its age distribution , as the number of open clusters in each age bin is , by hypothesis , correlated with the number of open clusters initially born at that time . once an age distribution is available , it is in principle easy to recover the star formation history that gave rise to the observed age distribution by using results from realistic @xmath0-body simulations . this method permits the reconstruction of the global star formation history of the galactic disk with a time resolution of 0.050 gyr over the last 2 gyr . our method does not assume any a priori structure or condition on the star formation rate and it basically consists of two steps : ( i ) construct a representative sample of open clusters . in principle , the optimal approach will be to construct a volume - limited open cluster sample in the solar vicinity ( see section 6 ) . the cluster sample spans a very large range of ages . ( ii ) construct the age distribution diagram for the sample . ( iii ) infer the star formation history from the diagram . observations of bursts of star formation in other galaxies indicate that intense star formation is always associated with production of large star clusters . the life span of larger clusters is longer ; therefore , an usually high number of open clusters at a given time interval can be interpreted as the result of an event of enhanced star formation at that given age . this is probably the main a priori assumption maintained throughout the paper although it appears to be strongly supported by observational evidence . if star cluster masses are sampled from an open cluster initial mass function , larger numbers translate into increased probability of formation of large star clusters . luminous , and therefore rich , young star clusters are found whenever there is vigorous star formation , whether it be in galaxy mergers or starburst galaxies . galaxies with very active star formation have proportionally more of their stars in clusters than in the field , with some of them devoting as much as 15 - 20% of their luminosity to clusters ( larsen & richtler , 2000 ) . presently available observational data strongly suggests that the cluster forming frequency is highest during violent bursts of star formation ( van den bergh , 2000 ) . the resulting star formation history comes directly from the age distribution ( figs . [ ay ] , [ ayz ] ) , in an approach which assumes that the most frequent ages of the open clusters indicate the epochs when the star formation was more intense if we take into account that star clusters are the elementary units of the star formation process . we have not included any volume correction because the number of clusters more distant than about 6 kpc is negligible for the two samples considered . the volume effect as well as evolutionary corrections are considered in section 6 ) . we refer basically to the clumps of clusters at about 0.35 , 0.70 , 1.13 , 1.50 , and 1.93 gyr . these clumps will be identified as burst 1 , 2 , 3 , 4 , and 5 , respectively . the actual existence of burst 1 , 2 and 3 is not significantly ( see the detailed discussion below ) affected by uncertainties in the cluster ages as errors are smaller than 0.3 gyr for all clusters ( younger than 1.2 gyr ) in the samples considered . however , for clusters older than 1.2 gyr age errors are larger , therefore the statistical significance of bursts 4 and 5 is not , in principle , as well established as for 1 , 2 , and 3 . on the other hand , if we assume that the galactic disk is not currently undergoing a burst of star formation and that the age distribution for stars younger than 0.2 gyr is representative of the monotonic ( quiescent ) star formation rate in the milky way disk ( i.e. it has not been polluted by an enhanced episode of star formation ) , then the fraction of open clusters that survives for up to about 0.25 gyr is about 92% . this corresponds to the characteristic life - time of clusters with @xmath2 . in other words , observations indicate that only 8% of all the open clusters are born with stellar populations larger than about 200 stars or masses @xmath58 . this feature can be interpreted as implying that star formation takes place preferentially in clusters of that size . this result , that most of the stars appear to have been formed in small star clusters , has already been pointed out by kroupa ( 1995a , b , c ) and de la fuente marcos ( 1997b ) . however , a single 3,000 stars open cluster can produce as many field stars as 60 small ( 50 stars ) clusters and , as pointed out above , the number of large clusters violently destroyed during cluster - giant molecular cloud interactions , even though difficult to estimate , can be very significant . on the other hand , during an episode of enhanced star formation even much larger clusters ( up to 40,000 stars ) can be formed . therefore it may also be possible that most of the field stars have been formed in large clusters , however we do not observe many large clusters today and the previous possibility remains rather speculative . in any case , the cumulative age distribution for the full open cluster sample shows a departure from the predictions of constant formation rate and exponentially declining dissolution rate . assuming that the galactic disk has never been polluted by any extragalactic stellar populations , our previous analysis suggests that superimposed on a relatively small level of quiescent star formation , mainly in small-@xmath0 star clusters , the star formation rate has experienced at least 5 episodes of enhanced star formation lasting about 0.2 gyr each , with production of larger clusters . this cyclic behavior appears to show a period of [email protected] gyr . although the observed cyclic behavior can be easily explained as triggered by density waves ( hernandez et al . , 2000a ; martos et al . , 2004 ) , the analysis in this section is also exploring the possibility of periodic , tidally triggered star formation bursts as an alternative ( or possibly concurrent ) scenario to explain the enhancements observed in the open cluster age distribution . there is evidence for a recent pericentric passage of the sagittarius dwarf spheroidal galaxy @xmath3 50 myr ago ( johnston et al . , 1999 ) . if this close interaction was able to tidally trigger star formation in the disk of the milky way , its effects should currently be observable and it could be considered as burst 0 . there are in fact observational signatures of unusually rich young open clusters ( or even protoglobulars ) in the disk of the milky way : e.g. cyg ob2 ( kndlseder , 2000 ; kndlseder et al . , 2002 ; comern et al . , 2002 ; hanson , 2003 ) . we will however neglect this possibility in our subsequent discussion . the most recent star formation bursts are also the most likely bursts to have occurred , since they took place in the very recent past , and so are less affected by the age errors . unfortunately , our analysis in section 2 appears to suggest that age errors are larger for younger clusters . the youngest ( 0.25 - 0.45 gyr ) burst found in this work appears around the same age in all the three samples , lasting 0.2 gyr that is the typical duration of bursts found in previous work . this burst is temporally coincident with a perigalactic passage of the small magellanic cloud ( hereafter smc ) with the milky way as predicted by lin et al . harris and zaritsky ( 2004 ) have studied the star formation history of the smc and have found four major bursts of star formation : 0.06 , 0.4 , 2.5 , and 8.4 gyr . models by zaritsky and harris ( 2004 ) indicate that the 0.4 gyr peak in the smc stellar age function coincides with one of the epochs of closest approach between smc and the milky way . during the interaction , the elevation in the star formation rate was significant ( @xmath24 factor of 2 ) over the quiescent rate of star formation in the smc . this burst appears at 0.6 - 0.8 gyr , lasting also 0.2 gyr , in webda and ncovocc but appears to be wider , 0.6 - 1.0 gyr , for the sample including salaris et al . ( 2004 ) , shifting the peak to 0.8 gyr . johnston et al . ( 1999 ) have found that the sagittarius dwarf galaxy had a pericentric passage @xmath3 0.7 gyr ago . this close interaction is temporally coincident with burst 2 . from webda , this burst started 1.3 gyr ago , lasting about 350 myr however , ncovocc gives the same starting age but lasting 0.3 gyr . the sample corrected with salaris et al . ( 2004 ) data differs also only in the duration , 0.2 gyr . this burst is temporally coincident with a perigalactic passage of the smc with the milky way ( lin et al . , 1995 ) . the three samples indicate that the burst started 1.6 gyr ago , lasting 0.2 gyr . this burst appears to be coincident with a close encounter between the magellanic clouds and the galaxy ( murai & fujimoto , 1980 ; gardiner et al . , 1994 ; lin et al . , 1995 ) . on the other hand , the sagittarius dwarf galaxy had another pericentric passage around the same time ( johnston , 1998 ; johnston et al . , 1999 ) . this is the oldest star formation burst identified in this work and it is also the most affected by the age errors . webda and ncovocc indicate that this burst started 2 gyr ago , lasting 0.15 gyr . the third sample suggests a burst lasting 0.1 gyr . this is the only burst that does not appear to coincide with any predicted or modelled interaction with a neighbour galaxy . in spite of the time coincidences , interpreting the observed periodicity in the age distribution of young , nearby open clusters as evidence in support of a tidally triggered star formation history is somewhat speculative . on the other hand , a close interaction with the magellanic clouds can not be compared dynamically to that of the sagittarius or the canis majoris dwarf galaxies . using our method without any corrections it is , however , difficult to estimate the evolution of the absolute star formation rate with the age of the disk . nevertheless , it is possible to attempt a rough estimate : the number of clusters in bursts 1 , 2 , 3 , 4 , and 5 from the webda sample is 73 , 31 , 28 , 10 , and 5 respectively , for the dias et al . sample is 73 , 30 , 28 , 12 , and 4 and for salaris et al . is 66 , 49 , 12 , 12 , and 4 . data seem to suggest that for webda and dias et al . burst 2 was likely the strongest . if under quiescent star formation only 2% of open clusters survive for more than 2 gyr , during burst 2 the star formation increased by at least a factor 4 . salaris et al . data also indicate that burst 2 was the most important , followed by burst 4 . during burst 2 the number of open clusters formed increased by a factor 8 at least . as pointed out before , the star formation history inferred for the milky way disk in the previous section is affected by selection effects and cluster age errors , therefore some features in the derived age distribution could be caused by the incompleteness of the sample . as a result of these limitations , the actual amplitude of the bursts is likely higher . working with a flux - limited sample makes it very difficult to estimate the evolution of the absolute star formation rate with the age of the disk . nevertheless , in this section we try to assess the accuracy of our results . the age errors affect considerably the duration of the star formation events , since they tend to scatter the ages of the star clusters originally born in a burst . we can expect that this error could smear out peaks and fill in gaps in the age distribution . in section 2 , we have tested our results using synthetic age distribution diagrams from the original data ( see fig . [ error ] ) . simulations appear to indicate that for age errors of about 15% or lower bursts 2 , 3 , 4 , and 5 remain relatively unaltered both in terms of age and duration , however burst 1 seems to split into two bursts , one at 0.2 gyr and another one at 0.5 gyr . for errors in age determination larger than about 25% only bursts at 0.2 , 0.5 , and 0.7 are recovered . however , our error analysis in section 2 appears to indicate that the older the cluster , the lower the error . therefore it could be possible that errors for clusters older than 1.5 gyr are in the range 15 - 25% and the five bursts can be recovered . our derivation of the star formation history of the milky way disk uses a sample of dynamically bound , disk star clusters or open clusters . this sample is assumed as not polluted by unbound star clusters or associations . however , associations dissolve on a time - scale of @xmath3 50 myr so even if some degree of contamination remains , the contribution to the final conclusions is negligible . it is only significant for the first bin in our age distribution diagram . on the other hand , it is known that a typical cluster in the outer disk will survive about twice as long as one in the inner disk ( janes et al . , the outer disk group is an homogeneous group whose members are dissolving on a time - scale on the order of 3 - 4 gyr . this contribution is however negligible as we restrict our analysis to clusters younger than 2 gyr . in our previous analysis we have assumed that this age distribution is depopulated from extragalactic objects , so no open clusters have been incorporated into the galactic disk from accreted galaxies . an alternative , plausible but rather speculative scenario with excess clusters coming from accreted dwarf galaxies is considered in section 8 . limitations on the applicability of our method to a wider time range appear in connection with the small number of known open clusters older than 2 gyr . with a resolution of 0.05 gyr , our results are not statistically significant for an age older than 2 gyr as the number of objects per age bin is too low . the need to provide reliable results limits the age range over which we can derive the star formation history to 0 - 2 gyr , with the time resolution considered throughout the paper . however , our method can be applied to a wider time range if the time resolution is lowered ( de la fuente marcos & de la fuente marcos , 2004 ) . contrary to what is sometimes assumed in the early stages of the star formation history of the milky way disk , which were ruled by accretion events and early mergers of satellite dwarf galaxies and star clusters that contributed to the formation of a substantial fraction of the old and metal - deficient stars in the galactic disk , during the last 0.2 gyr the star formation rate ( see fig . [ vrec ] ) has been remarkably stable with apparently no events of enhanced star formation . [ vrec ] appears to indicate that the open cluster sample younger than 0.2 gyr can be considered representative in the analysis of quiescent star formation in the milky way disk . the characteristic time - scale to assemble and _ give birth _ to an open cluster seems to be @xmath3 10 myr with ( at least ) about 80 new open clusters being formed every 10 myr or 8 clusters / myr within about 3 kpc from the sun . about 80% of newly formed clusters are destroyed during the first 20 myr of life . for a time resolution of 5 myr ( fig . [ vrec ] , small panel ) a very interesting feature is visible : clusters younger than 5 myr are scarce . in principle , this feature may be interpreted as a minimum in the star formation rate , however young clusters are still embedded in their parent molecular cloud and this makes them difficult to detect . dust obscuration appears to be the natural explanation for the origin of this feature . on the other hand , there is a sharp decrease in the number of open clusters observed for ages older than 15 - 20 myr . there is an initial very rapid rate of cluster dissolution , gradually declining thereafter . again the most simple interpretation is to assume that an enhanced star formation period started @xmath3 20 myr ago . however , this feature can be better explained if we consider stellar evolution in open clusters . for a metallicity @xmath59 = 0.02 , stellar evolution models ( e.g. schaller et al . , 1992 ; bressan et al . , 1993 ) indicate that stars with masses above 10 @xmath32 and older than about 20 myr have already exploded as supernovae . for a small cluster ( @xmath60 75 ) , the energy released during one supernova event can be equivalent to the binding energy of the entire cluster . therefore , early supernovae appear to be responsible for the violent destruction of about 80% of young open clusters . in the following section , we use this information to generate a power - law open cluster initial mass function with slope -2.7 . the frequency distribution of stellar masses at birth , the so - called stellar initial mass function ( hereafter simf ) , is a fundamental parameter to study the stellar mass spectrum . salpeter ( 1955 ) used the observed luminosity function for the solar neighbourhood and theoretical evolutionary times to derive a simf which may be approximated by a power - law : @xmath61 where @xmath62 is the number of stars per unit mass interval . the original value of @xmath63 found by salpeter is 2.35 , for masses between 0.4 and 10.0 @xmath32 . the determination of the simf is a cornerstone in astrophysics , since the simf determines the evolution , surface brightness , chemical enrichment , and baryonic content of all galaxies and it is also a fundamental link between stellar and galactic evolution . for recent reviews on the simf see kroupa ( 2002 ) and chabrier ( 2003 ) . as many of the stars in the disk of the milky way appear to be the result of a clustered mode of star formation , the open cluster initial mass function ( hereafter ocimf ) is also a fundamentally important distribution function to study cluster formation , chemical evolution of galaxies , and star formation in general . the fact that the simf and the ocimf can be intimately related was first pointed out by reddish ( 1978 ) . however , deriving the ocimf is very difficult as cluster mass determinations are affected by several observational and theoretical biases : obscuration by dust , flux limits , incomplete sampling of cluster haloes , lack of information on the evolutionary state of the cluster , and uncertainties in stellar models and cluster dissolution rates . although the ocimf is not well constrained it can be written as a power law of the type @xmath64 for @xmath65 , @xmath66 is the number of open clusters per unit mass interval . the cumulative distribution will be @xmath67 if the process by which open clusters are formed is also able to produce individual stars , it appears reasonable that both simf and ocimf should show similar slopes within the observational errors . reddish ( 1978 ) found a slope equal to -2.2 from a sample of 72 open clusters . vzquez and feinstein ( 1989 ) estimated the ocimf for the milky way disk using a sample of 130 clusters from lyng s ( 1987 ) catalogue . in order to find the distribution , they used the values of the masses quoted in the catalogue to obtain a slope @xmath68 for all the clusters in their sample . they also found that the slope is different for clusters with galactocentric distances smaller than the sun s and clusters with larger galactocentric distances , although this may be an artifact due to incompleteness of the data . the topic of the distribution of the masses of star clusters has been re - visited for various star - forming galaxies ( e.g. bik et al . , 2003 ; elmegreen et al . , 2000 ; whitmore et al . , 1999 ) . kroupa and boily ( 2002 ) have recently approached the subject from a theoretical point of view focusing mainly on large clusters . they found that the entire galactic population ii stellar spheroid can be generated if star formation proceeded via embedded clusters distributed like a power - law mass function with slope in the range ( 1.9 , 3.6 ) . in this section we will infer the ocimf for the quiescent milky way disk without considering the values of the cluster masses provided by the catalogues but taking into account the life - time scale obtained from simulations in section [ new ] and the open cluster sample described in section [ sample ] to derive the ocimf for the milky way disk . this method to construct an ocimf has been outlined in kroupa and boily ( 2002 ) . in section [ history ] it has been evidenced that the disk of the milky way may have experienced several episodes of enhanced star formation over the last 2 gyr , this implies that the ocimf can not be derived from the total sample of clusters , because the starburst will bias the ocimf towards massive clusters . however , if we restrict our analysis to clusters with ages less than about 200 myr , these clusters can be used to derive the ocimf . this choice implies that the star formation rate has remained almost constant during the last 200 myr ; this assumption appears to be reasonably supported by our previous analysis . for an average stellar mass of 0.4 @xmath32 , we obtain that @xmath3 65% of clusters have masses @xmath69 , about 17% with @xmath70 , 10% with @xmath71 , 6% with @xmath72 , and @xmath3 2% have masses @xmath73 , with a likely average of 150 @xmath32 . observations and simulations seem to suggest that @xmath74 ( @xmath75 ) and @xmath76 ( @xmath77 ) . these results are consistent with an open cluster population being sampled from a power - law ocimf with @xmath78 with a correlation coefficient @xmath79 ( see fig . [ ocimf ] ) . the error associated with this determination can be estimated from the standard deviation obtained for the life - time of simulated clusters and is @xmath80 . in this analysis we are not taking into consideration that a number of clusters ( those younger than @xmath81 myr ) may be still in the evolutionary stage driven by gas expulsion . this simplification is likely to have a minor impact on the overall results from this section because our results indicate that a significant fraction of open clusters formed during quiescence are in fact small and therefore short - lived . in the previous sections we have assumed that the open cluster sample under study is representative of the entire galactic disk , therefore the evolution of the star formation rate can be inferred from its age distribution , since the number of open clusters in each age bin has to be correlated with the number of objects initially at that time as a result of dissolution processes . the most reliable transformation of the open cluster age distribution into history of the star formation rate comprises two intermediate corrections , namely the volume and evolutionary corrections . as stated in section 4 , a scale height correction can be neglected because the studied age range is @xmath82 gyr . the applied corrections are explained in the following sections . as pointed out in section 3 , a correction due to the dynamical evolution of the cluster ( cluster disintegration ) is needed because our sample includes clusters with different initial populations ( @xmath10 ) . the more massive clusters have a life expectancy higher than the short - lived , small-@xmath0 clusters , thus the latter would be missing in the older age bins . it is however possible to correct for this effect using the ocimf obtained in the previous section as well as the results from section 3 . the corrections are given by the following formalism . the number of clusters born a time @xmath83 ago ( @xmath83 = 0 is present time ) is the current number observed at age @xmath83 divided by the fraction of surviving clusters expected for an age @xmath83 , @xmath84 . the fraction of surviving clusters changes with @xmath83 , therefore the life - time scale obtained in section 3 is used to estimate the correct fraction of surviving clusters to be applied . in detail : for the second bin we have corrected by the same number in bin 1 ; bins 3 - 4 have been corrected using @xmath84 = 0.18 ; @xmath84 = 0.08 has been used in bins 5 - 9 ; bins 10 - 16 have been corrected by @xmath84 = 0.02 . after 0.8 gyr , we have assumed that no cluster formed during quiescent star formation epochs may have survived ; therefore , the correction is made in a slightly different way : we add the reference level at @xmath83 = 0 to the standard correction with @xmath84 = 0.02 . the age histogram from webda and the corrected distribution appear in fig . the star formation history outlined in section 4 is somewhat altered : this burst appears at 0.20 - 0.35 gyr and it seems to be not very prominent . this burst appears at 0.6 - 0.8 gyr , lasting also 0.2 gyr . it appears to be much stronger than burst 1 . it started 1.3 gyr ago and lasted 0.35 gyr . it was the strongest . the unusually high level of enhanced star formation could be the result of the combined action of density waves and a strong tidal interaction with the smc . this burst appears at 1.4 - 1.6 gyr , lasting also 0.2 gyr . it appears to be similar in strength to burst 2 . this burst started 2 gyr ago , lasting 0.15 gyr . its activity was lower than for bursts 2 and 4 but much higher than 1 . since our original open cluster sample is not volume - limited , there could be a bias in the relative number of objects in each age bin : open clusters with different initial populations and metallicities have different integrated luminosities , thus the volume of space sampled varies from cluster to cluster . in principle , our method will produce optimal results for volume - limited open cluster samples . in this section we consider a cluster sample in the solar neighbourhood . the solar neighbourhood is defined as a volume centered on the sun that is much smaller than the overall size of the milky way galaxy and yet large enough to contain a statistically useful sample of stars ( see , e.g. , binney & tremaine , 1987 ) . the appropriate size of the volume depends on which stars or objects are going to be investigated : for white dwarfs , which are both common and faint , it may consist of a sphere of radius 10 pc centred on the sun , while for the bright but rare o and b stars , the solar neighbourhood may be considered to extend as far as 1 kpc from the sun . in our analysis we will consider the latter in order to retain a statistically significant sample of open clusters . there are 161 open clusters in our sample within 1 kpc of the sun ( and younger than 2 gyr ) . if we restrict our analysis to clusters closer than 1 kpc , the age distribution plotted in fig . [ sn ] is obtained . burst 1 is clearly identified at an age of 0.35 gyr with a duration of about 0.2 gyr . burst 2 is also present for an age of about 0.7 gyr with a duration of about 0.2 gyr . burst 3 appears at 1.15 gyr and is 0.1 gyr wide . unfortunately , it is difficult to recover burst 4 and 5 as the number of clusters older than 1.2 gyr within 1 kpc from the sun is very low ( 3 objects ) . in fact , our results appear to suggest that the solar neighbourhood experienced quiescent star formation at 1.2 - 2 gyr . from fig . [ sn ] , the star formation rate history in the solar neighbourhood seems to be slightly different , with not a well defined period and narrower maxima . it is , however , easy to reconcile these apparent differences . the age distribution features interpreted as bursts 1 , 2 , and 3 are clearly identified in fig . burst4 and 5 may also exist but there are too few objects to confirm the hypothesis . on the other hand , uncertainties in the open cluster ages are smaller for closer clusters , therefore maxima are narrower and show better contrast against steady , quiescent star formation . [ bestp ] shows the solar neighbourhood sample with evolutionary correction as described in the previous section . in this case the second burst is the strongest , followed by the third , the fifth , and the first . the fourth one is now the least important . the problem of deducing the star formation history of our galaxy has been primarily studied through chemical evolution models although other methods have also been used : hertzsprung - russell diagram inversion , stellar evolutionary tracks ( twarog , 1980 ; meusinger , 1991 ; bertelli & nasi , 2001 ) , chromospheric activity as measured by ca ii and k emission ( barry , 1988 ; soderblom et al . , 1991 ; rocha - pinto et al . 2000a , b ) , stellar kinematics ( gomez et al . , 1990 ; marsakov et al . 1990 ; chereul et al . , 1998 ; just , 2002 , 2003 ) , the main - sequence luminosity function ( scalo , 1987 ) , the white dwarf luminosity function ( noh & scalo , 1990 ; diaz - pinto et al . , 1994 ; isern et al . , 1999 ) , combining the metallicity distribution and age - metallicity relation of g dwarfs ( rocha - pinto & maciel , 1997 ) and the distribution of coronal emission as measured by x - ray luminosities ( micela et al . , 1993 ; lachaume et al . , 1999 ) . most of these studies have inferred an star formation history that is non - monotonic with time . although the majority of studies in this field use samples of stars in the solar neighbourhood with no stars more distant than about 100 pc being considered , it does not mean that the star formation history derived can only be applied to stars born in the solar neighbourhood . nearby stars older than about 0.2 gyr . come from birth sites which span a large range in galactocentric distances . wielen ( 1977 ) showed that the orbital diffusion coefficient deduced from the observed increase of velocity dispersion with age implies that such stars have suffered an rms azimuthal drift of about 2 kpc for an age of 0.2 gyr . considerable , but smaller , drift should occur also in the radial direction . wielen et al . ( 1996 ) , on the basis of the sun s metallicity and the radial metallicity gradient in the galactic disk , estimated that the sun has migrated outward by [email protected] kpc in the past 4.5 gyr . sellwood and binney ( 2002 ) have discussed the dynamics of radial migration due to transient spiral arms , and the relationship between radial migration and disk heating . they estimate that old stars formed in the solar neighbourhood should now be scattered nearly uniformly throughout an annulus of [email protected] kpc . in this sense the star formation history inferred for nearby stars is a measure of the global star formation history , at least at the sun s galactocentric radius . it is , however , not clear how the diffusion process affects entire star clusters not standalone stars and it is likely that star clusters experience little or no ( azimuthal , radial ) drift over the life - time . hernandez et al . ( 2000a ) have derived the star formation history of the solar neighbourhood over the last 3 gyr using data from the _ hipparcos _ catalogue to construct color - magnitude diagrams . they treated these diagrams using advanced bayesian analysis techniques ( hernandez et al . , 1999 ; hernandez et al . , 2000 ) to deduce the star formation rate history of this region . they recovered the star formation history with an unprecedented time resolution of 0.05 gyr . this high resolution makes it easy to compare our present results with those in their papers ( hernandez et al . 2000a , b , 2001 ) . their results indicate that the local star formation rate has an oscillatory component of period @xmath85 gyr superimposed on a small level of constant star formation activity . their inferred star formation history appears to be compatible with the observed distribution of stars . this cyclic behaviour is interpreted as the result of repeated encounters with the galactic arm density pattern . in their work , they find that the last episode of enhanced star formation happened about 0.5 gyr ago . another burst is found at about 1.3 gyr and yet another one at 1.9 gyr . their fig . 4 ( hernandez et al . , 2000a ) however suggests a double peak for the youngest burst and also for the one at 1.3 gyr . in fact , their results appear to indicate one burst at about 0.4 gyr ( our burst 1 ) and another one at 0.7 gyr ( our burst 2 ) . the 1.3 gyr burst may be the convolution of burst 3 and 4 found in our analysis . their third burst coincides in time with our fifth one . therefore , our derivation of the recent star formation history both in the solar neighbourhood and the galactic disk can be considered as fully consistent with the results found by hernandez et al . ( 2000a ) in spite of the fact that a completely different technique has been used in its derivation . llll & this work & hernandez et al . & rocha - pinto et al . + & & ( 2000a ) & ( 2000a ) + number of bursts & 5 & 3 & 1 + age of burst 1 & 0.4 gyr & 0.5 gyr & 0 - 1 gyr + age of burst 2 & 0.8 gyr & 1.3 gyr & - + age of burst 3 & 1.2 gyr & 1.9 gyr & - + age of burst 4 & 1.5 gyr & - & - + age of burst 5 & 1.9 gyr & - & - + [ comparison ] rocha - pinto et al . ( 2000a ) have derived a star formation history from the chromospheric activity - age distribution of a large sample comprising 552 stars with _ hipparcos _ parallaxes in the solar neighbourhood ( stars more distant than 80 pc were omitted ) , and have found intermittency in the star formation rate over 14 gyr . with a time resolution of 0.4 gyr , their history of the star formation rate plotted in fig . 2 ( rocha - pinto et al . , 2000a ) and table 1 ( rocha - pinto et al . , 2000b ) indicates that the solar neighbourhood experienced a burst of star formation 0 - 1 gyr ago ( burst a ) . an epoch of low star formation took place 1 - 2 gyr ago and a new , very strong burst was experienced 2 - 5 gyr ago ( burst b ) . an additional burst was found 7 - 9.5 gyr ago ( burst c ) . although their time resolution makes a direct comparison impossible , their burst a is compatible with a convolution of our bursts 1 and 2 , and possibly 3 . there are , however , strong discrepancies in the 1 - 2 gyr range . noh and scalo ( 1990 ) using the white dwarf luminosity function found peaks of star formation at 0.3 gyr and 1.8 gyr , that are compatible with our bursts 1 and 5 . barry ( 1988 ) using a technique similar to the one in rocha - pinto et al . ( 2000a ) , has also found three bursts using a volume - limited sample of 115 f - g stars , the most recent peaked in the last 0.5 gyr ( burst a ) . an epoch of low star formation took place 1 - 4 gyr ago and a new burst was experienced 4 - 6 gyr ago ( burst b ) . the oldest burst was found 7 - 11 gyr ago ( burst c ) . his burst a appears to be a convolution of our bursts 1 and 2 . using the main - sequence luminosity function , scalo ( 1987 ) found signatures of starbursts 5 - 6 and 0.3 gyr ago . his 0.3 gyr burst coincides with our burst 1 . twarog ( 1980 ) found that during the last 4 gyr the star formation rate has remained more or less constant but there was a sharp increase from 4 to 8 gyr ago . although the time resolution in these studies is rather low ( 0.2 - 1.0 gyr ) , their results seem to indicate that bursts 1 , 2 , and 5 have also been identified by other authors . starburst events are often seen in interacting galaxies , and it seems reasonable to infer a causal connection between galaxy interactions ( mergers and close encounters ) and starburst activity . it is now becoming increasingly clear that average galaxies engulf and devour smaller satellite galaxies as part of a more general process of hierarchical merging . our own galaxy , the milky way , may have been incorporating dwarf galaxies into its disk since it was formed . the galaxy is encircled by satellite galaxies that appear confined to one of two great streams across the sky ( lynden - bell & lynden - bell , 1995 ) . the most well - known of these are the magellanic clouds and the associated hi magellanic stream . all of these are expected to merge with the milky way in the distant future , largely due to the dynamical friction from the extended halo . the first strong evidence in favour of a hierarchical merging scenario for the milky way was presented by ibata et al . ( 1994 , 1995 , 1997 ) when they discovered the sagittarius dwarf galaxy , a low mass dwarf spheroidal galaxy about 25 kpc from the sun in an advanced state of disruption that is being absorbed by the disk of the milky way . in fact , our results appear to indicate that the sagittarius dwarf galaxy may have had some role in inducing enhanced star formation in the milky way disk . for about 10 years , the sagittarius dwarf galaxy remained as the only strong evidence that the milky way is incorporating satellite galaxies , but another object has been added recently . the canis major galaxy ( martin et al . , 2004 ; bellazzini et al . , 2004 ) appears to be a close relative of the sagittarius dwarf galaxy in terms of structural parameters . this new finding seems to confirm that part of the galactic disk could be extragalactic in origin . on the other hand , recent numerical models of disk galaxy formation ( abadi et al . , 2003a , b ) indicate that disrupted galaxy satellites might have contributed a significant fraction of the old stars in the disk of the milky way . navarro et al . ( 2004 ) have confirmed that the arcturus group , a dynamically - coherent group of 10 - 12 gyr old , metal - deficient stars located in the solar neighbourhood around the star arcturus , are very likely of extragalactic origin . these authors suggest that the group is part of the tidal debris of a 10@xmath86 @xmath32 disrupted satellite accreted by the milky way about 8 gyr ago . in this scenario , the peaks in the open cluster age distribution can be interpreted as signatures of merger events . this is however a non - straightforward interpretation as the variations found in the star formation rate appear to be periodic in nature , and mergers are expected to occur randomly in time . however , the actual sequence of events driving enhanced star formation episodes could be even more complex , with both tidal interactions and accretion events contributing to the global star formation history , as the unusual strength of burst 3 suggests . for the main open cluster sample used in this work ( mermilliod , 2003 ) , we find 97 clusters in the age range 0.2 - 0.6 gyr and 80 objects in the age range 0.6 - 2.0 gyr . there is a clear excess of clusters older than 0.6 gyr . strong star bursts are able to explain this feature but it could also be possible that some of these clusters may have been incorporated by the galactic disk during accretion events in the recent past . in any case , the accretion scenario can only be important for coplanar interactions with the disk of the milky way and large accreted star clusters , otherwise the star clusters surviving the merger would end up either destroyed or in the galactic halo . the chain of dynamical events that provoked the formation of the recently discovered canis major dwarf galaxy remnant ( martin et al . , 2003 ; bellazzini et al . , 2003 ) has been proposed as the process that built up the monoceros ring ( newberg et al . , 2002 ; yanny et al . , 2003 ; ibata et al . , 2003 ; crane et al . , 2003 ; rocha - pinto et al . , 2003 ; majewski et al . , 2004 ) . an unusually high spatial density of open clusters has been found in the field of the canis major galaxy . the list of open clusters includes , at least : ngc 2204 , 2243 , and 2477 , tombaugh 1 and 2 , berkeley 20 and 33 , melotte 66 , and am-2 . bellazzini et al . ( 2003 ) have also found that the position , distance , and stellar population of the old open clusters am-2 and tombaugh 2 strongly suggest that they are part of the canis major galaxy . both are older than 4 gyr and therefore they have not been included in our age distributions for clusters younger than 2 gyr . nevertheless , if the physical association between these open clusters and the merged dwarf galaxy is confirmed , it will be a clear example of _ intergalactic stellar pollution _ , with _ foreign _ star clusters being absorbed by the galactic disk . if true , it may have a major impact on the determination of the star formation history of the milky way , with a fraction of the observed stars and open clusters being actually outsiders coming from other star formation histories . in this paper we have shown that open clusters can be used as tracers to investigate the star formation history of the milky way disk . if a number of dwarf galaxies with their cohorts of star clusters have been captured and disrupted in the past , open clusters absorbed from these galaxies can be used to find the trail left by the parent galaxy . searching for groups of coeval old open clusters with similar distances located in the same region of the sky can help to identify the bulk of the accreted satellite . does the galaxy form stars continuously , or in bursts separated by epochs of relative quiescence ? if star formation occurs in bursts , what processes mediate the bursts ? these are just two of the many questions that we have tried to answer in this paper . we have applied a new method to carry out an objective reconstruction of the star formation history of the milky way disk over the last 2 gyr . a sample composed of 581 open clusters with known ages and distances was used in the derivation of this star formation history . peaks in the age distribution diagram of the cluster sample were interpreted as signatures of star formation bursts . our results indicate that the analyzed star formation rate presents two components : periodic episodes of enhanced star formation superimposed on a quiescent star formation level . a constant star formation history during the last 2 gyr can therefore be discarded . a uniform rate of cluster formation and an exponentially declining dissolution rate would have produced a single population of clusters that would follow a straight line in a logarithmic plot with a slope given by the characteristic life - time of the population . evidence for at least five epochs of enhanced star formation during the time interval studied has been found . the recent star formation history derived in this work is consistent with star formation rate histories deduced using a range of other techniques . interpreting the age distribution diagram suggests that quiescent ( non - enhanced ) star formation has proceeded in discrete , highly - correlated regions of activity producing clusters which dissipate their parent molecular cloud on a time - scale of a few myr ( @xmath9 10 myr ) . the majority of open clusters formed during the non - enhanced star formation periods are destroyed and their members become field stars on a time - scale @xmath9 20 myr . however , there appears to be a long - lasting component as well , since some of the clusters are able to survive for over 0.2 gyr . during bursts of star formation , the huge number of clusters formed make it possible to produce larger than average objects that could survive for several gyr . a number of simulations were done in order to estimate the impact of the age errors on the features found in the age distribution diagram . even for age errors larger than 30% some of these features ( the younger ones ) are still clearly identifiable . on the other hand , we have examined the possibility that the galactic bursts are coeval with features in the star formation history of the magellanic clouds and the sagittarius dwarf galaxy , as well as with close encounters between them and the milky way . although the degree of uncertainty is large , there are several coincidences that suggest tidal interactions can play some role as inductive forces of bursts of star formation in the milky way disk . distant galactic encounters may trigger significant star formation but quiescent star formation in the disk of the milky way appears to be characterized by a large number of small , short lived open clusters that contribute a small fraction of the total number of stars formed ( @xmath9 150 per cluster ) with a small number of large , relatively long lived clusters that may contribute most of the stars in the disk . there is , however , a cyclic behaviour in the burst sequence that may be better explained by the density wave hypothesis ( lin & shu , 1964 ) for the presence of spiral arms in late - type galaxies . a model like the one outlined in hernandez et al . ( 2000a ) and developed in martos et al . ( 2004 ) can explain easily the 0.4 gyr periodicity that we detect . heating by spiral structure can explain the main features of the age - velocity dispersion relation and other parameters of the velocity distribution in the solar neighbourhood , like why the stars in a single velocity - space moving group have a wide range of ages ( de simone et al . , 2004 ) . for a pattern speed @xmath87 = 20 km s@xmath88 kpc@xmath88 ( e.g. martos et al . , 2004 ) and a galactocentric distance @xmath6 = 8.5 kpc it implies an orbital period of about 1 gyr for the sun . if the enhanced star formation episodes are , in fact , due to the interactions with the spiral arms it means that our galaxy has two arms . this has been recently suggested by martos et al . ( 2004 ) using a completely different approach . on the other hand , studies on the nature of star formation triggering in large disks indicate that a number of mechanisms may operate concurrently : gravitational instabilities , shocks between colliding clouds of gas , enhanced pressure of the interstellar medium , strong stellar winds , supernova shocks ( star formation triggered by nearby bursts of star formation ) , density waves , shear forces produced by differential rotation , and interactions or mergers with other galaxies . star formation triggered by previous star formation events is a self - propagating process which may continue over a much longer period of time , hence it is possible that overall a relatively large fraction of star formation is triggered . on the other hand , if the star formation history inferred from the age distribution of open clusters is consistent with that deduced for individual stars then one can consider this as an argument in favour of most of the observed stars being formed in some kind of star clusters . since only a subsample of the young clusters are likely to survive , an obvious question is whether most of the field stars in a galaxy are originally formed in clusters . even in these young star forming regions many of the field stars are from clusters that have already been dissolved , hence the true percentage of stars that were originally in clusters is even higher , and might conceivably be @xmath89 100% . it is tempting to suggest that perhaps the majority of stars are formed in groups and clusters , and that the field stars are simply the remnants of the fainter , less dense clusters which have dissolved . in fact , infrared observations show that most star formation occurs in embedded star clusters within giant molecular clouds . these clusters have masses in the range 50 - 10@xmath90 @xmath32 ( lada & lada , 2003 ) . nevertheless , it is still possible to consider the existence of two channels in the star formation process : the dispersed mode and the clustered mode , by modifying slightly our concepts of dispersed and clustered . the dispersed mode would be represented by the numerous short lived ( @xmath41 30 myr ) small star clusters that are part of the quiescent star formation rate . these are different from classical open clusters and may be called _ fast living clusters _ or _ flusters _ in opposition to the fully fledged standard open clusters as the hyades or pleiades . the clustered mode would be represented by the relatively long lived open clusters . in our analysis of the last 0.2 gyr of star formation , the number of clusters in the age range 30 - 100 remains relatively constant as well as the fraction in the age range 100 - 200 . this feature suggests a real gap in the mass distribution of disk star clusters : very few clusters are formed in the mass range 20 - 30 @xmath32 ( @xmath0 = 50 - 75 ) . to explain the very good matching between the star formation history inferred from individual stars and that from open star clusters we have to assume that although most of open clusters ( @xmath91 90% ) are poor , and short lived ( @xmath92 200 ) most of the stars ( @xmath91 90% ) are formed in larger clusters ( @xmath39 200 ) . this conclusion , however , appears not to be supported by current observations . the third largest galaxy in the local group , triangulum galaxy ( m33 , ngc 598 ) may be undergoing star formation events similar to the ones described in this paper . m33 is a spiral galaxy about half the size of the milky way . chandar et al . ( 1999 ) have found 44 young clusters in m33 with ages @xmath4 100 myr and masses in the range @xmath93 to @xmath94 . currently , m33 appears to be forming many young compact clusters but less massive than an older cluster population formed in a previous starburst with characteristic mass @xmath95 . although our conclusions are uncertain as they are based on a relatively small ( 36% ) subsample of a larger but still incomplete sample composed of almost 2000 open clusters , they are similar to the ones previously found by a number of other independent studies . this confirms the validity of our approach to infer star formation histories ( recent and old ) , not just for the milky way disk but for any other galaxy . on the other hand , if the conclusions obtained in this paper are correct , the galactic disk will start another episode of enhanced star formation within the next very few myr if it is not already happening right now . we thank the referee whose comments greatly improved this paper . we are very grateful to s.j . aarseth for providing his computer code and for his comments on the manuscript . we thank the department of astrophysics of universidad complutense of madrid for providing excellent computing facilities at the centro de proceso de datos in moncloa . the authors thank s. van den bergh , l. o. lodn , j .- c . mermilliod , g. carraro , and i. platais for very helpful and interesting discussions . we also thank m. martos , x. hernndez and associated group for sharing with us their results on the galactic spiral pattern and its rotation speed in advance of publication . in preparation of this paper we made use of the open cluster database , operated at the _ institut dastronomie de luniversit de lausanne _ , switzerland , the new catalogue of optically visible open clusters and candidates , operated at the _ universidade de so paulo _ , brazil , the nasa astrophysics data system and the astro - ph e - print server . aarseth , s.j . , 1985 , in : multiple time scales , eds . brackbill & b.i . cohen , academic press , new york , p. 377 . aarseth , s.j . , 1994 , in : galactic dynamics and @xmath0-body simulations , eds . g. contopoulos , n.k . spirou , & l. vlahos , springer - verlag , berlin , p. 277 . aarseth , s.j . , 1999 , pasp , 111 , 1333 . aarseth , s.j . , 2003 , gravitational @xmath0-body simulation : tools and algorithms ( cambridge university press , cambridge ) . aarseth , s.j . & zare , k. , 1974 , cem , 10 , 185 . abadi , m.g . , navarro , j.f . , steinmetz , m. , & eke , v.r . , 2003a , apj , 591 , 499 . abadi , m.g . , navarro , j.f . , steinmetz , m. , & eke , v.r . , 2003b , apj , 597 , 21 . ahmad , a. & cohen l. , 1973 , j. comput . , 12 , 389 . ann , h.b . , lee , s.h . , sung , h. , lee , m.g . , kim , s .- l . , chun , m .- y . , jeon , y .- b . , park , b .- , & yuk , i .- s . , 2002 , aj , 123 , 905 . barry , d.c . , 1988 , apj , 334 , 436 . battinelli , p. , brandimarti , a. , & capuzzo - dolcetta , r. , 1994 , a&as , 104 , 379 . baume , g. , villanova , s. , & carraro , g. , 2003 , a&a , 407 , 527 . baumgardt , h. , 2001 , mnras , 325 , 1323 . baumgardt , h. & makino , j. , 2003 , mnras , 340 , 227 . bellazzini , m. , ibata , r.a . , monaco , l. , martin , n.f . , irwin , m.j . , & lewis , g.f . , 2004 , mnras , ( in press ) . van den bergh , s. , 2000 , pasp , 112 , 932 . bertelli , g. & nasi , e. , 2001 , aj , 121 , 1013 . bik , a. , lamers , h.j.g.l.m . , bastian , n. , panagia , n. , & romaniello , m. , 2003 , a&a , 397 , 473 . binney , j. & tremaine , s. , 1987 galactic dynamics ( princeton university press , princeton ) . bressan , a. , fagotto , f. , bertelli , g. , & chiosi , c. , 1993 , a&as , 100 , 647 . brown , a.g.a . , 2002 , in : modes of star formation and the origin of field populations , eds . grebel & w. brandner ( asp conference series ) , p. 150 . bruntt , h. , frandsen , s. , kjeldsen , h. , & andersen , m.i . , 1999 , a&as , 140 , 135 . carraro , g. , hassan , s.m . , ortolani , s. , & vallenari , a. , 2001 , a&a , 372 , 879 . carraro , g. , barbon , r. , & boschetti , c.s . , 2002 , mnras , 336 , 259 . carraro , g. & munari , u. , 2004 , mnras , 347 , 625 . chabrier , g. , 2003 , pasp , 115 , 763 . chandar , r. , bianchi , l. , ford , h.c . , salasnich , b. , 1999 , pasp , 111 , 794 . chereul , e. , crez , m. , & bienaym , o. , 1998 , a&a , 340 , 384 . clarke , c.j . , bonnell , i.a . , & hillenbrand , l.a . , 2000 , in : protostars and planets iv , eds . v. mannings , a.p . boss , & s.s . russell ( university of arizona press ) , p. 151 . comern , f. , et al . , 2002 , a&a , 389 , 874 . crane , j.d . , majewski , s.r . , rocha - pinto , h.j . , frinchaboy , p.m. , skrutskie , m.f . , & law , d.r . , 2003 , apj , 594 , l119 . dias , w.s . , alessi , b.s . , moitinho , a. , & lpine , j.r.d . , 2002 , a&a , 389 , 871 . dias , w.s . , alessi , b.s . , moitinho , a. , & lpine , j.r.d . , 2003a , in : galactic and stellar dynamics , eds . boily , p. patsis , s. portegies zwart , r. spurzem , & c. theis , eas publications series , p. 195 . dias , w.s . , alessi , b.s . , moitinho , a. , & lpine , j.r.d . , 2003b , new catalogue of optically visible open clusters and candidates version 1.4 , http://www.astro.iag.usp.br/ wilton . daz - pinto , a. , garcia - berro , e. , hernanz , m. , isern , j. , & mochkovitch , r. , 1994 , a&a , 282 , 86 . eggleton , p.p . , fitchett , m.j . , & tout , c.a . , 1989 , apj , 347 , 998 . elmegreen , b.g . , efremov , y. , pudritz , r.e . , & zinnecker , h. , 2000 , in : protostars and planets iv , eds . v. mannings , a.p . boss , & s.s . russell ( university of arizona press ) , p. 179 . esa , 1997 , the hipparcos catalogue , esa sp-1200 . freeman , k. & bland - hawthorn , j. , 2002 , ara&a , 40 , 487 de la fuente marcos , r. , 1993 , unpublished dissertation , universidad complutense de madrid . de la fuente marcos , r. , 1995 , a&a , 301 , 407 . de la fuente marcos , r. , 1996a , a&a , 308 , 141 . de la fuente marcos , r. , 1996b , a&a , 314 , 453 . de la fuente marcos , r. , 1997a , a&a , 322 , 764 . de la fuente marcos , r. , 1997b , ph.d . thesis , universidad complutense de madrid . de la fuente marcos , r. , 1998 , a&a , 333 , l27 . de la fuente marcos , r. & de la fuente marcos , c. , 2000 , ap&ss , 271 , 127 . de la fuente marcos , r. & de la fuente marcos , c. , 2002 , ap&ss , 371 , 1097 . de la fuente marcos , r. & de la fuente marcos , c. , 2004 , ( in preparation ) . gardiner , l.t . , sawa , t. , & fujimoto , m. , 1994 , mnras , 266 , 567 . gizis , j.e . , reid , i.n . , & hawley , s.l . , 2002 , aj , 123 , 3356 . gomez , a.e . , dehaye , j. , grenier , s. , jaschek , c. , arenou , f. , & jaschek , m. , 1990 , a&a , 236 , 95 . gray , d.f . , 1965 , aj , 70 , 362 . grocholski , a.j . & sarajedini , a. , 2002 , aj , 123 , 1603 . hanson , m.m . , 2003 , apj , 597 , 957 . harris , j. & zaritsky , d. , 2004 , aj , ( in press ) hernandez , x. , gilmore , g. , & valls - gabaud , d. , 2000 , mnras , 317 , 831 . hernandez , x. , valls - gabaud , d. , & gilmore , g. , 1999 , mnras , 304 , 705 . hernandez , x. , valls - gabaud , d. , & gilmore , g. , 2000a , mnras , 316 , 605 . hernandez , x. , valls - gabaud , d. , & gilmore , g. , 2000b , in : the evolution of the milky way : stars versus clusters , eds . f. matteucci & f. giovannelli ( kluwer academic publishers ) , p. 517 . hernandez , x. , valls - gabaud , d. , & gilmore , g. , 2001 , in : galaxy disks and disk galaxies , eds . funes & e.m . corsini ( asp conference series ) , p. 17 . ibata , r.a . , gilmore , g. , & irwin , m.j . , 1994 , nature , 370 , 194 . ibata , r.a . , gilmore , g. , & irwin , m.j . , 1995 , mnras , 277 , 781 . ibata , r.a . , wyse , r.f.g . , gilmore , g. , irwin , m.j . , & suntzeff , n.b . , 1997 , aj , 113 , 634 . ibata , r.a . , irwin , m.j . , lewis , g.f . , ferguson , a.m.n . , & tanvir , n. , 2003 , mnras , 340 , l21 . isern , j. , hernanz , m. , garcia - berro , e. , & mochkovitch , e. , 1999 , in : 11th european workshop on white dwarfs , eds . solheim & e.g. meistas ( asp conference series ) , p. 408 . janes , k.a . , tilley , c. , & lyng , g. , 1988 , aj , 95 , 771 . johnston , k.v . , 1998 , apj , 495 , 297 . johnston , k.v . , majewski , s.r . , siegel , m.h . , reid , i.n . , & kunkel , w.e . , 1999 , aj , 118 , 1719 . just , a. , 2002 , in : disks of galaxies : kinematics , dynamics and perturbations , eds . e. athanassoula , a. bosma , & r. mujica ( asp conference series ) , p. 117 . just , a. , 2003 , ap&ss , 284 , 727 . kndlseder , j. , 2000 , a&a , 360 , 539 . kndlseder et al . , 2002 , a&a , 390 , 945 . kroupa , p. , 1995a , mnras , 277 , 1491 . kroupa , p. , 1995b , mnras , 277 , 1507 . kroupa , p. , 1995c , mnras , 277 , 1491 . kroupa , p. , 2002 , science , 295 , 82 . kroupa , p. & tout , c.a . , 1992 , mnras , 259 , 223 . kroupa , p. & boily , c.m . , 2002 , mnras , 336 , 1188 . kroupa , p. , tout , c.a . , & gilmore , g. , 1990 , mnras , 244 , 76 . kroupa , p. , tout , c.a . , & gilmore , g. , 1991 , mnras , 251 , 293 . kroupa , p. , tout , c.a . , & gilmore , g. , 1993 , mnras , 262 , 545 . kroupa , p. , aarseth , s.j . , & hurley , j. , 2001 , mnras , 321 , 699 . kustaanheimo , p. & stiefel , e.l . , 1965 , j. reine angew . , 218 , 204 . lachaume , r. , dominik , c. , lanz , t. , & habing , h.j . , 1999 , a&a , 348 , 897 . lada , c.j . & lada , e.a . , 1991 , in : the formation and evolution of star clusters , ed . k. janes , asp conference series , p. 3 . lada , c.j . & lada , e.a . , 2003 , ara&a , 41 , 57 . larsen , s.s . , & richtler , t. , 2000 , a&a , 354 , 836 . lata , s. , pandey , a.k . , sagar , r. , & mohan , v. , 2002 , a&a , 388 , 158 . lin , c.c . & shu , f.h . , 1964 , apj , 140 , 646 . lin , d.n.c . , jones , b.f . , & klemola , a.r . , 1995 , apj , 439 , 652 . lynden - bell , d. & lynden - bell , r.m . , 1995 , mnras , 275 , 429 . lyng , g. , 1985 , in : the milky way galaxy , iau symp . 106 , eds . h. van woerden , r.j . allen , & w.b . burton ( dordrecht : reidel ) , p. 243 . lyng , g. , 1987 , catalogue of open cluster data , 5th ed . donnes stellaire , strabourg ) . ( available at http://cdsweb.u-strasbg.fr ) majewski , s.r . , 1993 , ara&a , 31 , 575 . majewski , s.r . , skrutskie , m.f . , weinberg , m.d . , ostheimer , j.c . , 2004 , apjl , ( in press ) . marsakov , v.a . , shevelev , y.g . , & suchkov , a.a . , 1990 , ap&ss , 172 , 51 . martin , n.f . , ibata , r.a . , bellazzini , m. , irwin , m.j . , lewis , g.f . , & dehnen , w. , 2004 , mnras , 348 , 12 . martos , m. , hernndez , x. , yez , m. , moreno , e. , & pichardo , b. , 2004 , mnras , ( submitted ) . mermilliod , j .- c . , 1988 , inform . cds , 35 , 77 . mermilliod , j .- c . , 1992a , in : astronomy from large databases ii , eds . a. heck & f. murtagh , eso conf . and work . proc . , 43 , 373 . mermilliod , j .- c . , 1992b , inform . cds , 40 , 115 . mermilliod , j .- c . , 1993 , in : databases for galactic structure , eds . philip , b. hauck , & a.r . upgren ( l. davis press : schenectady ) , p. 27 . mermilliod , j .- c . , 1995 , in : information and on - line data in astronomy , eds . d. egret & m.a . albrecht , ( dordrecht : kluwer ) , p. 127 . mermilliod , j .- c . , 1996 , in : the origins , evolutions , and destinies of binary stars in clusters , eds . milone & j .- c . mermilliod ( asp conference series ) , p. 475 . mermilliod , j .- c . , 2003 , open cluster database , http://www.unige.ch/webda . meusinger , h. , 1991 , ap&ss , 182 , 19 . meyer , m.r . , adams , f.c . , hillenbrand , l.a . , carpenter , j.m . , & larson , r.b . , 2000 , in : protostars and planets iv , eds . v. mannings , a.p . boss , & s.s . russell ( university of arizona press ) , p. 121 . micela , g. , sciortino , s. , & favata , f. , 1993 , apj , 412 , 618 . mikkola , s. , 1985 , mnras , 215 , 171 . mikkola , s. & aarseth , s.j . , 1993 , cemda , 57 , 439 . miller , g.e . & scalo , j.m . , 1979 , apjs , 41 , 513 . murai , t. & fujimoto , m. , 1980 , pasj , 32 , 581 . navarro , j.f . , helmi , a. , & freeman , k.c . , 2004 , apjl , ( in press ) . newberg , n.j . , et al . , 2002 , apj , 569 , 245 . & scalo , j. , 1990 , apj , 352 , 605 . pandey , a.k . , bhatt , b.c . , mahra , h.s . , & sagar , r. , 1989 , mnras , 236 , 263 . piskunov , a.e . , belikov , a.n . , kharchenko , n.v . , sagar , r. , & subramaniam , a. , 2004 , mnras , ( in press ) . popper , d.m . , 1980 , ara&a , 18 , 115 . prisinzano , l. , micela , g. , sciortino , s. , & favata , f. , 2004 , a&a , ( in press ) . reddish , v.c . , 1978 , stellar formation ( pergamon press ) . roberts , m.s . , 1957 , pasp , 69 , 59 . rocha - pinto , h.j . & maciel , w.j . , 1997 mnras , 289 , 882 . rocha - pinto , h.j . , scalo , j. , maciel , w.j . , & flynn , c. , 2000a , apj , 531 , l115 . rocha - pinto , h.j . , scalo , j. , maciel , w.j . , & flynn , c. , 2000b , a&a , 358 , 869 . rocha - pinto , h.j . , majewski , s.r . , skrutskie , m.f . , & crane , j.d . , 2003 , apj , 594 , 115 . sagar , r. , joshi , u.c . , & sinvhal , s.d . , 1983 , india , 11 , 44 . salaris , m. , weiss , a. , & percival , s.m . , 2004 , a&a , 414 , 163 . salpeter , e.e . , 1955 , apj , 121 , 161 . scalo , j.m . , 1986 , cosmic phys . , 11 , 1 . scalo , j.m . , 1987 , in : starbursts and galaxy evolution , eds . thuan , t. montmerle , & tran thanh van ( editions frontires : gif sur yvette ) , p. 445 . schaller , g. , schaerer , d. , meynet , g. , & maeder , a. , 1992 , a&as , 96 , 269 . sellwood , j.a . & binney , j.j . , 2002 , mnras , 336 , 785 . de simone , r.s . , wu , x. , & tremaine , s. , 2004 , mnras , ( in press ) . soderblom , d.r . , duncan , d.k . , & johnson , d.r . , 1991 , apj , 375 , 722 . spassova , n.m . & beav , p.v . , 1985 , ap&ss , 112 , 111 . taff , l.g . , 1974 , aj , 79 , 11 . terlevich , e. , 1983 , ph.d . thesis , university of cambridge . terlevich , e. , 1985 , in : dynamics of star clusters , eds . j. goodman & p. hut , d. reidel publ . , p. 471 . terlevich , e. , 1987 , mnras , 224 , 193 . twarog , b.a . , 1980 , apj , 242 , 242 . vzquez , r.a . & feinstein , a. , 1989 , rev . mexicana astron . astrof . , 17 , 3 . whitmore , b.c . , zhang , q. , leitherer , c. , fall , s.m . , schweizer , f. , & miller , b.w . , 1999 , aj , 118 , 1551 . wielen , r. , 1977 , a&a , 60 , 263 . wielen , r. , fuchs , b. , & dettbarn , c. , 1996 , a&a , 314 , 438 . yanny , b. , et al . , 2003 , apj , 588 , 824 . zaritsky , d. & harris , j. , 2004 , aj , ( in press ) . zombeck , m.v . , 1990 , handbook of astronomy and astrophysics , 2nd ed . , cambridge university press , cambridge .	we have derived the star formation history of the milky way disk over the last 2 gyr with a time resolution of 0.05 gyr from the age distribution diagram of a large sample of open clusters comprising more than 580 objects . by interpreting the age distribution diagram using numerical results from an extensive library of @xmath0-body calculations carried out during the last ten years , we reconstruct the recent star formation history of the milky way disk . under the assumption that the disk has never been polluted by any extragalactic stellar populations , our analysis suggests that superimposed on a relatively small level of constant star formation activity mainly in small-@xmath0 star clusters , the star formation rate has experienced at least 5 episodes of enhanced star formation lasting about 0.2 gyr with production of larger clusters . this cyclic behavior shows a period of [email protected] gyr and could be the result of density waves and/or interactions with satellite galaxies . on the other hand , the star formation rate history from a volume - limited sample of open clusters in the solar neighbourhood appears to be consistent with the overall star formation history obtained from the entire sample . pure continuous star formation both in the solar neighbourhood and the entire galactic disk is strongly ruled out . our results also indicate that , in the milky way disk , about 90% of open clusters are born with @xmath2 and the slope in the power - law frequency distribution of their masses is about -2.7 when quiescent star formation takes place . if the above results are re - interpreted taking into consideration accretion events onto the milky way , it is found that a fraction of the unusually high number of open clusters with ages older than 0.6 gyr may have been formed in disrupted satellites . problems arising from the selection effects and the age errors in the open cluster sample used are discussed in detail . galaxy : disk - galaxy : evolution - galaxy : open clusters and associations : general - solar neighbourhood - star formation
curvature - related level set models have been proposed in many problems of image and surface processing , due to their well understood analytical features and nice smoothing properties . in this paper , we consider one of such models , related to the problem of reconstructing a surface from a discrete set of points on it . for @xmath0 , let @xmath1 be a discrete set of points in @xmath2 , which are to be understood as points on a surface ( or a curve if @xmath3 ) , from which the surface itself has to be reconstructed . the set @xmath4 will be also termed as `` data set '' in what follows , and we assume that its points might be affected by noise . thus , the reconstructed surface is not expected to pass through each of the data set points , but rather ( depending on the criterion used to perform the reconstruction ) to provide some trade - off between an exact interpolation and a smooth behaviour . a level set model for this problem has been proposed by zhao et al . in @xcite , and leads to the following evolutive problem @xmath5 where @xmath6 is the euclidean distance from the set @xmath4 , and @xmath7 denotes the gradient operator . as customary in level set methods , the reconstructed surface at a given time @xmath8 is represented as the zero - level set of the solution @xmath9 , that is , @xmath10 in fact , is related to an energy functional in which the @xmath11 norm of the distance from @xmath4 is integrated on the whole surface ( see @xcite ) . more precisely , given a surface @xmath12 , we define the energy @xmath13 as the surface integral @xmath14 and look for a minimum of this functional ( this has the clear meaning of a compromise between the total surface of @xmath12 and its distance from the data set @xmath4 ) . once we define the evolution of an initial guess @xmath15 for the surface along the gradient flow of @xmath13 , and express the surface at time @xmath8 as the zero - level set of a function @xmath9 as in , we obtain . details about the derivation of this model may be found in @xcite , whereas well - posedness of can be proved in the framework of viscosity solutions ( see @xcite ) , which requires minimal regularity assumptions on the solution . here , we are interested in the stationary version of , that is , @xmath16 which plays the role of an euler lagrange equation for the energy @xmath13 and should be satisfied at local minima of the energy . in particular , a solution of will be obtained in what follows as a regime solution of for @xmath17 . we quote that similar techniques , retaining the regularizing effect of curvature - like terms , but possibly based on different evolution operators , are proposed for example in @xcite . of course , level set methods are not the only strategy used to solve the problem of surface reconstruction for example , segmentation techniques have been successfully proposed for this problem in recent years ( see @xcite ) . on the other hand , the largest amount of literature on the topic is probably devoted to least squares techniques . while we give up a complete review of this line of research , we mention that a relatively recent and successful technique in this framework makes use of radial basis functions ( rbf ) space reconstructions , whose application to surface reconstruction stems from pioneering works published in the late 90s @xcite ( see also @xcite for a general review ) . the aim of the present work is to investigate the application of semi - lagrangian ( sl ) numerical techniques to , focusing in particular on their implementation with rbf space reconstructions . the application of rbf techniques to sl schemes has gained a certain popularity , although it has been restricted so far to the case of hyperbolic problems ( see , e.g. , @xcite and the literature therein ) . on the other hand , semi - lagrangian schemes for curvature - related equations have been first proposed in @xcite and has gone through a number of improvements and applications ( see , in particular , @xcite for an in - depth convergence analysis and @xcite for two applications to image processing , along with @xcite for a general review on sl schemes ) . in this framework , rbf would be expected to provide a more flexible tool to construct a sparse space reconstruction . in fact , since their construction is not based on a space grid , radial basis functions allow in principle for a sparse implementation , as well as for local refinements , although we are not aware of general strategies which could effectively handle rbfs in very disordered geometries . we propose therefore a _ structured , but local _ rbf space reconstruction in which one could better focus on the region close to the data set , instead of working on a full ( and computationally more expensive ) grid . this will be the final goal of the paper . we mention that , despite being constructed by different numerical tools , our scheme implements a `` localization '' of the numerical effort , much in the same spirit of the multigrid / multilevel techniques shown in @xcite . we finally remark that , while a model like might not be rated as the cutting - edge technique for surface reconstruction , yet the coupling of sl schemes with localized rbf space reconstructions shows a good potential in terms of accuracy and computational cost , and seems to be valuably applicable to a wider class of problems ( in particular , level set models as the one under consideration ) . the outline of the paper is as follows . section [ schema ] will review the basic principles of construction of a sl scheme for , as well as the underlying ideas of rbf interpolation . section [ test ] will show numerical tests of increasing difficulty in two and three space dimensions , with both full and reduced grids . we sketch in this section the construction of a sl approximation to . we start by sketching the basic ideas on the two dimensional case , then give the three - dimensional version of the scheme , and last describe the main improvements and modifications for the case of rbf space interpolations . the main feature of the sl scheme under consideration is to be explicit , yet not constrained by the classic `` parabolic cfl '' condition , typical in the explicit treatment of diffusion terms . the degenerate diffusion performed by the curvature operator is treated by means of a convex combination of ( interpolated ) values of the numerical solution at the previous step , as we will soon show . before introducing the scheme , we rewrite the mean curvature operator in such a way that the derivation of the method will be more natural . let us recall that @xmath18 where @xmath19 is the identity matrix , and @xmath20 is a matrix which projects the diffusion on the tangent plane of each level surface . the projection @xmath21 is a matrix of rank @xmath22 with @xmath22 eigenvectors corresponding to the eigenvalue @xmath23 , and can be written as @xmath24 for a @xmath25 matrix @xmath26 having these eigenvectors as columns . in the case @xmath3 , there is only one eigenvector orthogonal to the gradient , namely @xmath27 in this case , the operator can be rewritten as @xmath28 , the projection matrix @xmath29 is a @xmath30 matrix of rank 2 spanning the two - dimensional space orthogonal to the gradient of @xmath31 . the two orthonormal eigenvectors of @xmath29 are : @xmath32 and , once set @xmath33 , the mean curvature operator can be rewritten as @xmath34 we will derive the scheme from this form , which corresponds to the probabilistic interpretation described in @xcite . in the two - dimensional case , the scheme has the structure @xmath35\left(x_j+\d t\ > d d(x_j)+\sqrt{2 \d t\ > d(x_j ) } \>\sigma^n_j\right)+ \\ \displaystyle \hspace{1.3 cm } + \frac{1}{2}i[u^n]\left(x_j+\d t\ > d d(x_j)-\sqrt{2 \d t\ > d(x_j)}\>\sigma^n_j\right ) & x_j \in { { \mathcal{g } } } , n\ge 0 \\ u^0_j = u^0(x_j ) & x_j \in { { \mathcal{g}}}\end{cases}\ ] ] where @xmath36(x)$ ] denotes a numerical interpolation of the solution @xmath37 computed at the point @xmath38 , @xmath39 denotes the set of grid nodes , and @xmath40 is an approximation of the unit normal to @xmath41 defined by @xmath42 for some consistent approximations @xmath43 , @xmath44 ( e.g. , centered partial incremental ratios of the numerical solution ) for the partial derivatives @xmath45 , @xmath46 , with @xmath47 denoting the euclidean norm of the approximate gradient at @xmath48 , i.e. , @xmath49 note that , at each node , performs an average of the solution at the two points @xmath50 . the term @xmath51 clearly represents an upwinding along the advection field @xmath52 , as usual in sl schemes . on the other hand , the degenerate operator related to curvature is treated by adding a further displacement @xmath53 , which has the effect of generating a diffusion along the tangent space of the level sets , according to the curvature operator . more explicitly , can be rewritten via minor algebraic manipulations as @xmath54(x_j+\d t\ > d d(x_j)+h_j ) - 2 i[u^n](x_j+\d t\ > d d(x_j ) ) + \nonumber \\ & & + i[u^n](x_j+\d t\ > d d(x_j)-h_j)\big ) + \frac{1}{\d t } \big(i[u^n](x_j+\d t\ > d d(x_j ) ) - u^n_j\big)\end{aligned}\ ] ] where @xmath55 is a finite increment along the tangent space ( approximately ) spanned by @xmath40 . in the right - hand side of , it is easy to recognize that the two terms are consistent with respectively the second and the first directional derivatives which appear in . a more detailed consistency analysis for the pure curvature case is carried out in @xcite . singular points at which @xmath56 ( where the definition would not make sense ) are to be treated by suitable techniques . a typical choice ( see @xcite ) is to replace the degenerate diffusion with an isotropic one ( e.g. , the heat operator ) whenever the approximate gradient is below a given threshold , say , @xmath57 , for a proper choice of the constant @xmath58 and the exponent @xmath59 . more explicitly , in a form parallel to , by computing the diffusion term as a 5-point laplacian we obtain @xmath60(x_j+\d t\ > d d(x_j)+\|h_j\|e_1 ) + i[u^n](x_j+\d t\ > d d(x_j)-\|h_j\|e_1 ) + \nonumber \\ & & + i[u^n](x_j+\d t\ > d d(x_j)+\|h_j\|e_2 ) + i[u^n](x_j+\d t\ > d d(x_j)-\|h_j\|e_2 ) - \nonumber \\ & & -4 i[u^n](x_j+\d t\ > d d(x_j ) ) \big ) + \frac{1}{\d t } \big(i[u^n](x_j+\d t\ > d d(x_j ) ) - u^n_j\big)\end{aligned}\ ] ] ( where @xmath61 and @xmath62 are the base vectors of @xmath63 ) , which results in a convex combination of the values @xmath36(x_j+\d t\ > d d(x_j ) \pm \|h_j\|e_i)$ ] and is therefore explicit and stable . this technique of treating singularities can be shown to be compatible with the definition of viscosity solution for ( see @xcite ) . following @xcite , we can also write a three - dimensional version of the scheme ( which will be used in the section on numerical tests to recover surfaces ) . the new form of @xmath64 will then be written as @xmath65 where @xmath66 and @xmath67 . using this matrix to perform an average of 4 points on the tangent plane , we write the scheme as @xmath68\left(x_j+\d t\ > d d(x_j)+\sqrt{2 \d t\ > d(x_j ) } \>\sigma^n_j\delta_i\right)\ ] ] in which the vectors @xmath69 @xmath70 are defined as @xmath71 for all combinations of the signs . via minor algebraic manipulations , can be rewritten as @xmath72(x_j+\d t\ > d d(x_j ) ) - u^n_j\big)+\\ & & + \frac{d(x_j)}{\|h^1_j\|^2}\big(i[u^n](x_j+\d t\ > d d(x_j)+h^1_j ) - 2 i[u^n](x_j+\d t\ > d d(x_j ) ) + i[u^n](x_j+\d t\ > d d(x_j)-h^1_j)\big ) + \nonumber \\ & & + \frac{d(x_j)}{\|h^2_j\|^2 } \big(i[u^n](x_j+\d t\ > d d(x_j)+h^2_j ) - 2 i[u^n](x_j+\d t\ > d d(x_j ) ) + i[u^n](x_j+\d t\ > d d(x_j)-h^2_j))\big ) . \nonumber \end{aligned}\ ] ] where @xmath73 @xmath74 in the second and third line of , it is possible to recognize the second finite differences along the directions @xmath75 and @xmath76 , which has the effect of generating a diffusion along the tangent space of the level sets , in agreement with the curvature operator . in the first line of , an upwind approximation of the transport term appears . in ( approximately ) singular conditions , i.e. , when @xmath77 , the diffusion term is switched to a 7-point laplacian by analogy with . while the sl scheme has proved to be robust and relatively accurate in a variety of applications , we study in this work an adaptation to this specific case . in , the interest is in following the zero - level set of the solution , which is in turn supposed to stay in the neighbourhood of the data set @xmath4 . the computational effort has therefore to be concentrated in this latter region computing an accurate solution away from the data set is useless . keeping this idea in mind , we implement with a space reconstruction in the form of a radial basis function ( rbf ) interpolation , which lends itself to a sparse implementation ( see @xcite ) . in particular , we have used here the matlab rbf interpolation toolbox described in @xcite . the general structure of the space reconstruction under consideration is @xmath78(x ) = c_0(u ) + c(u)\cdot x + \sum_i \lambda_i(u ) \phi_\rho(\|x - x_i\|)\ ] ] where the scalar @xmath79 , the vector @xmath80 and the coefficients @xmath81 ( all of which depend on @xmath31 ) are determined by imposing interpolation conditions , and a suitable closure of the system ( see @xcite ) . note that in this case , since @xmath82 $ ] has a globally defined form , the computation of the @xmath64 in might be carried out explicitly . in , the function @xmath83 provides the radial term . although ( as for the linear rbf defined by @xmath84 ) this term might not depend on further parameters , it is usually expected to depend on a shape constant @xmath85 ( e.g. , the `` variance '' in a gaussian function ) . whenever @xmath85 appears in the definition of @xmath86 , it is generally recognized that the choice of this parameter has a critical effect on the accuracy . moreover , optimal strategies for selecting @xmath85 exist for a regular distribution of the nodes @xmath87 , but not if the set @xmath39 of nodes is genuinely scattered . on the other hand , a structured , full - grid implementation of rbf space reconstruction gives essentially no advantage with respect to a more conventional interpolation . for all these reasons , we will implement the following strategy : * in order to have a clear guideline for choosing the scale parameter @xmath85 , nodes are distributed over a _ reduced , but structured set _ , i.e. , we start from a structured orthogonal grid and keep only the nodes belonging to some given neigbourhood of the data set . for example , we can use nodes @xmath88 for which @xmath89 for some suitable threshold @xmath90 . the nodes of the data set are also used as rbf nodes . as an example , [ test1dgrid ] shows data set points as circles in red , and rbf nodes as asterisks in blue . note that taking @xmath91 results in creating a reduced grid containing about @xmath92 layers of nodes both inside and outside @xmath4 . * it is known that rbf are particularly unsuitable to extrapolate a function at points far from the region covered by the nodes . in order to keep the nodes concentrated near the data set , still avoiding instabilities when far from this region , we introduce a set @xmath93 of _ anchor nodes _ in which the numerical solution has a prescribed value , possibly both inside and outside the curve / surface to be reconstructed . anchor nodes essentially work as boundary conditions , by locking the value of the numerical solution at `` far '' points where the computation of rbf interpolation can not be ( and in fact is not required to be ) accurate . + referring again to fig . [ test1dgrid ] , anchor nodes appear as the outer crosses in black . in our numerical tests , the anchor nodes have been placed on a frame surrounding the data set , following the geometry of the edges of the computational domain . in any of the tests , no anchor node has been used inside the data set . we will show in the section on numerical examples that this framework results in a considerable reduction of the computational grid , still preserving robustness of the algorithm . [ [ initial - condition ] ] initial condition + + + + + + + + + + + + + + + + + in order to speed up convergence , the initial condition should have a zero - level set @xmath94 _ as close as possible to the surface ( or curve ) to be reconstructed _ , although , of course , the surface itself is unknown . moreover , a natural choice for the initial condition @xmath95 would be the signed distance from the initial interface @xmath15 , that is , @xmath96 in practice , since we look for an asymptotic solution for , it turns out that neither the choice of @xmath15 nor the structure of @xmath97 is critical clearly , a better initial conditions results in a faster solution . in lack of a sharper information , a relatively conservative choice of the initial condition might be set as @xmath98 with a suitable choice of the constants @xmath99 in order to start from a zero - level set enclosing the data set , but not too far from it . this choice , which will be used in the numerical examples of the next section , avoids the cost of computing the signed distance , but maintains negative values for @xmath31 inside the surface @xmath100 , and positive values outside . we did not experience any critical behaviour of the scheme with respect to this choice , nor the need for a re - initialization of the level set function . in fact , the effect of the advection term in is to push the solution towards the data set , and this causes the solution to maintain a steep transition in the neighbourhood of the interface . we present in this section some numerical examples of shape reconstructions with the algorithm described above . in the two - dimensional case , we will also compare the reduced grid with the full grid scheme . note that , in two dimensions , it is known ( see @xcite ) that a polygonal line joining the points of the data set is a local minimizer of the energy , so this should be regarded as an `` exact solution '' for the 2d case . in all the test , we compute the normalized @xmath11-norm of the update between two successive iterates : @xmath101 this being an indication on the speed of convergence to the regime solution . note that , when plotted on a semi - log scale ( as it will be done in what follows ) , a linear decrease of @xmath102 suggests a contraction - like behaviour of the scheme . in the following examples , we will keep the data set in a quite sparse form . we found it discriminating to work with a relatively low number of points , and this will be our strategy in order to check the robustness of the numerical algorithm . we first consider a data set @xmath4 made of 24 points uniformly chosen on a heart - shaped curve and a grid @xmath39 of @xmath103 evenly spaced nodes on a domain @xmath104 ^ 2 $ ] , as shown in the left plot of fig . [ test1bgrid ] . we apply on the grid @xmath39 with time step @xmath105 and show ( in the right plot of fig . [ test1bgrid ] ) the trend of the normalized update between two successive iterations , the algorithm being stopped after 150 iterations . the final solution obtained by multiquadric rbfs ( with a scale factor @xmath106 , see @xcite ) is shown in the left plot of fig . [ test1bsol ] . the final solution obtained by linear rbfs and by linear rbfs applied on a finer @xmath107 grid are shown respectively in the center and in the right plot of fig . [ test1bsol ] . the reconstructed curve is very close to the theoretical forecast of a polygonal line , except for a slight smoothing of the non - convex , upper section of the shape ( see fig . [ test1bsol ] ) . regardless of the number of points in the data set , this gives the natural indication that the computational grid is still too coarse . note also that convergence of iterates tends to become quite slow , although it seems not to depend on the type of rbf used . in fact , we have included a single plot for both linear and multiquadric rbfs , whose convergence histories show an analogous behaviour . the same test is performed again on a reduced grid @xmath39 made of 106 point distributed around the data set according to with @xmath108 , as shown in the left plot of fig . [ test1dgrid ] , which also shows the anchor points ( marked as the outer black crosses ) together with the initial condition . the smaller area covered by the grid allows to decrease the total number of nodes to a fraction of about 12% of the original grid . the final solution obtained by multiquadric rbfs with @xmath106 and by linear rbfs are respectively shown in the left and right plots of fig . [ test1dsol ] . the result is essentially equivalent to what is obtained with the full grid ( fig . [ test1dsol ] ) . note that , comparing the right plots of fig . [ test1bgrid ] and [ test1dgrid ] , convergence of the iterative solver seems also to be comparable in terms of iteration number , but with a lower cost for a single iteration . as in the case of a full grid , convergence of the iterates for linear and multiquadric rbfs is very similar . we consider a data set @xmath4 made of 748 points chosen on a 3d heart shape in the set @xmath104 ^ 3 $ ] . we build a grid @xmath39 made by the dataset plus @xmath109 additional nodes uniformly distributed on a narrow band around the data set , computed according to with @xmath110 from a full grid of @xmath111 nodes . + we apply on the grid @xmath39 with time step @xmath112 . the right plot of fig . [ test2bsol ] shows the norm of the update between two successive iterations . the algorithm is stopped after 80 iteration , and the final solution is shown in the left plot of fig . [ test2bsol ] . dataset , grid and zero - level set of the initial condition are shown in fig . [ test2bgrid ] . we also analyze in this example the robustness of the method with respect to noisy data . to this end , we perturb the data set points by adding random displacements , uniformly distributed in @xmath113 ^ 3 $ ] . the recovered surfaces obtained by applying to three different levels of noisy datasets ( @xmath114 ) , still on the same space grid @xmath39 , with time step @xmath112 and 80 iterations , are shown in fig . [ test2bnoise ] . as a third test , we consider a data set @xmath4 made of 4020 points uniformly chosen on a shape made by two intersecting cubes in @xmath104 ^ 3 $ ] , with axes not aligned with the grid . we consider a grid @xmath39 containing the data set and @xmath115 additional nodes uniformly chosen in a narrow band , computed according to with @xmath116 from a full grid of @xmath117 nodes . we apply on the grid @xmath39 with time step @xmath105 . the algorithm is stopped after 100 iteration , and the final solution is shown in the left plot of fig . [ cubisol ] . data set , grid points and initial condition are shown in fig . [ cubigrid ] . finally , we test our algorithm on the more complex benchmark of a teapot - shaped surface ( whose data set has been taken from http://www.itl.nist.gov/iad/vug/sharp/benchmark/3dinterestpoint/ ) in the cube @xmath118 ^ 3 $ ] . we consider a data set @xmath4 made of 2602 points ( about @xmath119 of the original dataset ) and a grid @xmath39 containing the dataset and @xmath120 additional nodes chosen on a narrow band of the data set , computed with @xmath110 from a full grid of @xmath121 nodes . we apply on the grid @xmath39 with time step @xmath122 . the algorithm is stopped after 150 iterations , showing the final solution and the convergence history in the two plots of fig . [ test7sol ] , while the data set and the grid are shown in fig . [ test7grid ] . carr , r.k . beatson , j.b . cherrie , t.j . mitchell , w.r . fright , b.c . mccallum and t.r . evans,_reconstruction and representation of 3d objects with radial basis functions _ , proceedings of acm siggraph 2001 , 6776 , 2001 e. franchini , s. morigi , f. sgallari , _ implicit shape reconstruction of unorganized points using pde - based deformable 3d manifolds _ , numerical mathematics : theory , methods and applications , * 3 * ( 2010 ) , 405430 j. ye , i. yanowsky , b. dong , r. gandlin , a. brandt , s. osher , _ multigrid narrow band surface reconstruction via level set functions _ , proceedings of the 8th international symposium on visual computing isvc 2012 ( rethymnon , crete ) , 6170 , springer , 2012 h. zhao , s. osher , b. merriman and m. kang , _ implicit and non - parametric shape reconstruction from unorganized points using variational level set method _ , computer vision and image understanding , * 80 * ( 2000 ) , 295319	we propose a semi - lagrangian scheme coupled with radial basis function interpolation for approximating a curvature - related level set model , which has been proposed by zhao et al . in @xcite to reconstruct unknown surfaces from sparse data sets . the main advantages of the proposed scheme are the possibility to solve the level set method on unstructured grids , as well as to concentrate the reconstruction points in the neighbourhood of the data set , with a consequent reduction of the computational effort . moreover , the scheme is explicit . + numerical tests show the accuracy and robustness of our approach to reconstruct curves and surfaces from relatively sparse data sets .
the following pages contain the fusion of two published papers : _ what can we learn about gravitational wave physics with an elastic spherical antenna ? _ , published in the _ physical review _ ( prd * 52 * , 591 - 604 ( 1995 ) ) , and _ multiple mode gravitational wave detection with a spherical antenna _ , scheduled to appear in the july-2000 issue of _ monthly notices of the royal astronomical society _ ( mnras * 316 * , 173 - 194 ( 2000 ) ) , i first developed an analytic model to describe the interaction of a solid elastic sphere with an incoming flux of gravitational waves of the general _ metric _ class , then extended it to also address the multimode signal deconvolution problem by means of suitable layouts of _ resonant _ motion sensors . i think the model is quite complete as regards the major _ conceptual _ issues in a spherical gw detector , therefore think merging the two papers together makes positive sense . let me however stress that i do not mean to understate the fundamental problem of _ noise _ , which is here left aside ; rather , it is now being actively investigated in detail within the very convenient general framework set up below . _ what can we learn about gw physics _ is already five years old , and certain parts of it have been the subject of further research since . more specifically , this happens with brans - dicke absorption cross sections , see section 4 , where reference has been added to such newer work . in addition , two new tables and more mathematical detail have been added to its appendix b. the latter should help the interested reader with certain technicalities , while the tables and attendant new formulas are meant to provide explicit numerical values of the frequency spectrum of the sphere rather than the original _ only graphics _ information . i expect this to be useful reference in numerical and/or experimental determinations of these quantities . no changes have been included in the second paper relative to its mnras version . in merging the two articles together , however , i have found expedient to renumber the equations and bibliographic references into corresponding single streams . the bibliography has been updated and made into a unique list at the end of the file in alphabetical order . finally , i have included minor notation changes in order to reconcile otherwise small mistunings between both articles . alberto lobo + barcelona , june 2000 what can we learn about gw physics with an elastic spherical antenna ? + prd * 52 * , 591 - 604 ( 1995 ) + * jos alberto lobo * + departament de fsica fonamental + universitat de barcelona , spain + e - mail : [email protected] spherical antennae are considered by many to be the natural next step in the development of resonant gw detectors . the reasons for this new trend essentially derive from the _ improved sensitivity _ of a sphere which can be nearly an order of magnitude better than a cylinder having the same resonance frequency , see below and @xcite , and from its _ multimode capabilities _ , first recognised by forward @xcite and further elaborated in @xcite . although some of the most relevant aspects of detector sensitivity have already received attention in the literature , it seems to me that a sufficiently general and flexible analysis of the interaction between gw and detector has not been satisfactorily developed to date . this _ theoretical _ shortage has a number of _ practical _ negative consequences , too . traditional analysis , to mention but an example , is almost invariably restricted to general relativity or scalar tensor theories of gravity ; while it may be argued that this is already very general , any such argument is , as a matter of fact , understating the potentialities actually offered by a spherical gw antenna to help decide for or against any one specific theory of the gravitational field on the basis of _ experimental observation_. i thus propose to develop in this paper a full fledged mathematical formalism which will enable analysis of the antenna s response to a completely general gw , i.e. , making no _ a priori _ assumptions about which is the correct theory underlying gw physics ( other than , indeed , that it is a _ theory ) , _ and _ also making no assumptions about detector shape , structure or boundary conditions . considering things in such generality is not only `` theoretically nice '' it also brings about new results and a _ better understanding _ of older ones . for example , it will be proved that the sphere is the _ most efficient _ gw elastic detector shape , and that higher mode absorption cross sections _ scale independently of gw physics_. i will also discuss the direction of incidence deconvolution problem in the context of a general metric theory of gravity . the paper is organised as follows : section 2 is devoted to the development of the general mathematical framework , leading to a formula in which an elastic solid s response is related to the action of an arbitrary metric gw impinging on it . in section 3 the general equations are applied to the homogeneous spherical body , and a discussion of the _ deconvolution _ problem is presented as well . section 4 contains the description of the sphere s sensitivity parameters , specifically leading to the concept of _ multimode _ , or vector , _ transfer function _ , and to an analysis of the _ absorption cross section _ presented by this detector to a passing by gw . conclusions and prospects are summarised in section 5 , and two appendices are added which include mathematical derivations . in the mathematical model , i shall be assuming that the antenna is a solid elastic body which responds to gw perturbations according to the equations of classical non - relativistic linear elasticity theory @xcite . this is fully justified because gw - induced displacements will be very small indeed , and the speed of such displacements much smaller than that of light for any foreseeable frequencies . although our primary interest is a spherical antenna , the considerations which follow in the remainder of this section _ have general validity for arbitrarily shaped , isotropic elastic solids_. let @xmath2 be the displacement vector of the infinitesimal mass element sitting at point * x * relative to the solid s centre of mass in its unperturbed state , whose density distribution in that state is @xmath3 . let @xmath4 and @xmath5 be the material s elastic lam coefficients . if a volume force density @xmath6 acts on such solid , the displacement field @xmath2 is the solution to the system of partial differential equations @xcite @xmath7 with the appropriate initial and boundary conditions . a summary of notation and general results regarding the solution to that system is briefly outlined in the ensuing subsection , as they are necessary for the subsequent developments in this paper , and also in future work e.g . page and ss . below . for reasons which will become clear later on , we shall only be interested in driving forces of the separable type @xmath8 or , indeed , linear combinations thereof . the solution to ( [ 2.1 ] ) does not require us to specify the precise boundary conditions on @xmath2 at this stage , but we need to set the initial conditions . we adopt the following : @xmath9 where @xmath10 , implying that the antenna is at complete rest before observation begins at @xmath11=0 . the structure of the force field ( [ 2.2 ] ) is such that the displacements @xmath2 can be expressed by means of a _ green function _ integral of the form @xmath12 the deductive procedure whereby @xmath13 is calculated can be found in many standard textbooks see e.g. @xcite . the result is @xmath14 \sum_n\,\mbox{\large $ \frac{f_n}{\omega_n}$}\ , { \bf u}_n({\bf x})\sin\omega_nt & \ \ { \rm if}\ \ t\geq 0 \end{array}\right . \label{2.15}\ ] ] where @xmath15 and @xmath16 are the normalised _ eigen solutions _ to @xmath17 with suitable _ boundary conditions_. here @xmath18 represents an index , or set of indices , labelling the _ eigenmode _ of frequency @xmath19 . the normalisation condition is ( arbitrarily ) chosen so that @xmath20 where @xmath21 is the total mass of the solid , and the asterisk denotes complex conjugation . replacing now ( [ 2.15 ] ) into ( [ 2.6 ] ) we can write the solution to our problem as a series expansion : @xmath22 where @xmath23 equation ( [ 2.16 ] ) is the _ formal _ solution to our problem ; it has the standard form of an orthogonal expansion and is valid for _ any _ solid driven by a separable force like ( [ 2.2 ] ) and _ any _ boundary conditions . it is therefore _ completely general _ , given that type of force . before we go on , it is perhaps interesting to quote a simple but useful example . it is the case of a solid hit by a _ hammer blow _ , i.e. , receiving a sudden stroke at a point on its surface . exam of the response of a gw antenna to such perturbation is being used for correct tuning and monitoring of the device @xcite . if the driving force density is represented by the simple model @xmath24 where @xmath25 is the surface point hit , and @xmath26 is a constant vector , then the system s response is immediately seen to be @xmath27 with @xmath28 . a hammer blow thus excites _ all _ the solid s normal modes , except those _ perpendicular _ to @xmath26 , with amplitudes which are _ inversely proportional to the mode s frequency_. this is seen to be a rather general result in the theory of sound waves in isotropic elastic solids . an incoming gw manifests itself as a _ tidal _ force density ; in the long wavelength linear approximation @xcite it only depends on the `` electric '' components of the riemann tensor : @xmath29 where @xmath30 is the speed of light , and sum over the repeated index _ j _ is understood . in ( [ 3.1 ] ) tidal forces are referred to the antenna s centre of mass , and thus * x * is a vector originating there . note that i have omitted any dependence of @xmath0 on spatial coordinates , since it only needs to be evaluated at the solid s centre . the riemann tensor is only required to first order at this stage @xcite : @xmath31 where @xmath32 are the perturbations to flat geometry ) will run through space - time values 0,1,2,3 ; _ latin _ indices ( @xmath33 ) will run through space values 1,2,3 only . ] , always at the centre of mass of the detector . the form ( [ 3.1 ] ) is seen to be a sum of three terms like ( [ 2.2 ] ) but this three term `` straightforward '' splitting is not the most convenient , due to lack of invariance and symmetry . a better choice is now outlined . an _ arbitrary symmetric _ tensor @xmath34 admits the following decomposition : @xmath35 where @xmath36 are 5 linearly independent _ symmetric _ and _ traceless _ tensors , and @xmath37 is a multiple of the _ unit _ @xmath38 are uniquely defined functions , whose explicit form depends on the particular representation of the chosen . a convenient one is the following : & e_ij^(00 ) = ( ) ^ ( ccc 1 & 0 & 0 + 0 & 1 & 0 + 0 & 0 & 1 ) & [ 3.4a ] + & e_ij^(20 ) = ( ) ^ ( ccc -1 & 0 & 0 + 0 & -1 & 0 + 0 & 0 & 2 ) , e_ij^(21 ) = ( ) ^ ( ccc 0 & 0 & 1 + 0 & 0 & -i + 1 & -i & 0 ) & [ 3.4b ] + & e_ij^(22 ) = ( ) ^ ( ccc 1 & i & 0 + i & -1 & 0 + 0 & 0 & 0 ) & [ 3.4c ] the excellence of this representation stems from its ability to display the _ spin features _ of the driving terms in ( [ 3.1 ] ) . such features are characterised by the relationships @xmath39 where * n@xmath40x/@xmath41 is the radial unit vector , and @xmath42 are spherical harmonics @xcite . details about the above @xmath43-matrices are given in appendix a. in particular , the orthogonality relations ( [ a.9 ] ) can be used to invert ( [ 3.3 ] ) : ^(00)(t ) & = & e_ij^(00)s_ij(t ) [ 3.77a ] + s^(2m)(t ) & = & e_ij^(2m)s_ij(t ) , m=-2, ,2 [ 3.77b ] where an asterisk denotes complex conjugation . note that @xmath44 , where @xmath45 is the tensor s trace . we now take advantage of ( [ 3.3 ] ) to express the gw tidal force ( [ 3.1 ] ) as a sum of split terms like ( [ 2.2 ] ) : @xmath46 with f_i^(00)(x ) = e_ij^(00)x_j & , & g^(00)(t ) = e_ij^(00)r_0i0j(t)c^2 [ 3.8a ] + f_i^(2m)(x * ) = e_ij^(2m)x_j & , & g^(2m)(t ) = e_ij^(2m)r_0i0j(t)c^2 ( m = -2, ,2 ) [ 3.8b ] straightforward application of ( [ 2.16 ] ) yields the formal solution of the antenna response to a gw perturbation : @xmath47 \label{3.9}\ ] ] with the notation of ( [ 2.14 ] ) and ( [ 2.17 ] ) applied _ mutatis mutandi _ to the terms in ( [ 3.8 ] ) . equation ( [ 3.9 ] ) gives the response of an arbitrary elastic solid to an incoming weak gw , _ independently of the underlying gravity theory _ , be it general relativity ( gr ) or indeed any other _ metric _ theory of the gravitational interaction . it is also valid for _ any antenna shape _ and _ any boundary conditions _ , thus giving the formalism , in particular , the capability of being used to study the response of a detector which is _ suspended _ by means of a mechanical device in the laboratory site a situation of much practical importance . it is therefore _ very _ general . equation ( [ 3.9 ] ) also tells us that that only _ monopole _ and _ quadrupole detector modes _ can possibly be excited by a metric gw . the nice thing about ( [ 3.9 ] ) is that it _ fully _ displays the monopole - quadrupole _ structure _ of the solution to the fundamental differential equations ( [ 2.1 ] ) . in a non - symmetric body , all ( or _ nearly _ all ) the modes have monopole and quadrupole moments , and ( [ 3.9 ] ) precisely shows how much each of them contributes to the detector s response . a _ homogeneous spherical _ antenna , which is very symmetric , has a set of vibrational eigenmodes which are particularly well matched to the form ( [ 3.9 ] ) : it only possesses _ one _ series of monopole modes and _ one _ ( five - fold degenerate ) series of quadrupole modes see next section and appendix b for details . the existence of so _ few _ modes which couple to gws means that _ all the absorbed incoming radiation energy will be distributed amongst those few modes only _ , thereby making the sphere the _ most efficient _ detector , even from the sensitivity point of view . the higher energy cross section _ per unit mass _ reported for spheres on the basis of for example , finds here its qualitative explanation . the generality of ( [ 3.9 ] ) , on the other hand , means that _ this excellence of the spherical detector is there independently of which is the correct gw theory_. before going further , let me mention another potentially useful application of the formalism so far . cylindrical antennas , for instance , are usually studied in the _ thin rod _ approximation ; although this is generally quite satisfactory , equation ( [ 3.9 ] ) offers the possibility of eventually considering corrections to such simplifying hypothesis by use of more realistic eigenfunctions , such as those given in @xcite . recent new proposals for stumpy cylinder arrays @xcite may well benefit from the above approach , too . to explore the consequences of ( [ 3.9 ] ) in a particular case , the mode amplitudes @xmath16 and frequencies @xmath48 must be specified . from now on i will focus on a _ homogeneous sphere _ whose surface is free of tractions and/or tensions ; the latter happens to be quite a good approximation , even if the sphere is suspended in the static gravitational field @xcite . the normal modes of the free sphere fall into two families : so called _ toroidal _ where the sphere only undergoes twisting which keep its shape unchanged throughout the volume and _ spheroidal _ @xcite , where radial as well as tangential displacements take place . i use the notation @xmath49 for them , respectively ; note that the index @xmath18 of the previous section is a _ multiple _ index @xmath50 for _ each _ family ; @xmath51 and @xmath52 are the usual _ multipole _ indices , and @xmath53 numbers from 1 to @xmath54 each of the @xmath51-pole modes . the frequencies happen to be independent of @xmath55 , and so every one mode ( [ 3.10 ] ) is ( 2@xmath51 + 1)-fold degenerate . further details about these eigenmodes are given in appendix b. in order to see what ( [ 3.9 ] ) looks like in this case , integrals of the form ( [ 2.14 ] ) ought to be evaluated . it is straightforward to prove that they all vanish for the toroidal modes , the spheroidal modes contributing the only non - vanishing terms ; after some algebra one finds @xmath56 where a_n0 & = & -_0^r a_n0(r)r^3dr [ 3.12a ] + a_n2 & = & -_0^r r^3dr [ 3.12b ] the functions @xmath57 , @xmath58 are given in appendix b , and @xmath59 is the sphere s radius . to our reassurance , only the monopole and quadrupole _ sphere modes _ survive , as seen by the presence of the factors @xmath60 in ( [ 3.11 ] ) . the final series is thus a relatively simple one , even in spite of its generality , meaning _ spheroidal _ mode , to ease the notation since _ toroidal _ modes no longer appear in the formulae . ] : @xmath61\ , \qquad ( t>0 ) \label{3.16}\ ] ] where , it is recalled , @xmath62 equation ( [ 3.16 ] ) constitutes the sphere s response to an arbitrary tidal gw perturbation , and will be used to analyse the sensitivity of the spherical detector in the next section . before doing so , however , a few comments on the antenna s signal _ deconvolution capabilities _ , within the context of a completely general metric theory of gws , are in order . let us first of all take the fourier transform of ( [ 3.16 ] ) : @xmath63 this is seen to be @xmath64 \ + \nonumber \\[1 em ] & + & \!\frac{\pi}{i}\,\sum_{n=1}^\infty\,\frac{a_{n2}}{\omega_{n2 } } \left[\sum_{m=-2}^2\,{\bf u}_{n2m}({\bf x})\,g^{(2m)}(\omega)\right ] \left[\delta(\omega\!-\!\omega_{n2 } ) - \delta(\omega\!+\ ! \omega_{n2})\right ] \label{3.18}\end{aligned}\ ] ] where @xmath65 are the fourier transforms of @xmath66 , respectively : @xmath67 the @xmath68-function factors are of course idealisations corresponding to infinitely long integration times and infinitely narrow resonance line - widths but the essentials of the ensuing discussion will not be affected by those idealisations . if the measuring system were ( ideally ) sensitive to all frequencies , filters could be applied to examine the antenna s oscillations at each monopole and quadrupole frequency : a single transducer would suffice to reveal @xmath69 around the monopole frequencies @xmath70 , whilst _ five _ ( placed at suitable positions ) would be required to calculate the five degenerate amplitudes @xmath71 around the quadrupole frequencies @xmath72 . once the _ six _ functions @xmath65 would have thus been determined , inverse fourier transforms would give us the functions @xmath66 , and thereby the six riemann tensor components @xmath73 through inversion of the second of equations ( [ 3.8 ] ) , i.e. , as an expansion like ( [ 3.3 ] ) only with @xmath74 s instead of @xmath75 s . deconvolution would then be complete . well , not quite knowledge of the riemann tensor in the _ laboratory _ frame coordinates is not really sufficient to say the waveform has been completely deconvolved , unless we _ also _ know the _ source position _ in the sky . there clearly are two possibilities : 1 . the source position _ is _ known ahead of time by some other astronomical observation methods . let me rush to emphasise that , far from trivial or uninteresting , this is a _ very important _ case to consider , specially during the first stages of gw astronomy , when any reported gw event will have to be thoroughly checked by all possible means . + if the incidence direction is known , then a rotation must be applied to the just obtained quantities @xmath73 , which takes the laboratory into coincidence with the incoming wave propagation vector . a classification procedure must thereafter be applied to the so transformed riemann tensor in order to see which is the theory ( or class of theories ) compatible with the actual observations . such classification procedure has been described in detail in @xcite ; see also @xcite for an updated discussion . + the spherical antenna is thus seen to have the _ capability of furnishing the analyst sufficient information to discern amongst different competing theories of gw physics , whenever the wave incidence direction is known prior to detection_. 2 . the source position is _ not _ known at detection time . this makes things more complex , since the above rotation between the laboratory and gw frames can not be performed . + in order to deconvolve the incidence direction in this case , a specific theory of the gws _ must _ be assumed a given choice being made on the basis of whatever prior information is available or , simply , dictated by the the decision to probe a particular theory . wagoner and paik @xcite propose a method which is useful both for gr and bd theory , their idea being simple and elegant at the same time : since neither of these theories predicts the excitation of the @xmath52=@xmath761 quadrupole modes _ of the wave _ , the source position is determined precisely by the rotation angles which , when applied to the laboratory axes , cause the amplitudes of those _ antenna _ modes to vanish ; the rotated frame is thereby associated to the gw natural frame . + a generalisation of this idea can conceivably be found on the basis of a detailed and possibly rather casuistic analysis of the canonical forms of of the riemann tensor for a list of theories of gravity , along the following line of argument : any one particular theory will be characterised by certain ( homogeneous ) canonical relationships amongst the monopole and quadrupole components of the riemann tensor , @xmath66 , and so enforcement of those relations upon rotation of the laboratory frame axes should enable determination of the rotation angles or , equivalently , of the incoming radiation incidence direction . scalar - tensor theories e.g. have @xmath77 in their canonical forms , hence wagoner and paik s proposal for this particular case . + before any deconvolution procedure is triggered , however , it is very important to make sure that it will be _ viable_. more precisely , since the transformation from the laboratory to the ultimate canonical frame is going to be linear , _ invariants _ must be preserved . this means that , even if the source position is unknown , certain theories will forthrightly be _ vetoed _ by the observed @xmath73 if their predicted invariants are incompatible with the observed ones . to give but an easy example , if @xmath73 is _ observed _ to have a non - null trace @xmath78 , then a veto on gr will be readily served , and therefore no algorithm based on that theory should be applied @xcite . + i would like to make a final remark here . assume a direction deconvolution procedure has been successfully carried through to the end on the basis of certain gw theory , so that the analyst comes up with a pair of numbers @xmath79 expressing the source s coordinates in the sky . of course , these numbers will represent the _ actual _ source position _ only if the assumed theory is correct_. now , how do we know it _ is _ correct ? strictly speaking , `` correctness '' of a scientific theory is an _ asymptotic _ concept in the sense that the possibility always remains open that new facts be eventually discovered which contradict the theory , and so _ reliability _ of the estimate @xmath79 of the source position can only be assessed in practice in terms of the _ consistency _ between the assumed theory and whatever experimental evidence is available _ to date _ , including , indeed , gw measurements themselves . it is thus very important to have a method to verify that the estimate @xmath79 does not contradict the theory which enabled its very determination . + such verification is a _ logical _ absurdity if only _ one _ measurement of position is available ; this happens for instance if the recorded signal is a _ short burst _ of radiation , and so _ two antennas _ are at least necessary to check consistency in that case . the test would proceed as a check that the time delay between reception of the signal at both detectors is consistent with the calculated @xmath79 , even if the assumed theory is wrong , since the sphere deformations will be the same if caused by the same signal . ] , given their relative position and the wave propagation speed predicted by the assumed theory . if , on the other hand , the signal being tracked is a _ long duration _ signal , then a single antenna may be sufficient to perform the test by looking at the observed doppler patterns and checking them against those expected with the given @xmath79 . the above considerations have been made ignoring noise in the detector and monitor systems . a fundamental constraint introduced by noise is that it makes the antenna _ bandwidth limited _ in sensitivity . as a consequence , any deconvolution procedure is deemed to be incomplete or , rather , _ ambiguous _ @xcite , since information about the signal can possibly be retrieved only within a reduced bandwidth , whilst the rest will be lost . i thus come to a detailed discussion of the sensitivity of the spherical gw antenna in the next section . i will consider successively _ amplitude _ and _ energy _ sensitivities ; the first leads to the concept of _ transfer function _ , while the second to that of absorption _ cross section_. i devote separate subsections to analyse each of them in some detail . a widely used and useful concept in linear system theory is that of _ transfer function _ it is defined as the fourier transform of the system s impulse response , or as the system s impedance / admittance , and can be inferred from the frequency response function ( [ 3.18 ] ) . we recall from the previous section that the sphere is a multimode device due to its monopole and five - fold degenerate quadrupole modes . it is expedient to define a _ multimode _ or _ vector transfer function _ as a useful construct which encompasses all six different modes into a single conceptual block , according to @xmath80 where @xmath81 are the six driving terms @xmath65 given in ( [ 3.19 ] ) . the transfer function is @xmath82 , and its `` vector '' character alluded above is reflected by the _ multimode index _ @xmath83 . looking at ( [ 3.18 ] ) it is readily seen that ^(00)(x , ) & = & _ n=1^ * u_n00(x)[4.2a ] + z^(2m)(x * , ) & = & _ n=1^ * u_n2m(x)(m=-2, ,2 ) [ 4.2b ] as we observe in these formulae , the sphere s sensitivity to monopole excitations is governed by @xmath84 , and to quadrupole ones by @xmath85 . closed expressions happen to exist for @xmath86 and @xmath87 ; using the notation of appendix b , they are & = & [ 4.4a ] + & = & -[4.4b ] numerical investigation of the behaviour of these coefficients shows that they decay asymptotically as @xmath88 : @xmath89 likewise , it is found that the frequencies @xmath90 and @xmath91 diverge like @xmath53 for large @xmath92 , so that @xmath82 drops as @xmath93 for large @xmath94 . figures 6 and 7 display a symbolic plot of @xmath95 and @xmath96 , respectively , which illustrates the situation : monopole modes soon reach the asymptotic regime , while there appear to be 3 sub - families of quadrupole modes regularly intertwined ; the asymptotic regime for these sub - families is more irregularly reached . note also the perfectly regular alternate changes of phase ( by @xmath97 radians ) in both monopole and each quadrupole family . the sharp fall in sensitivity of a sphere for higher frequency modes ( @xmath98 ) indicates that only the lowest ones stand a chance of being observable in an actual gw antenna . i report in table i the numerical values of the relevant parameters for the first few monopole and quadrupole modes . the reason for the last ( fourth ) columns will become clear later . [ tab1 ] cccc\|cccc @xmath99 & @xmath100 & @xmath101 & @xmath102 & @xmath103 & @xmath104 & @xmath105 & @xmath106 + + & & & & 1 & 2.650 & 0.328 & 1 + & & & & 2 & 5.088 & 0.106 & 2.61 + 1 & 5.432 & 0.214 & 1 & & & & + & & & & 3 & 8.617 & @xmath1071.907@xmath108 & 27.95 + & & & & 4 & 10.917 & @xmath1079.101@xmath109 & 76.42 + 2 & 12.138 & @xmath1073.772@xmath108 & 6.46 & & & & + & & & & 5 & 12.280 & 1.387@xmath108 & 25.99 + & & & & 6 & 15.347 & 6.879@xmath109 & 67.87 + 3 & 18.492 & 1.600@xmath108 & 15.49 & & & & + let us calculate now the energy of the oscillating sphere . we first define the _ spectral energy density _ at frequency @xmath94 , which is naturally given by is the integration time assumed very large . the peaks in the @xmath68-functions diverge like @xmath110 . ] @xmath111 and can be easily evaluated : @xmath112 \label{4.8}\ ] ] the _ energy _ at any one spectral frequency @xmath113 is obtained by _ integration _ of the spectral density in a narrow interval around @xmath114 : @xmath115 in this case , @xmath116 the sensitivity parameter associated with the vibrational energy of the modes is the detector s _ absorption cross section _ , defined as the energy it absorbs per unit incident gw spectral flux density , or @xmath117 where @xmath118 is the number of joules per square metre and hz carried by the gw at frequency @xmath119 as it passes by the antenna . thus , for the frequencies of interest , _ abs(_n0 ) & = & ma_n0 ^ 2 [ 4.13a ] + _ abs(_n2 ) & = & ma_n2 ^ 2 [ 4.13b ] these quantities have very precise values , but such values can only be calculated on the basis of a _ specific underlying theory of the gw physics_. in the absence of such theory , neither @xmath118 nor @xmath65 can possibly be calculated , since they are _ not _ theory independent quantities . to date , only gr calculations have been reported in the literature @xcite . as i will now show , even though the fractions in the rhs of ( [ 4.13 ] ) are _ not _ theory independent , some very general results can still be obtained about the sphere s cross section within the context of metric theories of the gravitational interaction . to do so , it will be necessary to go into a short digression on the general nature of weak metric gws . no matter which is the ( metric ) theory which happens to be the `` correct one '' to describe gravitation , it is beyond reasonable doubt that any gws reaching the earth ought to be _ very weak_. the linear approximation should therefore be an extremely good one to describe the propagating field variables in the neighbourhood of the detector . in such circumstances , the field equations can be derived from a poincar invariant variational principle based on an action integral of the type @xmath120 where the lagrangian density @xmath121 is a _ quadratic _ functional of the field variables @xmath122 and their space - time derivatives @xmath123 ; these variables include the metric perturbations @xmath124 , plus any other fields required by the specific theory under consideration e.g . a scalar field in the theory of brans dicke , etc . the requirement that @xmath121 be quadratic ensures that the euler lagrange equations of motion are _ linear_. the energy and momentum transported by the waves can be calculated in this formalism in terms of the components @xmath125 of the canonical energy - momentum tensor ) will be sufficient . ] @xmath126 the flux energy density , or poynting , vector is given by @xmath127 , i.e. , @xmath128 where @xmath10 . any gw hitting the antenna will be seen plane , due to the enormous distance to the source . if k * is the incidence direction ( normal to the wave front ) , then the fields will depend on the variable @xmath129@xmath107k@xmath130x , so that the gw energy reaching the detector per unit time and area is @xmath131 where * x * is the sphere s centre position relative to the source which is _ fixed _ , and so its dependence can be safely dropped in the lhs of the above expression . the important thing to note in equation ( [ 4.18 ] ) is that it tells us that @xmath132 _ can be written as a quadratic form in the time derivatives of the fields @xmath133_. as a consequence , the spectral density @xmath118 , defined by @xmath134 can be ascertained to factorise as @xmath135 where @xmath136 is again a _ quadratic _ function of the fourier transforms @xmath137 of the fields @xmath133 . on the other hand , the functions @xmath65 in ( [ 4.13 ] ) which , it is recalled , are the fourier transforms of @xmath66 in ( [ 3.8 ] ) , contain _ second _ order derivatives of the _ metric _ fields @xmath32 , and therefore of _ all _ the fields @xmath133 as a result of the theory s field equations . since we are considering _ plane wave _ solutions to those equations , all derivatives can be reduced to _ time _ derivatives just like in ( [ 4.18 ] ) above . we can thus write @xmath138 with @xmath139 suitable _ linear _ combinations of the @xmath137 . replacing the last two equations into ( [ 4.13 ] ) and manipulating dimensions expediently , we come to the remarkable result that _ abs(_n0 ) & = & k_s()(k_n0a_n0)^2 [ 4.23a ] + _ abs(_n2 ) & = & k_q()(k_n2a_n2)^2 [ 4.23b ] where @xmath140@xmath141(2 + 2@xmath142@xmath143 , @xmath144 being the speed of sound in the detector s material , and @xmath145 its poisson ratio ; @xmath146 is the gravitational constant . the `` remarkable '' about the above is that the coefficients @xmath147 and @xmath148 _ are independent of frequency _ : they _ exclusively depend on the underlying gravitation theory _ , which i symbolically denote by @xmath149 . to see that this is the case , it is enough to consider a monochromatic incident wave : since the coefficients @xmath147 and @xmath148 happen to be _ invariant _ with respect to field amplitude _ scalings _ , this means they will _ only _ depend on the amplitudes relative weights , i.e. , on the field equations _ specific structure_. by way of example , it is interesting to see what the results for general relativity ( gr ) and brans dicke ( bd ) theory are . after somehow lengthy algebra it is found that @xmath150 k_q(\aleph ) = \mbox{\large $ \frac{16\,\pi^2}{15}$ } \end{array}\right . \label{4.25}\ ] ] and @xmath151^{-1 } \\[1.3 em ] k_q(\aleph ) = \mbox{\large $ \frac{16\,\pi^2}{15}$}\,\left[1 + \frac{1}{6}\,(3 + 2\omega)^{-2}\,k\right ] \left[1+\mbox{\large $ \frac{k\omega}{(3 + 2\omega)^2}$}\right]^{-1 } \end{array}\right . \label{4.26}\ ] ] in the latter formulae , @xmath152 is the usual brans dicke parameter @xmath119 @xcite , renamed here to avoid confusion with _ frequency _ , and @xmath153 is a dimensionless parameter , generally of order one , depending on the source s properties @xcite . as is well known , gr is obtained in the limit @xmath154 of bd @xcite ; the quoted results are of course in agreement with that limit . incidentally , an interesting consequence of the above equations is that the presence of a scalar field in the theory of brans and dicke causes _ not only the monopole _ sphere s modes to be excited , _ but also the @xmath52=0 quadrupole ones _ ; what we see in equations ( [ 4.26 ] ) is that _ precisely _ 5/6 of the total energy extracted from the scalar wave goes into the antenna s monopole modes , whilst there is still a remaining 1/6 which is communicated to the quadrupoles , independently of the values of @xmath152 and @xmath155 . this somehow non - intuitive result finds its explanation in the structure of the riemann tensor in bd theory , in which the _ excess _ @xmath156 with respect to general relativity happens _ not _ to be proportional to the scalar part @xmath157 , but to a combination of @xmath157 more in - depth analysis of these facts has been investigated in reference @xcite . equations ( [ 4.23 ] ) show that , no matter which is the gravity theory assumed , the sphere s absorption cross sections for higher modes _ scale _ as the successive coefficients @xmath158 and @xmath159 for monopole and quadrupole modes , respectively . in particular , the result quoted in @xcite that cross section for the second quadrupole mode is 2.61 times less than that for the first , assuming gr , is in fact valid , as we now see , _ independently of which is the ( metric ) theory of gravity actually governing gw physics_. the fourth columns in table [ tab1 ] display these scaling properties . it is seen that the drop in cross section from the first to the second monopole mode is as high as 6.46 . it should however be stressed that the frequency of such mode would be over 4 khz for a ( likely ) sphere whose fundamental _ quadrupole _ frequency be 900 hz @xcite . note finally the asymptotic cross section drop as @xmath160 for large @xmath53 cf . equation ( [ 4.6 ] ) and the ensuing paragraph . the main purpose of this paper has been to set up a _ sound _ mathematical formalism to address with as much generality as possible any questions related to the interaction between a resonant antenna and a weak incoming gw , with much special emphasis on the homogeneous _ sphere_. new results have been found along this line , such as the _ scaling _ properties of cross sections for higher frequency modes , or the sensitivity of the antenna to arbitrary metric gws ; also , new ideas have been put forward regarding the _ direction deconvolution _ problem within the context of an arbitrary metric theory of gw physics . less spectacularly , the full machinery has also been applied to produce _ independent _ checks of previously published results . the whole investigation reported herein has been developed with no _ a priori _ assumptions about any specific ( metric ) theory of the gws , and is therefore _ very _ general . `` too general solutions '' are often impractical in science ; here , however , the `` very general '' appears to be rather `` cheap '' , as seen in the results expressed by the equations of section 3 above . an immediate consequence is that solid elastic detectors of gws ( and , in particular , spheres ) offer , as a matter of principle , the possibility of probing _ any _ given theory of gw physics with just as much effort as it would take , e.g. , to probe general relativity : the vector transfer function of section 4 supplies the requisite theoretical vehicle for the purpose . an important question , however , has not been considered in this paper . this is the _ transducer _ problem : the sphere s oscillations can only be revealed to the observer by means of suitable ( usually electromechanical ) transducers . these devices , however , are not _ neutral _ , i.e. , they _ couple _ to the antenna s motions , thereby exercising a back action on it which must be taken into consideration if one is to correctly interpret the system s readout . preliminary studies and proposals have already been published @xcite , but further work is clearly needed for a more thorough understanding of the problems involved . progress in this direction is currently being made which i expect to report on shortly . the formalism developed in this paper provides basic support to that further work . it is a pleasure for me to thank eugenio coccia for his critical reading and comments on the manuscript . i also want to express gratitude to m montero and ja ortega for their assistance during the first stages of this work , and jmm senovilla for some useful discussions . i have received support from the spanish ministry of education through contract number pb931050 . i am greatly indebted with stephen merkowitz both for his kind supply of the _ tiga _ prototype data , and for continued encouragement and illuminating discussions . i am also indebted with curt cutler for addressing my attention to an initial error in the general equations of section 2 , and with eugenio coccia for many discussions . m.a . serrano gave me valuable help in some of the calculations of appendix [ app : b ] below , and this is gratefully acknowledged , too . i have received financial support from the spanish ministry of education through contract number pb96 - 0384 , and from institut destudis catalans . we say that the vectors ( [ a.1 ] ) are the _ natural basis _ for the @xmath51=1 irreducible representation of the rotation group ; they behave under arbitrary rotations precisely like the spherical harmonics @xmath164 . in particular , if a rotation of angle @xmath165 around the is applied to the original frame then higher rank tensors have specific multipole characteristics depending on the number of tensor indices , and the above basis lends itself to reveal those characteristics , too . for example , the five dimensional linear space of traceless symmetric tensors supports the @xmath51=2 irreducible representation of the rotation group , while a tensor s trace is an invariant . a general symmetric tensor can be expressed as an `` orthogonal '' sum of a traceless symmetric tensor and a multiple of the unit tensor . a convenient basis to expand any such tensor is the following : & e^(1)^(1),^(-1)^(-1 ) & [ a.4a ] + & e^(0)^(1)+e^(1)^(0 ) , e^(0)^(-1)+e^(-1)^(0 ) & [ a.4b ] + & e^(1)^(-1)+e^(-1)^(1 ) - 2e^(0)^(0 ) & [ a.4c ] + & e^(1)^(-1)+e^(-1)^(1 ) + e^(0)^(0 ) & [ a.4d ] the elements ( [ a.4a ] ) get multiplied by e@xmath167 in a rotation of angle @xmath165 around the , respectively , the ( [ a.4b ] ) by e@xmath168 , and ( [ a.4c ] ) and ( [ a.4d ] ) are invariant , as is readily seen . these properties define the `` spin characteristics '' of the corresponding tensors . also , the five elements ( [ a.4a])([a.4c ] ) are _ traceless _ tensors , while ( [ a.4d ] ) is the _ unit _ tensor . any symmetric tensor can be expressed as a linear combination of the six ( [ a.4 ] ) , and the respective coefficients carry the information about the weights of the different monopole and quadrupole components of the tensor . equations ( [ 3.4 ] ) in the text are the matrix representation of the above tensors in the cartesian basis @xmath161 , except that they are multiplied by suitable coefficients to ensure that the conditions this appendix is intended to give a rather complete summary of the frequency spectrum and eigenmodes of a uniform elastic sphere . although this is a classical problem in elasticity theory @xcite , some of the results which follow have never been published so far . also , its scope is to serve as reference for notation , etc . , in future work see ensuing article in page . where @xmath187 is the `` angular momentum '' operator , cf . @xcite , and @xmath188 and @xmath189 are also scalar functions . if ( [ b.1d ] ) and ( [ b.1e ] ) are now respectively substituted in ( [ b.1c ] ) , it is found that @xmath190 , @xmath191 , and @xmath192 satisfy helmholtz equations : in order to ensure regularity at the centre of the sphere , @xmath196=0 . here , @xmath197 is a _ spherical _ bessel function see @xcite for general conventions on these functions , and @xmath198 a spherical harmonic @xcite . finally thus , where @xmath200 are three constants which will be determined by the boundary conditions ( [ b.1 ] ) ( the denominators under them have been included for notational convenience ) . after lengthy algebra , those conditions can be expressed as the following system of linear equations : 1 . _ toroidal _ modes . these are characterised by + @xmath202 + the frequencies of these modes are _ independent _ of @xmath4 , and thence independent of the material s poisson ratio . their amplitudes are + @xmath203 + with + @xmath204 + and @xmath205 a dimensionless normalisation constant determined by the general formula ( [ 2.10 ] ) ; @xmath206 is the @xmath53-th root of the first equation ( [ b.11 ] ) for a given @xmath207 . _ spheroidal _ modes . these correspond to + @xmath208 + and @xmath209=0 . the frequencies of these modes _ do _ depend on the poisson ratio , and their amplitudes are + @xmath210 + where @xmath57 and @xmath58 have the somewhat complicated form + a_nl(r ) & = & c(n , l)[b.16a ] + b_nl(r ) & = & c(n , l ) [ b.16b ] + with accents denoting derivatives with respect to implied ( dimensionless ) arguments , + @xmath211 + and @xmath212 a new normalisation constant . it is understood that @xmath213 and @xmath214 are obtained after the ( transcendental ) equation ( [ b.14 ] ) has been solved for @xmath119 cf . equation ( [ b.4 ] ) . in actual practice equations ( [ b.11 ] ) and ( [ b.14 ] ) are solved for the _ dimensionless _ quantity @xmath215 , which will hereafter be called the _ eigenvalue_. in view of ( [ b.4 ] ) , the relationship between the latter and the measurable frequencies ( in hz ) is given by it is more useful to express the frequencies in terms of the poisson ratio , @xmath217 , and of the speed of sound @xmath218 in the selected material . for this the following formulas are required see e.g. @xcite : tables [ tab2 ] and [ tab3 ] respectively display a set of values of @xmath222 for _ toroidal _ and _ spheroidal _ modes . while gws can only couple to quadrupole and monopole modes , it is important to have some detailed knowledge of analytical results , as the sphere s frequency spectrum is rather involved . it often happens , both in numerical simulations and in experimental determinations , that it is very difficult to disentangle the wealth of observed frequency lines , and to correctly associate them with the corresponding eigenmode . complications are further enhanced by partial degeneracy lifting found in practice ( due to broken symmetries ) , which result in even more frequency lines in the spectrum . accurate analytic results should therefore be very helpful to assist in frequency identification tasks . .list of a few _ spheroidal eigenvalues _ , ordered in columns of ascending harmonics for each multipole value . spheroidal eigenvalues depend on the sphere s material poisson ratio although this dependence is weak . in this table , values are given for @xmath224 . note that the table contains _ all _ eigenvalues less than or equal to 11.024 yet is not exhaustive for values larger than that one ; this would require to stretch the table horizontally beyond @xmath51=10 see figure i for a qualitative inspection of trends in eigenvalue progressions . [ tab3 ] [ cols="^,^,^,^,^,^,^,^,^,^,^,^",options="header " , ] + tuning & 3241 & ( 3241 ) & ( 0 ) & 4 reson . & 3159 & 3155 & @xmath225 + no reson . & 3223 & ( 3223 ) & ( 0 ) & & 3160 & 3156 & @xmath226 + & 3224 & ( 3224 ) & ( 0 ) & & 3168 & 3165 & @xmath225 + & 3236 & ( 3236 ) & ( 0 ) & & 3199 & 3198 & @xmath227 + & 3238 & ( 3238 ) & ( 0 ) & & 3236 & 3236 & @xmath228 + & 3249 & ( 3249 ) & ( 0 ) & & 3285 & 3286 & @xmath229 + 1 reson . & 3167 & 3164 & @xmath230 & & 3310 & 3310 & @xmath228 + & 3223 & 3223 & @xmath228 & & 3311 & 3311 & @xmath228 + & 3236 & 3235 & @xmath172 & & 3319 & 3319 & @xmath228 + & 3238 & 3237 & @xmath172 & 5 reson . & 3152 & 3154 & @xmath231 + & 3245 & 3245 & @xmath228 & & 3160 & 3156 & @xmath232 + & 3305 & 3307 & @xmath233 & & 3163 & 3162 & @xmath234 + 2 reson . & 3160 & 3156 & @xmath235 & & 3169 & 3167 & @xmath230 + & 3177 & 3175 & @xmath236 & & 3209 & 3208 & @xmath172 + & 3233 & 3233 & @xmath228 & & 3268 & 3271 & @xmath237 + & 3236 & 3236 & @xmath228 & & 3304 & 3310 & @xmath238 + & 3240 & 3240 & @xmath228 & & 3310 & 3311 & @xmath229 + & 3302 & 3303 & @xmath229 & & 3313 & 3316 & @xmath237 + & 3311 & 3311 & @xmath228 & & 3319 & 3321 & @xmath233 + 3 reson . & 3160 & 3155 & @xmath239 & 6 reson . & 3151 & 3154 & @xmath240 + & 3160 & 3156 & @xmath235 & & 3156 & 3155 & @xmath234 + & 3191 & 3190 & @xmath172 & & 3162 & 3162 & @xmath228 + & 3236 & 3235 & @xmath172 & & 3167 & 3162 & @xmath232 + & 3236 & 3236 & @xmath228 & & 3170 & 3168 & @xmath236 + & 3297 & 3299 & @xmath231 & & [ 3239 ] & [ 3241 ] & [ 6 ] + & 3310 & 3311 & @xmath241 & & 3302 & 3309 & @xmath242 + & 3311 & 3311 & @xmath228 & & 3308 & 3310 & @xmath233 + & & & & & 3312 & 3316 & @xmath243 + & & & & & 3316 & 3317 & @xmath241 + & & & & & 3319 & 3322 & @xmath237 equation ( [ 8.5 ] ) can now be readily solved , once the resonator positions are fed into the matrices @xmath244 . such positions correspond to the pentagonal faces of a truncated icosahedron . merkowitz @xcite gives a complete account of all the measured system frequencies as resonators are progressively attached to the selected faces , beginning with one and ending with six . figure [ fig8 ] graphically displays the experimentally reported frequencies along with those calculated theoretically by solving equation ( [ 8.5 ] ) . in table 1 i give the numerical values . as can be seen , coincidence between theoretical predictions and experimental data is remarkable : the worst error is 0.2% , while for the most part it is below 0.1% . this is a few parts in 10@xmath245 , which is precisely the magnitude of @xmath246 , as specified in equation ( [ 8.7 ] ) . therefore _ discrepancies between theoretical predictions and experimental data are exactly as expected _ , i.e. , of order @xmath246 . in addition , it is also reported in reference @xcite that the 11-th , weakly coupled mode of the _ tiga _ ( enclosed in square brackets in table 1 ) has a practically zero amplitude , again in excellent agreement with the general theoretical predictions about modes beyond the tenth see paragraph after equation ( [ 8.5 ] ) . this is a remarkable result which encouraged a better fit by estimates of _ next order _ corrections , i.e. , @xmath247 of ( [ 4.11.a ] ) . as it turned out , however , matching between theory and experiment does not consistently improve in the next step . this is not really that surprising , though , as m&j explicitly state @xcite that control of the general experimental conditions in which data were obtained had a certain degree of tolerance , and they actually show satisfaction that @xmath2481% coincidence between theory and measurement is comfortably accomplished . but 1% is _ two orders of magnitude larger than @xmath249 _ cf . equation ( [ 8.7]) , so failure to refine our frequency estimates to order @xmath249 is again fully consistent with the accuracy of available real data . a word on a technical issue is in order . merkowitz and johnson s equations for the _ tiga _ @xcite are identical to the equations in this paper to lowest order in @xmath249 . remarkably , though , their reported theoretical estimates of the system frequencies are not quite as accurate as those in table 1 @xcite . the reason is probably this : in m&j s model these frequencies appear within an algebraic system of 5+@xmath250 linear equations with as many unknowns which has to be solved ; here instead the algebraic system has only @xmath250 equations and unknowns , actually equations ( [ m3.16 ] ) . this is a very appreciable difference for the range of values of @xmath250 under consideration . while the roots for the frequencies _ mathematically _ coincide in both approaches , in actual practice they are _ estimated _ , generally by means of computer programmes . it is here that problems most likely arise , for the numerical reliability of an algorithm to solve matrix equations normally decreases as the rank of the matrix increases . where @xmath249 can be defined e.g. as the ratio of the _ average _ resonator mass to the sphere s mass . it is immediately obvious from equation ( [ 8.9 ] ) that mass non - uniformities of the resonators only affect the main equations in _ second order _ , since resonator mass non - uniformities result , as we see , in corrections of order @xmath252 to @xmath252 itself , which is the very parameter of the perturbative expansions . the system is thus clearly _ robust _ to mismatches in the resonator masses of the type ( [ 8.9 ] ) . the same happens if the _ locations _ of the resonators have tolerances relative to a _ pre - selected _ distribution . for let @xmath253 be a set of resonator locations , for example the _ tiga _ or the _ phc _ positions , and let @xmath254 be the real ones , close to the former : the values @xmath256 determine the eigenvalues @xmath257 in equation ( [ 5.2 ] ) , and they also appear as arguments to the spherical harmonics in the system response functions of sections [ sec : srgw][sec : hs ] . it follows from ( [ 8.10 ] ) by continuity arguments that inspection of the equations of sections [ sec : srgw][sec : hs ] shows that both @xmath257 and @xmath258 _ always _ appear within lowest order terms , and hence that corrections to them of the type ( [ 8.105 ] ) will affect those terms in _ second order _ again . we thus conclude that the system is also _ robust _ to small misalignments of the resonators relative to pre - established positions . to assess the consequences of this , however , we must go back to equation ( [ 3.18 ] ) and see what the coefficients in its series solutions of the type ( [ 4.11.a ] ) are . the procedure is very similar to that of section [ s8.1 ] , and will not be repeated here ; the lowest order coefficient @xmath260 is seen to satisfy the algebraic equation which reduces to ( [ 5.1 ] ) when all the @xmath262 s vanish , as expected . this appears to potentially have significant effects on our results to lowest order in @xmath252 , but a more careful consideration of the facts shows that it is probably unrealistic to think of such large tolerances in resonator manufacturing as implied by equation ( [ 8.11 ] ) in the first place . in the _ tiga _ experiment , for example @xcite , an error of order @xmath252 would amount to around 50 hz of mistuning between resonators , an absurd figure by all means . in a full scale sphere ( @xmath24840 tons , @xmath2483 metres in diameter , @xmath248800 hz fundamental quadrupole frequency , @xmath249@xmath24810@xmath263 ) the same error would amount to between 5 hz and 10 hz in resonator mistunings for the lowest frequency . this is probably excessive for a capacitive transducer , but may be realistic for an inductive one . with this exception , it is thus more appropriate to consider that resonator mistunings are at least of order @xmath249 . if this is the case , though , we see once more that the system is quite insensitive to such mistunings . summing up the results of this section , one can say that the resonator system dynamics is quite _ robust _ to small ( of order @xmath252 ) changes in its various parameters . the important exception is of course the effect of suspension drilling , which do result in significant changes relative to the ideally perfect device , but which can be relatively easily calculated . the theoretical picture is fully supported by experiment , as _ robustness _ in the parameters here considered has been reported in the real device @xcite . where @xmath165 is a suitable label . it is recalled from reference @xcite that , in such circumstances , a formal solution can be written down for equation ( [ 2.1.a ] ) in terms of a _ green function integral _ , whereby the following orthogonal series expansion obtains : here , @xmath48 and @xmath16 are the eigenfrequencies and associated normalised wave - functions of the free sphere . also , @xmath18 is an abbreviation for a multiple index @xmath50 . the generic index @xmath83 is a label for the different pieces of interaction happening in the system . i quote the result of the calculations of the terms needed in this paper : where the label `` external '' explicitly refers to agents acting upon the system from outside . two kinds of such external actions are considered in this article : those due to gws and those due to a calibration hammer stroke signal . specifying x*=@xmath268 in the lhs of ( [ m3.6 ] ) and multiply on either side by @xmath253 , the following is readily found : the following bare sphere responses to gws and hammer strokes ( equations!([1.1 ] ) and ( [ 2.5 ] ) , respectively ) can be calculated by direct substitution . the results as best presented as laplace transform domain functions : we must first isolate @xmath284 , then find inverse laplace transforms to revert to time domain quantities . substituting the values of @xmath285 and @xmath286 from ( [ 6.2.a ] ) and ( [ 4.85 ] ) into ( [ a2.1 ] ) we find clearly thus , the _ poles _ of @xmath291 must be determined in the first place . it is immediately clear from equation ( [ 6.4 ] ) that there are no poles at either @xmath293=0 , or @xmath293=@xmath294 , or @xmath293=@xmath295 , for there are exactly compensated infinities at these locations . the only possible poles lie at those values of @xmath293 for which the matrix in square brackets in ( [ 6.4 ] ) is not invertible , and these of course correspond to the zeroes of its determinant , i.e. , there are infinitely many roots for equation ( [ m3.18 ] ) , but _ analytic _ expressions can not be found for them . _ perturbative _ approximations in terms of the small parameter @xmath249 will thus be applied instead . it is assumed that since @xmath249 is a small parameter , the _ denominators _ of the fractions in the different terms in square brackets in ( [ m4.9 ] ) must be _ quantities of order @xmath249 _ at the root locations for the determinant to vanish at them . a distinction however arises depending on whether @xmath300 is close to @xmath301 or to the other @xmath302 . there are accordingly two categories of roots , more precisely : ) in the complex @xmath293plane for @xmath152 tuned to the lowest frequency of the sphere s spectrum . note that @xmath303 is closer to @xmath304 than @xmath305 are to @xmath306 ; this is a consequence of the differences being proportional to @xmath249 and @xmath252 , respectively , as seen in equations ( [ m4.11 ] ) . [ fig9],width=491 ] figure [ fig9 ] provides a visual characterisation of these roots . the coefficients @xmath260 , @xmath247 , ... and @xmath307 , @xmath307 , ... can be calculated recursively , starting form the first , by substitution of the corresponding series expansions into equation ( [ m4.9 ] ) . the lowest order terms are easily seen to be given by respectively . both equations ( [ 5.1 ] ) and ( [ 5.4 ] ) are algebraic eigenvalue equations . as shown in appendix [ app : c ] , the matrix @xmath310 has at most @xmath311 non - null positive eigenvalues all the rest up to @xmath250 are identically zero . as a final step we must evaluate ( [ 6.6 ] ) . this is accomplished by standard textbook techniques ( see e.g. @xcite ) ; the algebra is quite straightforward but rather lengthy , and i shall not delve into its details here , but quote only the most interesting results . it appears that the _ dominant _ contribution to @xmath290 comes from the poles at the locations ( [ 4.11.a ] ) , whereas all other poles only contribute as higher order corrections ; generically , @xmath290 is seen to have the form this appendix presents a few important properties of the matrix @xmath315 for arbitrary @xmath51 and resonator locations @xmath253 ( @xmath316=1, ,@xmath250 ) which are useful for detailed system resonance characterisation . equation ( [ ma.3 ] ) indicates that the _ rank _ of the matrix @xmath319 can not exceed @xmath311 , as there are only @xmath311 kets @xmath321 . so , if @xmath250@xmath324@xmath311 then it has at least @xmath325 identically null eigenvalues there can be more if some of the @xmath253 s are parallel , as this causes rows ( or columns ) of @xmath319 to be repeated . in other words , @xmath328 are the eigenvalues of the @xmath311@xmath320@xmath311 matrix @xmath331 , which is positive definite because so is the `` scalar product '' @xmath332 . all of them are therefore strictly positive . lobo j.a . and ortega j.a . , in coccia e. , pizzella g. , ronga f. , eds . , _ proceedings of the first edoardo amaldi conference on gravitational waves _ ( frascati 1994 ) , pag 449 , world scientific , singapore ( 1995 ) . lobo j.a . , in e. coccia , g. pizzella , g. veneziano , eds , _ proceedings of the second edoardo amaldi conference on gravitational waves _ ( _ cern _ , geneva 1997 ) , pag 168 - 179 , world scientific singapore ( 1998 ) . a spherical gw antenna is a natural multi - mode device with very rich potential capabilities to detect gws on earth . but such detector is not just a bare sphere , it requires a set of _ motion sensors _ to be practically useful . it appears that transducers of the _ resonant _ type are the best suited ones for an efficient performance of the detector . resonators however significantly interact with the sphere , and they affect in particular its frequency spectrum and vibration modes in a specific fashion , which must be properly understood before reliable conclusions can be drawn from the system readout . the main objective of this paper has been the construction and development of an elaborate theoretical model to describe the joint dynamics of a solid elastic sphere and a set of _ radial motion _ resonators attached to its surface at arbitrary locations , with the purpose to make predictions of the system characteristics and response , in principle with arbitrary mathematical precision . the solutions to the equations of motion have been shown to be expressible as an ascending series in powers of the small `` coupling constant '' @xmath249 , the ratio of the average resonator mass to the mass of the larger sphere . the _ lowest order _ approximation corresponds to terms of order @xmath252 and , to this order , previous results @xcite are recovered . this , i hope , should contribute to clarify the nature of the approximations inherent in earlier approaches , and to better understand the physical reason for their remarkable accuracy @xcite . in addition , the methods of this paper have permitted us to discover that there can be in fact transducer layouts alternative to the highly symmetric _ tiga _ , and having potentially interesting practical properties . an example is the _ phc _ distribution , which is based on a pentagonally symmetric set of 5 rather than 6 resonators per quadruopole mode sensed . this transducer distribution has the property that _ mode channels _ can be constructed from the resonators readouts , much in the same way as in the _ tiga _ @xcite . in the _ phc _ however a new and distinctive characteristic is present : different _ wave amplitudes _ selectively couple to different _ detector modes _ having different frequencies , so that the antenna s mode channels come at different rather than equal frequencies . phc _ philosophy can be extended to make a _ multi - frequency _ system by using resonators tuned to the first two quadrupole harmonics of the sphere _ and _ to the first monopole , an altogether 11 transducer set . the assessment of _ symmetry failure _ effects , as well as other parameter departures form ideality , has also been subjected to analysis . the general scheme is again seen to be very well suited for the purpose , as the theory transparently shows that the system is _ robust _ against relative disturbances of order @xmath249 or smaller in any system parameters , also providing a systematic procedure to assess larger tolerances up to order @xmath252 . the system is shown to still be robust to tolerances of this order in some of its parameters , whilst it is not to others . included in the latter group is the effect of spherical symmetry breaking due to system suspension in the laboratory , which causes _ degeneracy lifting _ of the sphere s eigenfrequencies , which split up into multiplets . a strong point is that , by use of mostly analytic algorithms , it has been possible to accurately reproduce the reportedly measured frequencies of the _ lsu _ prototype antenna @xcite with the predicted precision of four decimal places . the also reported robustness of the system to resonator mislocations @xcite is too in satisfactory agreement with the theoretical predictions . the perturbative approach here adopted is naturally open to refined analysis of the system response in higher orders in @xmath249 . for example , one can systematically address the weaker coupling of non - quadrupole modes , etc . it appears however that such refinements will be largely masked by _ noise _ in a real system , as shown by merkowitz and johnson @xcite , and this must therefore be considered first . so the next step is to include noise in the model and see its effect . stevenson @xcite has already made some progress in this direction , and partly assessed the characteristics of _ tiga _ and _ phc _ , but more needs to be done since not too high signal - to - noise ratios should realistically be be considered in an actual gw detector . in particular , _ mode channels _ are at the basis of noise correlations and dependencies , as well as the errors in gw parameter estimation @xcite . i do expect the analytic tools developed in this article to provide a powerful framework to address the fundamental problems of _ noise _ in a spherical gw antenna which , to my knowledge , have not yet received the detailed attention they require .	gravitational waves ( gw ) are propagated perturbations of the space - time geometry , and they show up locally as _ tides _ , i.e. , local gravity gradients which vary with time . these gradients are determined by _ six _ often called _ electric _ components of the riemann tensor , which in a local coordinate frame are @xmath0 , ( @xmath1 ) . a gw detector is a device designed to generate information on those quantities on the basis of suitable measurements . both weber bars and laser interferometers are _ single mode _ antennas , i.e. , they generate a _ single _ readout which is the result of the combined action of those six gw s riemann tensor amplitudes on the system sensor s output port . a _ spherical _ detector is _ not _ limited in that fashion : this is because its _ degenerate _ oscillation eigenmodes are uniquely matched to the structure of the above riemann tensor components . this means that a solid spherical body is a natural _ multimode _ gw detector , i.e. , it is fully capable of delivering six channel outputs , precise combinations of which completely deconvolve the six gw amplitudes . the present article is concerned with the theoretical reasons of the remarkable properties of a spherical gw detector . the analysis proceeds from _ first principles _ and is based on essentially no _ ad hoc _ hypotheses . the mathematical beauty of the theory is outstanding , and abundant detail is given for a thorough understanding of the fundamental facts and ideas . experimental evidence of the accuracy of the model is also provided , where possible . * multimode gravitational wave detection : * * the spherical detector theory * * jos alberto lobo * + departament de fsica fonamental + universitat de barcelona , spain + e - mail : [email protected]
long range correlations are a well known property of non - equilibrium systems in their steady state@xcite . for one dimensional diffusive systems , which satisfy fourier s law , these correlations ( between pairs of points separated by a macroscopic distance ) scale like the inverse @xmath0 of the system size @xmath1@xcite . they can also be related to the non local nature of the the large deviation functional of the density profiles@xcite . these long range correlations have been calculated in a few cases , using in particular fluctuating hydrodynamics@xcite . it was observed that when a diffusive system approaches a second order phase transition , the factor in front of these @xmath0 correlations becomes singular@xcite . the goal of the present work is to analyse these long range correlations at and in the neighborhood of a phase transition . to do so , we consider the @xmath2 model , one of the simplest diffusive systems which undergoes a phase transition . the main result obtained in the present work is that the long range correlations at the second - order phase transition of the @xmath2 model have an anomalous dependence on the system size which can be understood by an effective theory @xcite for the amplitude of the slow density mode which becomes unstable at the transition . this effective theory , which was developed in a recent work @xcite , led us to predict an anomalous size dependence of the cumulants of the particle current at the transition . here , we also show that , away from the transition , these cumulants satisfy fourier s law ( i.e. are proportional to @xmath0 ) , with prefactors which become singular at the transition . the outline of the paper is as follows : in section [ sec : abc ] , we first recall some known properties of the @xmath2 model as well as our previous results@xcite on the cumulants of the current , which exhibit an anomalous fourier s law at the second - order transition . in sections [ sec : relax ] and [ sec : steady ] , we show how the effective theory for the slow density mode yields a @xmath3 size dependence of the correlation function in the critical regime . in section [ sec : cur ] , we calculate the cumulants of the particle current ( all the cumulants in the flat phase and the first two cumulants in the modulated phase ) . they all decay like @xmath0 , with prefactors which become singular at the transition , and which match the expressions obtained directly in the critical regime in @xcite . the @xmath2 model is a one - dimensional lattice gas , where each site is occupied by one of three types of particles , @xmath4 , @xmath5 or @xmath6 . neighboring sites exchange particles at the rates @xmath7 with an asymmetry @xmath8 . this model has been studied on a closed and on an open interval ( with particle reservoirs at each end)@xcite as well as on a ring with periodic boundary conditions@xcite : in this article , we consider the latter case by studying a ring of @xmath1 sites . for @xmath9 , all configurations are equally likely in the steady state ; on the other hand , for @xmath10 , the particles of the same species tend to gather @xcite . when @xmath11 scales as @xmath12 the dynamics of the model becomes diffusive@xcite : for each species @xmath13 , one can define a rescaled density profile @xmath14 , @xmath15 = \rho_a(k / l , t / l^2 ) \equiv \rho_a(x,\tau)\,.\ ] ] because the microscopic dynamics conserves the numbers of particles , the @xmath16 are related to their associated currents @xmath17 by the conservation laws @xmath18 by assuming that the underlying microscopic regions of the system are local equilibrium @xcite , one can show that the currents @xmath17 satisfy noisy , biased fick s laws : @xmath19 where @xmath20 and @xmath21 designate the previous and next species with respect to @xmath22 , and where the @xmath23 are gaussian white noises such that @xmath24 ( note that these correlations imply @xmath25 due to the fact that @xmath26 is identically zero ) . at leading order in @xmath1 , the noise in can be neglected , so that leads to the evolution equations@xcite @xmath27 for low @xmath28 , the flat density profiles @xmath29 are a stable solution of these equations . they become linearly unstable@xcite when @xmath28 reaches @xmath30 given by @xmath31 indicating a second - order transition at @xmath32 : above @xmath30 , the steady - state profiles are modulated . it has been argued that these steady - state density profiles are time - independent@xcite . this stability analysis does not rule out the possibility of a first - order phase transition taking place at some @xmath33 . a more detailed analysis of the neighborhood of @xmath30 shows that for @xmath34 , the transition should be first order@xcite , with @xmath35 while no analytical expression for @xmath36 is known in this case , it has been studied numerically in @xcite . in @xcite , we considered the @xmath2 model in the region @xmath37 where the transition is expected to be second order . we found that , for a system of size @xmath1 , there exists a critical regime @xmath38 for which the dynamics of the density profiles is dominated by those of their first fourier mode . let @xmath39 be the first fourier mode of the density of species @xmath4 . then , near the transition , the first fourier modes of the other densities , @xmath40 and @xmath41 , are related to @xmath42 by @xmath43 and the evolution of @xmath44 can be described by a langevin equation in a quartic potential in terms of the diffusive time @xmath45 : @xmath46r_a + { \mu_a(\tau)\over\sqrt{l}}\ ] ] with @xmath47 and with @xmath48 is a complex gaussian white noise : @xmath49 by rescaling @xmath50 by @xmath51^{1/4 } f(\bar\tau)\,,\ ] ] one can see that becomes @xmath52 where @xmath53 and with @xmath54 hence , in the critical regime @xmath55 , the amplitude @xmath42 of the first fourier mode varies on a time scale @xmath56 , with an amplitude in @xmath57 . in @xcite , we showed that , due to these slow fluctuations of @xmath44 , the integrated particle current @xmath58 of @xmath4 particles during time @xmath59 through a section in the system exhibits anomalous fluctuations at the transitions , reminiscent of those that can be numerically observed in momentum - conserving mechanical models@xcite : @xmath60 we also argued that the coefficients @xmath61 can be expressed in terms of @xmath62-point correlation functions of the solution @xmath63 of : @xmath64 therefore in @xcite we reduced the calculations of the cumulants of @xmath58 in the critical regime to the study of the langevin equation of a single particle evolving in a quartic potential . relaxation of the first - mode amplitude @xmath65 measured using for systems of size @xmath66 started in a segregated initial configuration , for @xmath67 and @xmath68 . our analysis of the first mode of the densities predicts a @xmath69 decay ( see ) . ] in this section , we study the decay of the first fourier mode in the critical regime when its initial amplitude is much larger than its steady - state values @xmath70 . this is the case when one starts from a non steady - state initial condition : here we consider such a relaxation , starting from a fully segregated initial configuration of the type @xmath71 . according to @xcite , we expect all the higher fourier modes of the densities to relax on the hydrodynamic time scale @xmath45 . after this initial relaxation , the density profiles should be dominated by a first fourier mode with amplitude @xmath44 evolving according to , ; moreover , as long as @xmath42 remains much larger than @xmath72 , the noise term in can be neglected and the evolution of @xmath42 reduces to @xmath73 r_a\,.\ ] ] when @xmath74 ( i.e. when @xmath75 ) , @xmath42 should thus decay as a power law : @xmath76 in figure [ fig : decay ] , we measured numerically the amplitude @xmath77 from its definition for systems of @xmath78 particles for @xmath67 and @xmath68 . the power - law decay should be valid in a rather limited range of time ( @xmath79 ) , and is rather difficult to observe . fortunately , the higher fourier modes seem to relax fast enough for the power - law decay to occur already for @xmath80 : our data for increasing sizes seems to converge to the power - law in the whole range @xmath81 . figure [ fig : decay ] also shows a departure from this power law at a rescaled time @xmath82 increasing with @xmath1 : in the next section , we show that this is due to the increasing effect of the noise term in as @xmath42 decreases . the damping term of the evolution equation for @xmath42 that we obtained around @xmath68 vanishes on the tricritical line @xmath83 ( see eq . ) . in this case , it is necessary to push the analysis of @xcite further in order to obtain an effective equation for @xmath42 . one can show that , when @xmath84 , is replaced by @xmath85 along this tricritical line ( @xmath75 , @xmath84 ) , the @xmath86 decay of @xmath42 in the critical regime should become in @xmath87 . it has been shown that the @xmath2 model exhibits long - range steady - state correlations , scaling as @xmath0 , in the flat phase @xmath88 @xcite . in the modulated phase @xmath89 , these correlations should be of order @xmath90 due to the modulation of the steady - state profiles . in this section , we show , using our effective dynamics , for the first mode @xmath42 , that these correlations scale as @xmath91 in the critical regime ; we also show that temporal correlations decay on the slow time scale @xmath92 at the transition . finally , we briefly comment on the behavior on the tricritical line @xmath84 . -\sum r_a^2 $ ] measured for systems of size @xmath93 , with @xmath67 and @xmath68 . our analysis of the first mode of the densities predicts cosine correlations of an amplitude scaling as @xmath91 . ] [ fig : steady ] as explained in @xcite , the density fluctuations in the critical regime are dominated by those of the first fourier mode @xmath42 . thus one can calculate to leading order the steady - state correlations of the densities at the critical point : @xmath94 - \sum_a r_a^2 & \simeq \sum_a { \left\langle}\left(r_a + r_a { { \rm e}}^{2{{\rm i}}\pi k / l}+cc.\right)\left(r_a + r_a { { \rm e}}^{2{{\rm i}}\pi l / l}+cc.\right){\right\rangle}_c\,.\nonumber\end{aligned}\ ] ] the langevin equation can be expressed as a fokker - planck equation over the probability density of the rescaled first mode @xmath63 , + @xmath95 $ ] : @xmath96 + { 1\over 4 } { \nabla^2 } p\ ] ] with @xmath97 and @xmath98 as defined in . from this equation , it is easy to see that @xmath63 is isotropically distributed in the steady state : @xmath99\,.\ ] ] therefore , @xmath100 and @xmath101 it is easy to compute @xmath102 this leads to @xmath103 - \sum r_a^2 \simeq 2 \sqrt{3\delta r_a r_b r_c\over \lambda l } c_1(\bar \gamma ) \cos 2\pi{k - l\over l}\,.\ ] ] therefore , steady - state correlations should scale as @xmath91 in the critical regime . in figure [ fig : steady ] , we compare our prediction to numerical measurements of systems of @xmath104 particles , for @xmath68 and @xmath67 . + + * remark : * the @xmath105 and @xmath106 limits of our expression for the equal - time correlations can both be checked from known results . in @xcite , exact expressions for the equal - time density correlations have been calculated in the flat phase ( @xcite , eq . ( 23 ) ) : they diverge as @xmath107 , leading to @xmath108 - \sum r_a^2 \simeq { 6r_a r_b r_c \over \delta l}{\beta_\over \beta_-\beta}\cos 2\pi{k - l\over l}\,,\ ] ] which is compatible with @xmath109 for @xmath105 in . on the other hand , equation ( 39 ) of @xcite establishes that the modulated steady - state profiles in the disordered phase are , for @xmath110 , of the form @xmath111 with @xmath112 and @xmath113 an arbitrary phase . this leads to @xmath108 - \sum r_a^2 \simeq { \delta^2\over \lambda}{\beta-\beta_\over \beta_}\cos 2\pi{k - l\over l}\,,\ ] ] which is compatible with @xmath114 for @xmath115 in and . -\sum r_a^2 $ ] for systems of size @xmath116 , with @xmath67 and @xmath68 , as functions of @xmath117 . our effective , fluctuating evolution equation for the first fourier mode leads us to predict an exponential decay . ] [ fig : decayexp ] in the previous section , we have seen that , on a time scale @xmath79 , the first fourier mode of the densities should decay as a power law when the system relaxes from a non steady state initial configuration . here we consider the decay of the steady - state density correlations @xmath118 : the noisy evolution equation implies that they should decay on the `` slow '' time scale @xmath119 . they can be expressed in terms of the rescaled first mode @xmath63 as @xmath120 -\sum r_a^2\propto { 1\over \sqrt{l}}{\left\langle}f(0)f^(\bar\tau ) + cc.{\right\rangle}_c\ ] ] with @xmath121 the rescaled time . let @xmath122 be the operator of the fokker - planck equation over the probability density of @xmath63 , @xmath123 : @xmath124 for large @xmath125 , correlations such as @xmath126 should decay as @xmath127 , where @xmath128 is the second largest eigenvalue of @xmath122 ( the largest being @xmath129 ) . thus , we expect time correlations to decay on the time scale @xmath130 as @xmath131-\sum r_a^2 \propto { 1\over \sqrt{l } } \exp\left [ -\alpha { t\over l^{5/2 } } \right]\ ] ] with @xmath132 . while there is no known analytical expression for @xmath128 , one can determine it numerically by approximating the operator @xmath122 over a finite subspace of @xmath133 of growing dimension . because the steady - state ( and thus the @xmath129-eigenvector of @xmath122 ) is @xmath134 , we consider the subspace @xmath135 where @xmath136 is a polynomial of degree less than some finite @xmath137 in @xmath138 . we need to define a scalar product @xmath139 for which @xmath122 is hermitian : to do so , we choose @xmath140 for which the matrix elements of @xmath122 read @xmath141 for @xmath142 , we constructed an orthonormal basis of the subspace of the @xmath143 with respect to the scalar product for @xmath144 ; then , we computed the matrix elements of @xmath145 in this basis . the second largest eigenvalue of these successive hermitian matrices appears to converge to @xmath146 in figure [ fig : decayexp ] , we compare this prediction with the results of simulations in the steady - state for @xmath147 , @xmath67 and @xmath68 . the exponential decay also seems to predict the cut - off of the power law decay of figure [ fig : decay ] observed in systems relaxing from an arbitrary initial condition to the stationary state . as in the deterministic case the cubic term of the fluctuating evolution equation for the first fourier mode vanishes on the tricritical line @xmath83 . the fluctuating correction of the deterministic tri - critical evolution equation is the same as those of : @xmath148 this leads to the rescaling @xmath149{6 r_a^4r_b r_c}\over l^{1/6 } } g(\tilde\tau ) \mbox { with } { { { \rm d}}g\over { { \rm d}}\tilde\tau } = ( \tilde\gamma-\|g\|^4 ) g + \mu(\tilde\tau ) \,,\ ] ] with @xmath150 hence , the tricritical regime occurs for @xmath151 : it is characterized by fluctuations of the first fourier mode of amplitude @xmath152 on a time scale @xmath153 . in its steady state , we expect spatial correlations to scale as @xmath154-\sum r_a^2 \propto { 1\over l^{1/3}}\cos 2\pi{k - l\over l}\,,\ ] ] while time correlations should decay exponentially on the time scale @xmath153 , as @xmath155-\sum r_a^2 \propto { 1\over l^{1/3}}\exp\left[-\alpha { t\over l^{8/3}}\right]\,.\ ] ] in this section , we study the cumulants of the integrated current of particles of type @xmath4 , @xmath58 , as @xmath156 . in contrast to @xcite , where we studied the critical regime @xmath157 , here we compute the cumulants to leading order in @xmath1 for a fixed @xmath158 : the expressions we obtain diverge at @xmath30 . for @xmath107 , we find @xmath159 while we find that , for @xmath160 , @xmath161 we also show that these expressions are compatible with the @xmath162 limits of the expressions of the cumulants , in terms of the functions @xmath163 derived in the critical regime in @xcite . we start with the assumption that the steady state density profiles @xmath164 , which are the long - time limits of the leading - order , deterministic hydrodynamic equations , are time - independent , both in the flat and in the modulated phase . the associated particle currents @xmath165 are thus homogeneous , @xmath166 , and can be computed by integrating in space , yielding @xmath167 then , the average integrated current @xmath168 through any position in the system will behave in the long - time limit like @xmath169 , so that @xmath170 for @xmath171 , we thus obtain @xmath172 . for @xmath89 , the analytic expression of the @xmath164 is rather complicated@xcite . however , their limit as @xmath110 takes a simple form ( see @xcite , or the @xmath115 limit of ): @xmath173 with @xmath174 and with @xmath175 , @xmath176 related to @xmath42 by . this leads to an analytical expression for @xmath168 when @xmath110 : @xmath177\,.\ ] ] in order to compute fluctuations of the current @xmath58 , the noise terms of the biaised fick s law have to be taken into account . these fluctuating hydrodynamics can be reformulated as a large deviation principle known as the macroscopic fluctuation theory ( mft)@xcite , which gives the probability of observing a given evolution @xmath16 of the density profiles : @xmath178 \propto \exp\left[-l\iint dxd\tau { 1\over 2 } ( j - q).\sigma^{-1}(j - q)\right]\ ] ] where @xmath179 , @xmath180 ( with @xmath181 the deterministic part of ) and with @xmath182 the generating function @xmath183 can then be calculated as the solution of a variational problem : @xmath184\,.\ ] ] for @xmath185 , the optimum in the above is the solution to the deterministic equations , which satisfy by definition @xmath186 : these are @xmath29 ( for @xmath187 ) or @xmath188 ( for @xmath189 ) , and @xmath190 defined by . for @xmath171 , the optimal profiles are still flat even for @xmath191 ( as we will see below , they are a stable minimum ) : @xmath192 with @xmath193 , @xmath194 , and @xmath11 and @xmath195 taking their constant values for @xmath196 . this leads to @xmath197 \equiv { t\over l}f_{\rm flat}(\lambda)\ ] ] and to @xmath198 on the other hand , the optimal profiles in vary with @xmath199 for @xmath200 , with non - trivial optimization equations . we will therefore restrict ourselves to the calculation of @xmath201 , for which it is sufficient to compute to second order in @xmath199 . in order to do so , we consider profiles close to the @xmath202 optimum , @xmath203 and @xmath204 . because arbitrary translations of these profiles are also optimal , we need to consider profiles moving at a small velocity @xmath205 : this amounts to @xmath206 with @xmath193 and @xmath207 . because @xmath208 , this leads to the following expression for @xmath11 : @xmath209 @xmath210 and @xmath211 . then , the right - hand side of becomes @xmath212\end{gathered}\ ] ] with @xmath213 . the optimization equations over @xmath214 , @xmath205 and @xmath215 can be written as @xmath216 with @xmath217 . the right - hand side of then reads @xmath218\ ] ] it can be expressed completely in terms of the optimal @xmath219 solution of by using @xmath220 so that the generating function is given to second order in @xmath199 by @xmath221\ ] ] by determining the solution @xmath219 of and integrating numerically , @xmath222 can be predicted to leading order in @xmath1 for @xmath189 . this prediction diverges as @xmath110 : because @xmath223 is known analytically in this limit , we were able to calculate @xmath219 exactly , @xmath224 leading to a simple expression for @xmath222 in this limit : @xmath225 as shown above , the macroscopic fluctuation theory predicts , for large @xmath1 , gaussian fluctuations of @xmath58 in the flat phase @xmath88 , with a non - singular variance as @xmath107 . here , we calculate the next @xmath0 corrections to the generating function , by generalizing to the @xmath2 model the approach followed in @xcite in the case of a single conserved quantity . these corrections , of order @xmath226 , can be obtained by considering the large - deviation principle as a functional integral : @xmath227d[j_a ] \exp\left[l \iint dxd\tau\left(\lambda j_a - { 1\over 2}(j - q).\sigma^{-1}(j - q)\right)\right]\ ] ] where the integral takes place over all profiles @xmath228 compatible with the conservation law @xmath229 . fluctuations around the optimum profile then give corrections to the saddle - point expression . we now consider such fluctuations , expressing them in terms of their fourier modes : @xmath230\\ \delta j(x,\tau ) & = \sum_{k,\omega } \omega\left[\alpha_{k\omega}{{\rm e}}^{{{\rm i}}(kx-\omega t ) } + cc.\right ] \end{aligned}\right.\ ] ] with @xmath231 the amplitude of the fluctuations of wave number @xmath232 and pulsation @xmath233 , which take discrete values : @xmath234 expanding to second order in the @xmath235 , we obtain @xmath236 \exp\left[{t\over l}f_{\rm flat } ( \lambda ) -\sum_{k,\omega } \alpha_{k\omega}^ . m_{k\omega}\alpha_{k\omega}\right]\ ] ] with @xmath237 the dominant - order generating function and + @xmath238 , with @xmath239 integrating the gaussian variables @xmath235 gives the corrections @xmath240 with @xmath241 and @xmath6 an additive constant fixed by the condition @xmath242 for @xmath202 . in the @xmath243 limit , the sum over @xmath233 can be replaced by an integral : @xmath244 because only the first mode of the fluctuations , @xmath245 , becomes unstable as @xmath246 , we expect the divergence in the cumulants at the transition to only affect this @xmath245 term . thus we take the following limit in @xmath247 : @xmath248 ( to determine the cumulants ) , @xmath249 ( we expect slow fluctuations to be responsible for the divergence ) , and @xmath250 . then @xmath247 takes the simplified expression @xmath251\equiv q(\lambda,\omega)\,.\ ] ] the integral of @xmath252 is apparently divergent : however , since @xmath253 its divergent part is canceled out by adjusting @xmath6 so that @xmath242 for @xmath202 . hence , the part of the generating function which becomes singular as @xmath156 reads @xmath254\,,\ ] ] leading to a divergence of the @xmath62-th cumulant of @xmath58 scaling as , with @xmath255^n \,.\ ] ] in @xcite , we found that the fluctuations of the first fourier mode on the slow time scale @xmath256 lead to anomalous fluctuations of the integrated current of @xmath58 in the critical regime . more precisely , we derived @xmath257^n c_n(\bar\gamma)\label{cncrit}\end{aligned}\ ] ] with @xmath98 and @xmath163 as defined in and . because @xmath63 evolves in a quartic potential , only @xmath258 can be easily calculated . however , and predict the dependence of the cumulants in @xmath259 in terms of the unique parameter @xmath98 : one can easily check that and are consistent with this dependence , for @xmath260 in this paper we have shown that the long - range correlations of the @xmath2 model near the second - order phase transition decay like the @xmath3 power of the system size @xmath1 . in the entire critical regime , these correlations @xcite can be understood from the evolution @xcite of the amplitude of the first fourier mode given by the langevin equation of a particle in a quartic potential . we have also computed the cumulants of the current of particles away from the transition , showing that the become singular at the transition in a way which matches the results of our previous work @xcite where these cumulants were computed in the critical regime . it would be interesting to see whether other diffusive systems , at a phase transition , display correlation functions and current fluctuations with behaviors similar to those we discovered here for the @xmath2 model . deterministic one - dimensional systems , in particular those which conserve momentum , are known to exhibit an anomalous fourier s law @xcite , with cumulants of the current @xcite and correlations @xcite scaling as a non - integer power of the system size . although these systems are much more difficult to study than the @xmath2 model ( for which one only needs to follow the dynamics of a single mode ) , it would be interesting to see whether the approximations that have been used so far , such as the mode - coupling approach @xcite , could predict the power - law dependence of these current and density fluctuations . _ acknowledgments . _ bd acknowledges the support of the french ministry of education through the anr 2010 blan 0108 01 grant . 10 m. r. evans , y. kafri , h. m. koduvely , and d. mukamel , `` phase separation and coarsening in one - dimensional driven diffusive systems : local dynamics leading to long - range hamiltonians , '' _ phys . e _ , vol . 58 , no . 3 , pp . 27642778 , 1998 . l. bertini , a. de sole , d. gabrielli , g. jona - lasinio , and c. landim , `` towards a nonequilibrium thermodynamics : a self - contained macroscopic description of driven diffusive systems , '' _ j. stat . _ , vol . 135 , pp . 857872 , 2009 . l. delfini , s. lepri , r. livi , c. meja - monasterio , and a. politi , `` nonequilibrium dynamics of a stochastic model of anomalous heat transport : numerical analysis , '' _ j. phys . a : math . theor . _ , vol . 43 , p. 145001 , 2010 . a. gerschenfeld , b. derrida , and j. l. lebowitz , `` anomalous fourier s law and long range correlations in a 1d non - momentum conserving mechanical model , '' _ j. stat . _ , vol . 141 , pp . 757766 , 2010 .	non - equilibrium diffusive systems are known to exhibit long - range correlations , which decay like the inverse @xmath0 of the system size @xmath1 in one dimension . here , taking the example of the @xmath2 model , we show that this size dependence becomes anomalous ( the decay becomes a non - integer power of @xmath1 ) when the diffusive system approaches a second - order phase transition . this power - law decay as well as the @xmath1-dependence of the time - time correlations can be understood in terms of the dynamics of the amplitude of the first fourier mode of the particle densities . this amplitude evolves according to a langevin equation in a quartic potential , which was introduced in a previous work to explain the anomalous behavior of the cumulants of the current near this second - order phase transition . here we also compute some of the cumulants away from the transition and show that they become singular as the transition is approached .
in @xcite , the first author presented an algorithm for the rigorous computation of @xmath0-functions associated to automorphic forms . the algorithm is efficient when one desires many values of a single @xmath0-function or values of many @xmath0-functions with a common @xmath4-factor . in this paper , we explore the prototypical case of a family of @xmath0-functions to which that does not apply , namely maass cusp forms in the eigenvalue aspect . as described in @xcite , one of the main challenges when computing @xmath0-functions is the evaluation of the inverse mellin transform of the associated @xmath4-factor . rubinstein @xcite describes an algorithm based on continued fractions that performs well in practice , but for which it seems to be very difficult to obtain rigorous error bounds . on the other hand , the algorithm in @xcite , following dokchitser @xcite , uses a precomputation based on simpler power series expansions that are easy to make rigorous ; it works well for motivic @xmath0-functions of low weight , but suffers from catastrophic precision loss when the shifts in the @xmath4-factor grow large , as is the case for maass forms . a well - known similar problem occurs when one attempts to evaluate an @xmath0-function high up in the critical strip . rubinstein , following an idea of lagarias and odlyzko @xcite , has demonstrated that this can be dealt with effectively by multiplying by an exponential factor to compensate for the decay of the @xmath4-factor ; specifically , for a complete @xmath0-function @xmath5 of degree @xmath6 , one works with @xmath7 for a suitable @xmath8 . this idea can be made to work for general @xmath0-functions , including those associated to maass forms ( albeit with the problems related to precision loss noted above , if the @xmath4-factor is not fixed ) , and molin @xcite has worked out rigorous numerical methods in quite wide generality . for the specific case of maass cusp forms , vishe @xcite ( see also @xcite ) has shown that the `` right '' factor to multiply by to account for the variation in both @xmath2 and the @xmath4-factor shifts is not the exponential function @xmath9 , but rather the hypergeometric function @xmath10 where @xmath11 denotes the parity of the maass form , and @xmath12 is its laplacian eigenvalue . to understand the motivation for this factor , consider first the case of a classical holomorphic cuspform @xmath13 , for which the @xmath0-function is defined via the mellin transform @xmath14 since @xmath13 is holomorphic and vanishes in the cusp , we can change the contour of integration from @xmath15 to @xmath16 for some @xmath17 ; writing @xmath18 for @xmath19 , we obtain @xmath20 thus , rubinstein s exponential factor arises naturally from a contour rotation . for a maass form @xmath13 of weight @xmath21 and even parity ( say ) , we similarly have the integral representation @xmath22 since @xmath13 is no longer holomorphic in this case , we can not simply rotate the contour in this expression , but we are free to start with the rotated integral @xmath23 and try to relate it back to @xmath5 . as we show in [ s : prelim ] , this can be done , and the two are related essentially by the factor . we analyze this strategy in greater detail in [ s : prelim ] , but the upshot is that to compute maass form @xmath0-functions for a wide range of values of @xmath24 and @xmath2 , it suffices to compute @xmath25 for suitable values of @xmath26 and @xmath27 . in turn , using modularity to move each point to the fundamental domain , the problem reduces to computing the @xmath28-bessel function @xmath29 for various @xmath24 and @xmath30 . fortunately , that is a problem that underlies all computational aspects of maass cusp forms and has been well studied ; see , for instance , @xcite . in [ s : results ] , using as input the rigorous numerical maass form data of @xcite we compute values of the corresponding maass form @xmath0-functions and use the resulting numerical data to test conjectures about the distribution of zeros of maass form @xmath0-functions in the @xmath2- and @xmath24-aspects . in particular , we show that the phenomenon of zero repulsion around @xmath31 that strmbergsson observed @xcite disappears in the large eigenvalue limit . we derive rigorous error estimates and use the interval arithmetic package mpfi @xcite throughout our computations to manage round - off errors . thus , modulo bugs in the software or hardware , our numerical results are rigorous . let @xmath32 be a cuspidal maass newform and hecke eigenform of weight @xmath21 and level @xmath33 . then @xmath13 has a fourier expansion of the form @xmath34 where @xmath35 is the @xmath36th hecke eigenvalue of @xmath13 , @xmath12 is its eigenvalue for the hyperbolic laplacian @xmath37 , @xmath38 indicates the parity , and @xmath39 moreover , @xmath13 is related to its dual @xmath40 via the fricke involution , so that @xmath41 for some complex number @xmath42 with @xmath43 . associated to @xmath13 we have the @xmath0-function @xmath44 , defined for @xmath45 by the series @xmath46 it follows from that @xmath44 continues to an entire function and satisfies a functional equation relating its values at @xmath47 and @xmath48 . to see this , let @xmath49 denote the gauss hypergeometric function @xmath50 and consider the family of @xmath4-factors @xmath51 where @xmath52 and @xmath17 is a parameter . by ( * ? ? * sec . 6.699 , eqs . 3 and 4 ) , we have @xmath53 for @xmath54 . ( note that for a maass form with odd reflection symmetry , viz . @xmath55 , has a removable singularity at @xmath56 ; this is related to the fact that the complete @xmath0-function is the mellin transform of @xmath57 rather than @xmath13 . ) making the substitution @xmath58 , we can express the complete @xmath0-function @xmath59 as the mellin transform of the maass form along a ray in the upper half plane : @xmath60 splitting the integral at @xmath61 and employing completes the analytic continuation : @xmath62 using that @xmath63 and making the substitution @xmath64 , we obtain the functional equation : @xmath65 in particular , @xmath66 for @xmath67 . we describe an algorithm based on the fast fourier transform that allows one to evaluate @xmath68 quickly , if one is interested in many points . the integral is essentially a fourier transformation , @xmath69 similarly for the integral , @xmath70 in order to use the fast fourier transform to compute @xmath71 we first need to discretize the integral . to that end , let @xmath72 be parameters such that @xmath73 is an integer . in the poisson summation formula , @xmath74 we solve for @xmath75 , which results in @xmath76 @xmath77 in [ s : bounds ] we will derive precise bounds for this error term . let @xmath78 be the analytic conductor , defined by @xmath79 note that @xmath80 satisfies the recurrence @xmath81 . further , we define @xmath82 so that @xmath83 . @xcite[lem : booker4.1 ] for @xmath47 in the strip @xmath84 , @xmath85 the estimate in lemma [ lem : booker4.1 ] is not optimal since , for @xmath86 for large @xmath2 , the right - hand side grows quadratically in @xmath2 , whereas the convexity estimate would give @xmath87 . moreover , for @xmath88 and @xmath89 the bound becomes useless , since @xmath90 , whereas @xmath91 . nevertheless , the estimate is clean and uniform in all parameters , and suffices for our purposes if we keep away from @xmath89 . [ cor : l ] for @xmath47 in the strip @xmath92 , @xmath93 with @xmath94 recall that @xmath95 satisfies the recurrence , reflection , and duplication formulas @xmath96 hence , for @xmath38 and @xmath67 , @xmath97 which yields @xmath98 for @xmath99 . on the critical line we have @xmath100 and by the phragmn - lindelf theorem , @xmath101 thus , in the strip @xmath102 we have @xmath103 using the kim sarnak estimate @xcite @xmath104 with @xmath105 in the euler product gives @xmath106 inserting the last three bounds in lemma [ lem : booker4.1 ] yields the corollary . [ lem : gamma ] for @xmath107 with @xmath108 and @xmath109 , @xmath110 with @xmath111 for @xmath112 we have the integral representation . since @xmath113 , it is enough to prove the lemma for non - negative @xmath2 . making the change of variables @xmath114 and moving the contour of integration back to the real line , we get @xmath115 and @xmath116 we bound the first integral using the estimates @xmath117 @xcite and @xmath118 , and the second integral using @xmath119 and @xmath120 . [ lem : sum gamma ] for @xmath107 with @xmath108 , @xmath121 , and @xmath109 , @xmath122 using lemma [ lem : gamma ] together with @xmath123 yields the stated bound . [ lem : sum lambda ] for @xmath107 with @xmath124 , @xmath121 , and @xmath109 , @xmath125 by corollary [ cor : l ] and lemma [ lem : gamma ] , @xmath126 applying the estimate @xmath127 and summing up results in the stated bound . [ lem:^lambda ] for @xmath128 , @xmath129 , @xmath130 , @xmath131 , and @xmath132 , @xmath133 applying @xmath134 , @xmath135 , and @xmath136 in the fourier expansion of the maass form gives @xmath137 and by fricke involution @xmath138 for @xmath139 , @xmath140 writing @xmath141 , @xmath142 , @xmath143 , with @xmath144 and @xmath145 , we have @xmath146 summing the resulting geometric series gives @xmath147 for @xmath148 , and similarly for the sum over @xmath149 . for @xmath128 , @xmath150 writing @xmath151 , @xmath152 , @xmath153 , with @xmath144 and @xmath154 , we have @xmath155 and summing up gives @xmath156 [ lem:^gamma ] for @xmath128 , @xmath129 , @xmath130 , @xmath157 , and + @xmath158 , @xmath159 we have @xmath135 and @xmath136 , so that @xmath160 writing @xmath141 , @xmath142 , @xmath143 , with @xmath144 and @xmath145 , we have @xmath161 and summing the geometric series yields @xmath162 for @xmath163 . the argument is slightly different for the sum over @xmath149 . using that @xmath164 @xcite and @xmath136 , we have @xmath165 for @xmath166 . the following two lemmata show that we can circumvent division by zero in @xmath167 . [ lem : psi lemma ] for any @xmath168 with @xmath169 , @xmath170 is an increasing function of @xmath171 . it suffices to show that @xmath172 . suppose first that @xmath173 . then , differentiating binet s formula for @xmath174 , we have @xmath175 for @xmath173 , this is easily seen to be minimum at @xmath176 , so we obtain @xmath177 for @xmath178 we use the reflection formula @xmath179 to see that @xmath180 and apply the above to obtain a bound for @xmath181 . we calculate that @xmath182 and with a little calculus we see that this is bounded in modulus by @xmath183 . thus , altogether we have @xmath184 for all @xmath171 and @xmath185 . note that the right - hand side is strictly increasing for @xmath186 . using the known value @xmath187 , it is straightforward to see that @xmath188 is positive for @xmath169 . [ lem : min gamma ] for @xmath189 , @xmath128 , @xmath190 , @xmath191 , and @xmath192 , @xmath193 we follow the proof of @xcite , generalizing it and making the implied constants explicit . using ( * ? ? ? * sec . 9.132 , eq . 1 ) and writing @xmath194 we get @xmath195 by lemma [ lem : psi lemma ] , for @xmath196 and @xmath197 , @xmath198 next , @xmath199 and for @xmath128 , @xmath200 hence for @xmath201 , @xmath202 where @xmath203 stands for a value of absolute size at most @xmath204 . @xmath189 implies that @xmath197 or @xmath205 . thus @xmath206 adjusting the phase factor @xmath207 in suitably , i.e. taking @xmath208 and @xmath209 with @xmath210 , respectively , completes the proof . we compute values of @xmath0 on a grid , but we are ultimately interested in the zeros , which are not regularly spaced . to zoom in on the zeros , we interpolate @xmath211 with @xmath212 and @xmath213 , respectively . the function @xmath214 has the advantage that it decays rapidly at @xmath215 and is approximately bandlimited , which allows us to use the whittaker shannon sampling theorem @xcite @xmath216 truncating the sum over @xmath217 and bounding the error of truncation , we get an effective interpolation formula for @xmath214 . [ lem : sum lambda truncated ] for @xmath218 , @xmath219 , @xmath220 , and @xmath221 , @xmath222 for @xmath223 , we have @xmath224 , and for @xmath225 , we bound @xmath226 by . hence , @xmath227 [ lem : sum gamma truncated ] for @xmath228 , @xmath219 , @xmath220 , and @xmath221 , @xmath229 the proof is similar to that of the previous lemma , except that for @xmath225 we bound @xmath226 by lemma [ lem : gamma ] . [ lem : int hatlambda ] for @xmath218 , @xmath128 , @xmath230 , @xmath231 , and @xmath221 , @xmath232 by fourier convolution , @xmath233 with @xmath234 , as defined in ( [ lambda]a ) . using and writing @xmath141 gives @xmath235 changing the order of integration , we have @xmath236 let @xmath237 $ ] and set @xmath238 . for @xmath239 , we have @xmath240 and @xmath241 , while for @xmath242 , @xmath243 . moreover , for @xmath239 , @xmath144 , @xmath196 , @xmath244 while for @xmath242 , @xmath245 thus , @xmath246 identifying the integrals with the complementary error function and the incomplete gamma function completes the proof . [ lem : int hatgamma ] for @xmath228 , @xmath128 , @xmath230 , @xmath231 , and @xmath221 , @xmath247 by fourier convolution , @xmath248 with @xmath249 , as in ( [ gamma]a ) . for @xmath239 , we have @xmath164 @xcite , @xmath136 , and @xmath250 . hence , writing @xmath251 , we have @xmath252 for @xmath253 , we use @xmath135 , and @xmath136 to obtain @xmath254 therefore , after interchanging the order of integration , @xmath255 where we have employed @xmath256 for @xmath257 , and @xmath243 otherwise . evaluating the integrals completes the proof . for each maass form @xmath0-function under consideration , we compute rigorously many values on the critical line . for instance , figure [ fig : z ] shows a graph of @xmath258 for the first even maass form @xmath0-function on @xmath259 . graph of @xmath260 for the first even maass form @xmath0-function on @xmath261 . ] [ cor : gamma zeros ] let @xmath262 be the set of zeros on the critical line of the @xmath4-factor . for values of @xmath263 and @xmath264 chosen according to lemma [ lem : min gamma ] , @xmath265 this follows immediately from lemma [ lem : min gamma ] . fixing the value of @xmath26 , say @xmath208 , there is a risk of hitting a zero of @xmath266 ( to within the internal precision ) when evaluating @xmath267 for some specific value of @xmath2 . in all of our computations , we never observed this in practice , i.e. we never had to deal with division by zero . however , computing at finite absolute precision , we sometimes come close to a zero of @xmath266 and experience some loss of precision in the division by @xmath268 . since we also compute for a second value of @xmath26 , @xmath209 , chosen according to lemma [ lem : min gamma ] , we may always ensure the accuracy of the computed values of @xmath269 . for each maass form @xmath0-function under consideration , we rigorously compute all zeros on the critical line up to some height . the search for zeros is faciliated by the following lemma . [ lem : t ] ( a ) let @xmath270 be real valued , and assume it has consecutive simple zeros at @xmath271 , @xmath272 and @xmath273 , with @xmath274 . then @xmath275 such that the following holds : @xmath276 \(b ) let @xmath277 be real valued , and assume it has a zero of order @xmath36 at @xmath271 , with @xmath278 . then @xmath279 such that the following holds : @xmath280 for @xmath281 , but @xmath282 . the lemma follows from elementary analysis and the intermediate value theorem . [ alg : t ] let @xmath283 be real valued . let @xmath284 be a strictly increasing sequence of real numbers . refine the sequence @xmath285 until all zeros of @xmath269 are isolated , i.e. there is at most one zero per interval @xmath286 $ ] . [ rem : t]if the quadrants of @xmath287 for consecutive @xmath288 are not ordered as given in lemma [ lem : t](a ) , there is either a zero of order greater than @xmath289 which is to be investigated according to lemma [ lem : t](b ) , or the sequence is not yet fine enough . we expect the sequence to be fine enough if for successive @xmath288 the quadrants of @xmath287 do change by at most by @xmath289 , and when they change , they do so in agreement with lemma [ lem : t ] . there is no proof that the expectation in remark [ rem : t ] holds , and one can construct sequences @xmath285 that contradict the expectation . nevertheless , with some reasonable choices in the construction of the sequence @xmath285 and its refinements , the expectation turns out to be reliable in practice . namely , for every maass form @xmath0-function that we considered , we never overlooked any zero , as proven after the fact using turing s method . turing s method for verifying the riemann hypothesis for arbitrary @xmath0-functions is described in @xcite . for @xmath2 not the ordinate of some zero or pole of @xmath290 , let @xmath291 by convention , we make @xmath292 upper semicontinuous , i.e. when @xmath2 is the ordinate of a zero or pole , we define @xmath293 . we select a particular branch of @xmath294 by using the principal branch of @xmath295 . with this choice , set @xmath296 we further define @xmath297 which relates to the number of zeros in the critical strip up to height @xmath2 . for @xmath298 let @xmath299 denote the multiset of zeros with imaginary part in @xmath300 $ ] , and let @xmath301 denote their number , counting multiplicity , @xmath302 then , we have @xmath303 @xcite[thm : booker4.6 ] for @xmath304 and @xmath305 , define @xmath306 and @xmath307 suppose @xmath272 and @xmath273 satisfy @xmath308 for some @xmath309 , and set @xmath310 then @xmath311 [ turing method ] for @xmath312 , assume @xmath313 is a given multiset of zeros with imaginary part in @xmath300 $ ] , i.e. @xmath314 . let @xmath315 if @xmath316 exceeds the right - hand side of the bound in theorem [ thm : booker4.6 ] , then the set @xmath317 contains all zeros with imaginary part in @xmath318 $ ] . @xmath319 . if @xmath317 were a proper subset of @xmath320 , then we would have @xmath321 , whence @xmath322 $ ] . but the integral of the latter is bounded by theorem [ thm : booker4.6 ] . we consider consecutive maass cusp forms on @xmath323 . booker , strmbergsson , and venkatesh @xcite have rigorously computed the first @xmath324 maass cusp forms on @xmath261 to high precision . bian @xcite has extended these computations to a larger number of maass cusp forms . the readily available list of rigorously computed maass cusp forms is consecutive for the first @xmath325 maass cusp forms , which covers all maass cusp forms whose laplacian eigenvalue @xmath326 falls into the range @xmath327 . previous numerical computations of some non - trivial zeros for a few even maass form @xmath0-functions were made by strmbergsson @xcite . we extend his results by rigorously computing , for each of the first @xmath325 consecutive maass form @xmath0-functions on @xmath261 , many values of @xmath269 , including all non - trivial zeros up to @xmath328 , at least . at the time of strmbergsson s work , even the numerical data pertaining to the maass cusp forms for @xmath261 was not rigorously proven to be accurate , so he had no reason to carry out his computations of the zeros with more than heuristic estimates for the error . making use of the rigorous data sets described above , we have rigorously verified the correctness of strmbergsson s results . in particular his lists of zeros are consecutive and accurate . moreover , we confirm his observation of a zero - free region on the critical line for @xmath2 near @xmath24 , when @xmath24 is small . we note that some theoretical results , such as cho s theorem @xcite on simple zero of maass form @xmath0-functions , assumed the correctness of strmbergsson s numerical results . with our verification , cho s theorem becomes unconditional . our lists of zeros contain more than @xmath330 consecutive non - trivial zeros per maass form @xmath0-function . all these zeros are simple . the first several zeros of the first five maass form @xmath0-functions are listed in table [ tab : zeros ] . [ thm : finite grh ] for @xmath13 a maass cusp form on @xmath261 with spectral parameter @xmath327 , all non - trivial zeros with @xmath331 of the corresponding maass form @xmath0-function are simple and on the critical line . for each maass form @xmath0-function we prove , using corollary [ turing method ] , that the corresponding list of rigorously computed zeros is consecutive for @xmath331 , and that all the zeros are indeed simple and on the critical line . according to a conjecture of montgomery @xcite , the distribution of non - trivial zeros should follow random matrix theory ( rmt ) predictions . in case of maass form @xmath0-functions , the distribution of non - trivial zeros is expected to conform to that of eigenvalues of large random matrices from the gaussian unitary ensemble ( gue ) @xcite . this raises the question of how gue statistics relate to the zero - free region around @xmath332 observed by strmbergsson @xcite are the gue statistics asymptotically correct in the large @xmath2 aspect only ? we investigate this question by distinguishing between zeros with small and large absolute ordinate , respectively . for a given maass form @xmath0-function there are only a finite number of zeros with small ordinate , and the resulting statistics would be poor . knowing the zeros for many maass form @xmath0-functions , we can evaluate on a common scale the distribution of zeros for each @xmath0-function and collate the statistics of many of them together . let @xmath13 be a maass cusp form with spectral parameter @xmath24 and parity @xmath11 . consider the zeros of the associated maass form @xmath0-function . we unfold the zeros , @xmath333 in order to obtain rescaled zeros @xmath334 with a unit mean density . then @xmath335 defines the sequence of nearest - neighbor spacings , which has mean value @xmath289 as @xmath336 . now , the distribution of nearest - neighbor spacings is given by @xmath337 where the index @xmath13 denotes the corresponding maass cusp form . distributions of rescaled nearest - neighbor spacings are expected to be independent of the specific parameter values of corresponding maass cusp forms @xmath13 and can be collated by writing @xmath338 rescaled nearest - neighbor spacing distributions @xmath339 , @xmath340 , @xmath341 , and @xmath342 , respectively , in comparison with the wigner surmise @xmath343 . only the distribution of zeros that are in absolute size closest to the value of the spectral parameter might show a stronger level repulsion than the wigner surmise . in all other cases , the distribution of zeros closely resembles gue statistics . ] to distinguish between zeros with small and large absolute ordinate , we define the respective nearest - neighbor spacings distributions , @xmath344 where @xmath36 is a non - negative integer , as well as their collated versions @xmath345 for the first @xmath325 maass form @xmath0-functions on @xmath261 , the resulting distributions are displayed in figure [ fig : p ] , in comparison with the wigner surmise @xmath346 as is visible , the distribution of zeros resembles gue statistics for both small and large absolute ordinate , and there appears to be no distinction between the statistics of the two cases . only the distribution of zeros that are in absolute size closest to the value of the spectral parameter might show a stronger level repulsion than the wigner surmize . however , it is unclear whether this seemingly stronger level repulsion is just an artefact of the limited number of @xmath325 spacings that contribute to the histogram of @xmath340 . if we take three times as many spacings into account , as is the case with @xmath341 , we again find a close resemblance to the poisson distribution . we speculate that the gue statistics hold for all @xmath2 , not only in the large @xmath2 aspect . average number @xmath347 of non - trivial zeros for the first even maass form @xmath0-function on @xmath348 . for comparison , the locations of zeros and unfolded zeros are also included as tics . clearly visible is the negative density region ( @xmath349 ) for @xmath2 around @xmath24 . zeros are pulled away from this region resulting in a zero - free region in the @xmath2 aspect , but not in the unfolded zeros . ] since the gue statistics are based on rescaled zeros , @xmath350 , they do not contradict a zero - free region on the critical line . the density of zeros is described by @xmath351 , and according to the @xmath4-factor , the density of zeros is smaller for @xmath2 in a neighborhood of @xmath24 . in particular , for small values of the spectral parameter @xmath24 , the density @xmath351 becomes negative for @xmath2 near @xmath24 ; see figure [ fig : phi ] . there are finitely many maass form @xmath0-functions on @xmath261 that have such a region where @xmath351 is negative . by inspection , we find that no zero falls into a negative density region . moreover , the zeros seem to be repelled away from the negative density regions resulting in the zero - free region around @xmath24 . distributions of the rescaled first zero in dependence of the parity of the maass form @xmath0-function . amongst the @xmath325 maass cusp forms under consideration , @xmath352 of them are even with respect to reflection in the imaginary axis , @xmath88 , and @xmath353 of them are odd , @xmath55 . in comparison with the wigner surmise @xmath343 , close to the origin of the plots , the first zero shows a stronger level repulsion irrespective of the parity . ] finally , we investigate the repulsion from zero of the rescaled first zero @xmath354 in dependence of the parity @xmath11 of the maass form @xmath0-function . for this we consider the distributions of the rescaled first zero , @xmath355 for @xmath356 . the resulting distributions are displayed in figure [ fig : p1 ] . close to the origin of the plots , they show a stronger level repulsion than the wigner surmise @xmath357 . h. l. montgomery , the pair correlation of zeros of the zeta function . analytic number theory ( proc . sympos . pure math . * 24 * , st . louis univ . , st . louis , mo . , 1972 ) , pp . 181193 . soc . , providence , r.i . , 1973 . m. rubinstein , computational methods and experiments in analytic number theory , in _ recent perspectives in random matrix theory and number theory _ , pp . 425506 , london math . lecture note ser . * 322 * , cambridge , cambridge univ . press , 2005 .	let @xmath0 be a degree-@xmath1 @xmath0-function associated to a maass cusp form . we explore an algorithm that evaluates @xmath2 values of @xmath0 on the critical line in time @xmath3 . we use this algorithm to rigorously compute an abundance of consecutive zeros and investigate their distribution .
the study of high-@xmath0 superconductivity is one of the most attractive realms in the last two decades . since the angle - resolved photoemission spectroscopy ( arpes ) directly probes the electronic occupied states , it has become an important technique to investigate the electronic properties of cuprates@xcite . recently , the oxygen isotope effect has been detected with arpes in bi@xmath1sr@xmath1cacu@xmath1o@xmath3 ( bi2212 ) by two groups@xcite , and a common feature is noticed that the spectra are shifted with the @xmath4o/@xmath5o substitution , providing direct evidence for electron - phonon ( @xmath2-ph ) coupling in this material . however , since the first report by gweon _ et al._@xcite , this isotopic band shift has become an controversial issue@xcite , as the observed shift is up to 40 mev , much larger than the isotopic energy change of oxygen phonon , @xmath6 5 mev@xcite . very recently , douglas _ et al._@xcite repeat the experiment , but find the shift is only 2@xmath73 mev . thus it turns out to be an interesting problem whether the large shift observed by gweon _ et al . _ is possible or not in the cuprates . in this paper we shall look into this isotope induced band shift from a theoretical point of view . in the cuo@xmath1 plane of cuprates , as shown in fig . 1 , the electronic transfer is affected by the vibration of oxygen atoms between the initial and final cu sites , resulting in an off - diagonal type @xmath2-ph coupling . in order to have an insight into the isotope effect of bi2212 , we start from a half - filled hamiltonian including the electron - electron ( @xmath2-@xmath2 ) repulsion and the above mentioned off - diagonal @xmath2-ph coupling ( @xmath8 and @xmath9 throughout this paper ) : @xmath10 where @xmath11 ( @xmath12 ) is the creation ( annihilation ) operator of an electron with spin @xmath13 at the cu site @xmath14 on a square lattice ( see in fig . the electrons hop between two nearest neighboring cu sites , denoted by @xmath15 , with a transfer energy @xmath16 . @xmath17 is the strength of coulomb repulsion between two electrons on the same cu site with opposite spins . the oxygen phonon is assumed to be of the einstein type with a frequency @xmath18 and a mass @xmath19 . @xmath20 ( @xmath21 ) is the mass change factor of phonon due to the isotope substitution . in the third term , @xmath22 is the dimensionless coordinate operator of the oxygen phonon locating between the nearest - neighboring cu sites @xmath14 and @xmath23 , and the sum denoted by @xmath24 just means a summation over all the phonon sites in the lattice . conduction plane in cuprates . the copper atom ( white circle ) is on the @xmath14-th site of a simple square lattice . the oxygen atom ( black circle ) is located between the nearest - neighboring cu sites , and @xmath25 denotes its displacement from the equilibrium position . ] in the conduction plane of cuo@xmath1 , the electronic hopping integral @xmath26 can be expanded to the second order terms with respect to the phonon displacements @xmath22 as @xmath27 where @xmath28 is the bare hopping energy and @xmath29 the off - diagonal quadratic @xmath2-ph coupling constant . here we note the linear @xmath2-ph coupling does not occur owing to the lattice symmetry of present model . whereas the inter - site @xmath2-@xmath2 interaction is included in the screened values of @xmath28 and @xmath29 . in this section , we develop a path - integral theory for a model with both @xmath2-@xmath2 and @xmath2-ph interactions . by making use of the trotter s decoupling formula , the boltzmann operator is written as , @xmath30 applying the hubbard - stratonovitch transformation@xcite and the gaussian integral formula@xcite , we decouple the two - body parts , so that the @xmath2-@xmath2 and @xmath2-ph correlated terms are replaced by a two - fold summation over the auxiliary spin and lattice configurations , which is the so - called path - integral . in this way , the boltzmann operator is rewritten into the path - integral form as , @xmath31 \right\ } \prod_l \left [ \|x_q(l,\theta ) \rangle \langle x_q(l,0)\| \right ] \right),\end{aligned}\ ] ] @xmath32 \left[a^{\dag}_{l \sigma}(\tau ) a_{l ' \sigma}(\tau ) + a^{\dag}_{l ' \sigma}(\tau ) a_{l \sigma}(\tau ) \right ] \nonumber\\ & & - \sqrt{u \over \delta } \sum_l x_m(l , \tau ) [ n_{l \uparrow}(\tau ) - n_{l \downarrow}(\tau ) ] , \\ \omega(x_q ) & \equiv & \sum_{\langle l , l ' \rangle } \left\ { { \lambda \over 2 \omega_0 } \left [ { \partial x_q(l , l',\tau ) \over \partial \tau } \right]^2 + { 1 \over 2 } \omega_0 x^2_q(l , l',\tau ) \right\}.\end{aligned}\ ] ] here , @xmath33 and @xmath34 correspond to the auxiliary spin and lattice field , respectively , and @xmath35 symbolically denotes the integrals over the path @xmath36 synthesized by @xmath33 and @xmath34 . @xmath37 is the time interval of the trotter s formula , @xmath38 , and @xmath39 is the absolute temperature . @xmath40 in eq . ( 4 ) is the time ordering operator . then we define the time evolution operator [ @xmath41 as @xmath42.\end{aligned}\ ] ] in terms of the boltzmann operator ( 4 ) and time evolution operator ( 7 ) , we define the free energy [ @xmath43 of the given path as @xmath44.\end{aligned}\ ] ] while , the partition function ( @xmath45 ) and total free energy ( @xmath46 ) are given as @xmath47 according to refs . [ 6 ] and [ 7 ] , we also define the one - body green s function [ @xmath48 on a path @xmath36 as @xmath49 where @xmath50 is the heisenberg representation of @xmath12 . it is really time - dependent and defined by @xmath51 meanwhile , the ordinary green s function [ @xmath52 can be obtained by the path - integral as @xmath53 this path - integral is evaluated by the quantum monte carlo ( qmc ) simulation method . if the qmc data of green s function @xmath54 is obtained , we can immediately calculate its fourier component [ @xmath55 as @xmath56 where @xmath57 is the momentum of the outgoing photo - electron . from this fourier component @xmath58 , we derive the spectral function [ @xmath59 by solving the integral equation @xmath60 finally , the normalized spectral intensity is obtained as , @xmath61 where the fermi - dirac function @xmath62 $ ] is imposed . we now present the qmc results on a 4@xmath634 square lattice , where @xmath28 is set as the unit of energy , and @xmath18=1.0 is used . for the qmc simulation , we impose a little large isotopic mass enhancement , @xmath64=1 and @xmath20=2 , to suppress the numerical error . in this calculation , we determine the binding energy @xmath65 by the moment analysis of the spectral intensity as @xmath66 . correspondingly , the isotope induced band shift is calculated by @xmath67 . with @xmath29 on a 4@xmath634 square lattice , when @xmath17=2.0 , @xmath68=10 , @xmath64=1 , @xmath20=2 . the filled circles are from qmc , and the solid curve from hfa as a guide for eyes . ( b ) the variations of @xmath69 with @xmath17 on a 4@xmath634 square lattice at @xmath68=10 , @xmath64=1 and @xmath20=2 . three different values of @xmath29 are used to show the @xmath29-dependence of @xmath69 . the discrete symbols are the results of qmc , and continuous curves by hfa as a reference . ] in fig . 2(a ) , we plot the ratio @xmath69 versus @xmath29 , at @xmath17=2.0 and @xmath68=10 , where @xmath70 is the band shift at the @xmath71 point of brillouin zone [ @xmath72=(0,0 ) ] , and @xmath73 is the isotopic change of phonon energy . the filled circles are calculated by qmc , and the solid curve by the mean - field theory with hartree - fork approximation ( hfa ) as a guide for eyes . here both theories figure out an increase of @xmath69 with @xmath29 , which means if the @xmath2-ph coupling is strong enough , a large band shift can be generated in the cost of a small @xmath73 . in fig . 2(b ) , the ratio @xmath69 versus @xmath17 are shown for three different @xmath29 s , where the discrete symbols and continuous curves are the qmc and hfa results , respectively . it can be seen that the ratio @xmath69 increases with @xmath29 . meanwhile , for a fixed @xmath29 , the ratio declines slightly as @xmath17 increases . this behavior indicates that the band shift is owing to the @xmath2-ph coupling , whereas the presence of @xmath17 partially reduces this effect . in terms of figs . 2(a ) and 2(b ) , one can see the band shift is actually a measure of the @xmath2-ph coupling strength in the system . if the result of ref . [ 2 ] is correct , the @xmath2-ph coupling must be strong in bi2212 . on the contrary , ref . [ 3 ] shows that the coupling can not be very large . in summary , by using the path - integral qmc method , we study the isotopic shift in the arpes of bi2212 based on a model including both @xmath2-@xmath2 and off - diagonal quadratic @xmath2-ph interactions . our calculation demonstrates that the band shift is primarily triggered by the @xmath2-ph coupling , while the presence of @xmath2-@xmath2 repulsion tends to suppress this effect .	we develop a path - integral theory to study the angle - resolved photoemission spectra ( arpes ) of high-@xmath0 superconductors based on a two - dimensional model for the cuo@xmath1 conduction plane , including both electron - electron ( @xmath2-@xmath2 ) and electron - phonon ( @xmath2-ph ) interactions . comparing our result with the experimental one of bi@xmath1sr@xmath1cacu@xmath1o@xmath3 , we find that the experimentally observed isotopic band shift in arpes is due to the off - diagonal quadratic @xmath2-ph coupling , whereas the presence of @xmath2-@xmath2 repulsion partially suppresses this effect .
identifying thermoelectric materials composed of environmentally friendly and/or plentiful elements is one current thrust of the thermoelectric community . recent successes in this area include several zintl phases , such as ca@xmath0alsb@xmath0 and ca@xmath10al@xmath11sb@xmath12.@xcite simple binaries such as fesb@xmath11 and fesi may be promising for low - temperature applications.@xcite in addition to environmental concerns and abundance issues , low - density materials are also of interest for applications due to their potential for large specific powers , and the associated reduced mass requirements . for instance , mg@xmath11si is an example of a material that possesses all of the above desired properties and good thermoelectric performance.@xcite the higher manganese silicides ( mnsi@xmath13 , @[email protected] ) are another group of low - cost , abundant materials with promising thermoelectric performance.@xcite it is this combination of abundant and low - density materials that brought us to examine another silicide , nisi@xmath0p@xmath1 , of which little is known beyond the crystallographic structure . nisi@xmath0p@xmath1 is isostructural with cu@xmath0sbse@xmath1,@xcite which has recently been shown to possess moderate thermoelectric performance . at the materials level , thermoelectric performance is quantified via the dimensionless figure of merit @xmath9=@xmath15/@xmath16 , where @xmath17 is the seebeck coefficient , @xmath2 the electrical resistivity , and @xmath18 the thermal conductivity . in cu@xmath0sbse@xmath1 , @xmath9 reaches between 0.7 and 0.9 near @xmath3650k , with the exact value depending on the particular alloying and doping employed.@xcite nisi@xmath0p@xmath1 and cu@xmath0sbse@xmath1 are isostructural with the mineral famatinite cu@xmath0sbs@xmath1 , which is a member of the stannite ( cu@xmath11fesns@xmath1 ) group . these materials possess a tetragonal unit cell ( space group @xmath19@xmath202@xmath21 , no . 121 ) based on the zinc - blende structure with all atoms being tetrahedrally coordinated ( though the tetrahedral coordinations are slightly distorted with bond angles deviating from the ideal 109.5@xmath22 ) . in cu@xmath0sbse@xmath1 , cu and sb are tetrahedrally coordinated by se , whereas in nisi@xmath0p@xmath1 , ni and si are surrounded by p. in this work , we characterize nisi@xmath0p@xmath1 , and find it to be a narrow gap semiconductor where holes naturally dominate conduction . this material is too resistive to yield large thermoelectric performance , and thus doping studies were undertaken . at first , we attempted to substitute co for ni , though this did not appear to produce extrinsic holes . thus , in analogy with si and sige alloys , we attempted to use boron as a hole dopant , and nominal compositions nisi@xmath23b@xmath24p@xmath1 were found to produce extrinsic hole concentrations . if boron substitutes for silicon , it does so with a relatively limited solubility , as the maximum hole concentration observed in these samples is approximately 5@xmath510@xmath6@xmath7 at 200k . it is also possible the extrinsic hole concentration is associated with off - stoichiometry of the nisi@xmath0p@xmath1 phase induced by the nominal compositions employed . with some modest success at doping , it was clear that the lattice thermal conductivity severely limits the thermoelectric performance . thus , we attempted to alloy in 20% ge on the si site , and this resulted in only a small decrease in @xmath18 ( primarily at low temperatures ) . the net result is a rather low figure of merit @[email protected] at 700k . further improvements in performance may be achieved with higher doping levels , as the material has a relatively high band mass , though a reduction in the lattice thermal conductivity is also necessary . therefore , this material is unlikely to be pursued in further detail unless theoretical calculations provide insights into doping or suggest anomalous performance for other regions of the electronic structure . polycrystalline samples were prepared by first arc - melting high - purity ni , si , b , and ge as appropriate for the nominal compositions listed in table [ tab1 ] . the resulting ingot was hand ground to a fine powder , and combined with red phosphorus in a silica ampoule that was subsequently sealed under vacuum . the ampoule was heated to 450@xmath22c , held at this temperature for 10h , then heated to 900@xmath22c and held for 100h to allow phosphorus to react through the vapor phase . after cooling , the ampoule was shaken vigorously and heated to 900@xmath22c for another @xmath350 - 100h to ensure full reaction of the phosphorus . during the first heating , a rate of 25@xmath22c / h was employed , and 50@xmath22c / h was used for the second heat treatment . indications of decomposition and significant volatilization of phosphorus were observed when sintering of a cold - pressed pellet at 1000@xmath22c was attempted , and thus @xmath3900@xmath22c was used as a high temperature limit . after heating , the powders were hand ground and consolidated with a hot press . hot pressing was performed in a graphite die within a graphite furnace , using pressures near 10,000psi and a maximum temperature near 900@xmath22c@xmath2550@xmath22c . after exposure to the maximum force / temperature for @xmath31h , the force was removed for a one hour stress - free anneal , followed by cooling to 675@xmath22c over one hour , at which temperature the furnace was turned off to cool more rapidly . this procedure resulted in samples with densities greater than 92% of the crystallographic density . the samples are relatively stable in air , with only a very slight tarnish appearing after several months . grinding was generally done in a helium glove box , though grinding in air prior to hot pressing did not appear to affect the transport properties provided the reaction was complete . the acrid smell of phosphorus was experienced upon cutting or grinding the final specimens . the polycrystalline samples were found , by powder x - ray diffraction , to contain mainly the desired nisi@xmath0p@xmath1 phase . the undoped nisi@xmath0p@xmath1 sample contained a small amount ( @xmath262 wt.% ) of ni@xmath11si . rietveld refinement of the room temperature powder diffraction data using the published structural model@xcite yielded lattice parameters of @xmath27=5.155(1)@xmath28 and @xmath29=10.345(1)@xmath28 . boron substitution resulted in a negligible change in lattice parameters . with increasing nominal concentrations of boron , the impurity phase(s ) changed from ni@xmath11si to bp and nisi , with up to 3% and 6.4% , respectively , observed in sample 4,b in table [ tab1 ] . as discussed below , transport measurements suggest that the boron solubility limit was reached by the nominal composition of nisi@xmath30b@xmath31p@xmath1 . it is also possible that the phase width of nisi@xmath0p@xmath1 is reached at this nominal composition and boron does not incorporate into the lattice . in the sample with nominal composition nisi@xmath11ge@xmath32b@xmath33p@xmath1 , the diffraction revealed about 87% purity of the main phase , with bp ( 7wt.% ) and ge ( 6 wt.% ) as the primary secondary phases . the lattice parameters indicate some ge incorporated into the lattice , though , with refinement yielding @xmath27=5.169(1)@xmath28 and @xmath29=10.367(1)@xmath28 . .nominal compositions and lattice parameters of the nisi@xmath0p@xmath1 and doped - nisi@xmath0p@xmath1 samples presented in this study . [ cols=">,<,^,^ " , ] [ tab1 ] transport data suggested very high hole concentrations are necessary to achieve large thermoelectric performance , and thus we attempted to synthesize samples with large concentrations of boron . the secondary phases seem to have little influence on the transport properties , as similar results were obtained on the different samples . to allow for an appreciation of the variations in properties across batches , we report data on several boron containing samples that possess similar transport properties and thus likely possess similar amounts of boron in the primary phase . p@xmath1 , as well as b- and ge - doped compositions as listed in table [ tab1].,width=288 ] seebeck coefficient @xmath17 , electrical resistivity @xmath2 , and hall coefficient @xmath34 measurements were utilized to investigate the electrical properties ; the hall carrier density is @xmath35=1/@xmath36 . below 300k , these measurements were performed in a quantum design physical property measurement system ( ppms ) . the thermoelectric measurements , including the thermal conductivity @xmath18 , were performed using the ppms thermal transport option ( tto ) , and the hall effect was measured using the resistivity option on the ppms . for tto measurements , gold - plated copper leads were attached to the sample using silver epoxy ( h20e epo - tek ) , and silver paste ( dupont 4929n ) was utilized to obtain low resistance contacts for hall effect measurements performed using 0.025 mm pt wires . hall coefficients were obtained from a fit of the hall resistance versus magnetic field , with maximum fields of @xmath256tesla employed . high temperature thermoelectric performance was assessed using an ulvac zem-3 m8 to measure @xmath17 and @xmath2 . the thermal conductivity @xmath18 ( above room temperature ) was calculated via @xmath37 where @xmath38 is the density . the specific heat capacity @xmath39 was approximated using the dulong petit limit of @xmath40= 3r , calculated according to nominal composition . the differences between @xmath40 and @xmath41 are assumed to be negligible at the temperatures of interest , and this estimation likely results in a slight overestimation of @xmath9 at the highest temperatures . the thermal diffusivity @xmath42 was obtained on @xmath310 mm diameter discs between 1.5 and 2 mm thick , in an anter fl-5000 with a graphite furnace ; data analysis followed astm 1461 for flash diffusivity . no sign of oxidation occurred during the high temperature measurements , and data collected on cooling were consistent with those collected on heating . nisi@xmath0p@xmath1 was found to behave similar to a narrow gap semiconductor with a low concentration of charge carriers . as shown in figure [ fig : tto ] , activated electrical conduction is observed for nisi@xmath0p@xmath1 below room temperature . between 150 and 250k , a simple activation energy was obtained from fitting to @xmath2=@xmath43exp[@xmath44/kt ] , resulting in @xmath44=0.11ev , which is equivalent to a fundamental band gap of 0.22ev . however , the fit to this simple activated behavior becomes poor when wider temperature ranges are considered . consistent with the semiconducting - like behavior , a large seebeck coefficient is observed in nisi@xmath0p@xmath1 , as shown in figure [ fig : tto]b . also consistent with this behavior , the hall carrier density @xmath35 increases with increasing temperature for nisi@xmath0p@xmath1 , as shown in figure [ fig : hall]a . we emphasize that this is a single band @xmath35 , and may not reflect the true carrier density if electrons are also contributing to electrical transport . in order to probe the potential for thermoelectric performance , we attempted to hole - dope nisi@xmath0p@xmath1 . we first tried the substitution of co for ni , but this sample had a higher resistivity than undoped nisi@xmath0p@xmath1 at 300k , and also displayed activated conduction below room temperature . we then found that boron successfully increased the hole concentration in samples with nominal compositions of nisi@xmath23b@xmath24p@xmath1 . however , synthesis with @xmath14=0.02 , 0.10 , 0.20 , 0.40 all yielded similar transport properties , as observed in figure [ fig : tto ] . one explanation for this trend is that the solubility limit of boron has been reached . hole - doped samples of nisi@xmath23b@xmath24p@xmath1 behave metallic above @xmath3 150k with @xmath2 increasing with increasing @xmath45 . similarly , the seebeck coefficients of these samples increase with increasing @xmath45 , consistent with extrinsically doped samples . the seebeck coefficients remain rather large in these doped samples , with @xmath17@xmath3250@xmath4v / k obtained at 300k . the hall carrier concentration and mobility @xmath46 of boron- and boron / germanium - doped nisi@xmath47p@xmath1 are shown in figure [ fig : hall ] . as observed , the values of @xmath35 are relatively temperature independent , and fairly similar for the different nominal boron concentrations . this is consistent with similar @xmath2 and @xmath17 in the various samples . this result suggests that , if boron substitutes for silicon , the solubility limit of boron has been reached . if lattice - incorporated boron atoms are fully ionized and contribute one hole to the system , then a solubility limit of @xmath30.2%b per silicon is obtained from the hall carrier concentration of approximately 5@xmath510@xmath6holes/@xmath48 at 200k . this seems reasonable considering the solubility limits of boron in silicon , which are @xmath3 2@xmath510@xmath6b/@xmath48 at 700@xmath22c and @xmath3 1@xmath510@xmath49b/@xmath48 at 900@xmath22c.@xcite the small increase of @xmath35 above @xmath3200k may be the result of some carriers being activated from other impurities or across the band gap . the hole mobility of boron - doped nisi@xmath0p@xmath1 is found to be approximately 12@xmath8/v / s at 300k . the mobility increases with decreasing temperature , consistent with a carrier mean free path limited by phonon scattering . however , this behavior only occurs down to @xmath3125k , where the mobility reaches a maximum of approximately 18@xmath8/v / s . while @xmath46 decreases upon cooling below 125k , @xmath35 remains relatively constant , which may suggest that additional carrier scattering mechanisms are important at low @xmath45 . ionized impurity scattering generally leads to a mobility that increases as @xmath50 , though the low - t rise in @xmath46 occurs more slowly than @xmath50 in boron - doped nisi@xmath0p@xmath1 . it is also possible that anderson localization occurs due to disorder in the lattice , causing a mobility gap that leads to an increase in @xmath2 at low-@xmath45 . also , we note that the high - temperature decay occurs more slowly than the @xmath51 behavior expected for a lightly doped semiconductor ( or even the @xmath52 behavior expected for a degenerate semiconductor ) . therefore , it is possible impurity band conduction is important in these materials or grain boundary scattering influences transport . p@xmath1 , as well as b- and ge - doped compositions as listed in table [ tab1].,width=288 ] the hole mobility of boron - doped nisi@xmath0p@xmath1 is less than that of isostructural cu@xmath0sbse@xmath1 . the mobility of undoped cu@xmath0sbse@xmath1 is @xmath341@xmath8/v / s at 300k , with a hole concentration of 7.5@xmath510@xmath53@xmath7.@xcite the mobility of undoped nisi@xmath0p@xmath1 is not shown in figure [ fig : hall ] because the single band model likely fails for the undoped composition . at a more comparable hole concentration of 6.3@xmath510@xmath6@xmath7 in cu@xmath0sb@xmath54sn@xmath55se@xmath1 the mobility is approximately 35@xmath8/v / s , and it is 31@xmath8/v / s in cu@xmath0sb@xmath56sn@xmath57se@xmath1 at 300k.@xcite given that the carrier mobility in doped nisi@xmath0p@xmath1 is dominated by phonon scattering at room temperature , it seems unlikely that the boron dopant is leading to this large difference in carrier mobility . therefore , the hole mobility in nisi@xmath0p@xmath1 appears to be inherently lower than in the copper - based analogues . to better understand the differences between transport in nisi@xmath0p@xmath1 and cu@xmath0sbse@xmath1 , we consider a parabolic band analysis of the hall and seebeck coefficients ( equations [ eqn : alpha][eqn : f ] ) . analysis of the hall and seebeck coefficient data allows the carrier mass @xmath58 to be obtained . when phonon scattering limits the relaxation time of charge carriers , the mobility is expected to scale as @xmath4@xmath59@xmath60 . therefore , even small changes in the band mass can have significant influence on carrier transport . within a single parabolic band model , the seebeck coefficient is a function of the reduced electrochemical potential @xmath61 and scattering parameter @xmath62 : @xmath63 the scattering parameter @xmath62 is used to incorporate the energy dependence of the carrier relaxation time , @xmath64 . the value of @xmath62 influences @xmath17 , but @xmath65 does not . in this analysis , we assume acoustic phonon scattering limits @xmath66 , corresponding to @xmath62=0 . the hall carrier density , @xmath35=1/@xmath67e where @xmath67 is the hall coefficient , is given by @xmath68 analysis of @xmath35 allows the effective mass @xmath58 to be determined for a given @xmath61 obtained from analysis of the seebeck coefficient . carrier scattering influences @xmath35 through the hall factor @xmath69 , which is a function of @xmath61 and @xmath62 similar to the seebeck coefficient : @xmath70 for simplification purposes , these expressions utilize the fermi integrals @xmath71 ) defined by @xmath72}. \label{eqn : f}\ ] ] p@xmath1 , as well as b- and ge - doped compositions as listed in table [ tab1].,width=288 ] employing the above equations , we find that the effective mass is approximately [email protected]@xmath73 in boron - doped nisi@xmath0p@xmath1 at room temperature ; @xmath73 is the free electron mass . at 200k , where @xmath35 is still relatively independent of temperature , this process yields 2.9 - 3.7@xmath73 for the various samples of boron - doped nisi@xmath0p@xmath1 . from literature values for @xmath35 and @xmath17 in cu@xmath0sbse@xmath1 ( corresponding to the @xmath4 values noted above),@xcite we obtain @xmath74 depending on the particular sample . therefore , it is clear that the band mass is much larger in nisi@xmath0p@xmath1 than in cu@xmath0sbse@xmath1 , and this likely contributes to the lower mobility . this large band mass results in a relatively large seebeck coefficient in nisi@xmath0p@xmath1 , but the impact on carrier mobility is detrimental to the thermoelectric performance of nisi@xmath0p@xmath1 . the thermal conductivity of nisi@xmath0p@xmath1 is approximately 10w / m / k at 300k . boron doping reduces the low - temperature value of @xmath18 slightly . in an attempt to further reduce @xmath18 , we substituted ge for si , in addition to the boron doping . this results in some additional suppression of @xmath18 at low-@xmath45 , though little change is observed at higher temperatures . interestingly , @xmath18 does not display much temperature dependence below room temperature , even in undoped nisi@xmath0p@xmath1 . this suggests there may be some disorder in the system , or perhaps grain boundaries or secondary phases are reducing @xmath18 at low-@xmath45 . we note that due to the large values of @xmath2 , the lattice thermal conductivity @xmath75 dominates @xmath18 for all samples below @xmath3600k . the high temperature transport properties are plotted in figure [ fig : zem ] . interestingly , the decrease in @xmath2 with increasing @xmath45 for undoped nisi@xmath0p@xmath1 turns into a nearly temperature independent @xmath2 above @xmath3400k , and the corresponding seebeck coefficient is large and relatively temperature independent . the boron - doped samples display typical doped semiconductor behavior at high temperatures , with the resistivity and seebeck coefficients increasing with increasing temperature until the effects of minority carrier activation are observed near 650k . the thermal conductivity decreases with increasing @xmath45 in all samples , consistent with the dominant role of phonon - phonon scattering in these crystalline materials . as noted above , the substitution of 20% ge ( nominal composition ) did not result in a significant reduction in @xmath18 above 300k . for comparison , the thermal conductivity of cu@xmath0sbs@xmath1 reaches approximately 1.5w / m / k at 600k.@xcite p@xmath1 , as well as b- and ge - doped compositions as listed in table [ tab1].,width=288 ] the maximum in @xmath76 , which is relatively unpronounced in these materials , can be utilized to estimate the thermal band gap by @xmath77=2@xmath78.@xcite this analysis suggests that the band gap is between 0.43 and 0.50ev , which is about two times larger than that obtained from the analysis of @xmath79 at lower temperatures . it is possible that one of these estimates of @xmath77 is influenced by the activation of carriers from an impurity band . without high temperature hall effect data the source for this discrepancy is difficult to isolate . the thermoelectric power factor , @xmath80 is plotted in figure [ fig : zt ] along with the figure of merit @xmath9 . relatively low thermoelectric performance is observed in these materials . this is due in part to the relatively high thermal conductivity for the corresponding mobility values . boron - doping greatly improves both the power factor and @xmath9 , and additional improvements would likely come with higher doping levels . this is particularly the case for @xmath81-type nisi@xmath0p@xmath1 , because it is a large band mass material and the optimum carrier density increases with increasing band mass and increasing lattice thermal conductivity.@xcite as noted , however , boron substitution only permits modest carrier densities to be reached , and using a transition metal dopant was not successful in increasing the hole concentration . it is possible that altering the si / p ratio would permit higher doping levels to be achieved , and may also allow @xmath82-type behavior to be probed . however , it seems unlikely that this material will achieve the @xmath9 values observed in isostructural copper compounds , which have higher mobility due to lower band masses , as well as lower lattice thermal conductivities . polycrystalline nisi@xmath0p@xmath1 was found to be a narrow gap semiconductor , with an activation energy for the electrical resistivity on the order of 100mev below room temperature . boron - doping is successful in producing extrinsic holes , though the corresponding hole concentration is limited to approximately 5@xmath510@xmath6@xmath7 , suggesting a possible solubility limit of boron in nisi@xmath0p@xmath1 . the transport properties of the boron - doped samples behave as expected for moderately doped semiconductors , and parabolic band analysis suggests a fairly high band mass of @xmath33@xmath73 . the thermoelectric figure of merit only reaches @xmath30.1 at 700k , due in large part to a lower than desired hole mobility and carrier concentration , and a moderately large thermal conductivity . further improvements in thermoelectric performance may be possible with a different doping scheme or through additional phonon scattering , but are unlikely to yield large thermoelectric efficiencies . this research was supported by the u. s. department of energy , office of basic energy sciences , materials sciences and engineering division ( a.f.m . , m.a.m ) , and the us department of energy , eere , vehicle technologies , propulsion materials program ( h.w . ) .	the hall and seebeck coefficients , electrical resistivity and thermal conductivity of polycrystalline nisi@xmath0p@xmath1 were characterized from 2 to 775k . undoped nisi@xmath0p@xmath1 behaves like a narrow gap semiconductor , with activated electrical resistivity @xmath2 below room temperature and a large seebeck coefficient of @xmath3 400@xmath4v / k at 300k . attempts to substitute boron for silicon resulted in the production of extrinsic holes , yielding moderately - doped semiconductor behavior with @xmath2 increasing with increasing temperature above @xmath3 150k . hall carrier densities are limited to approximately 5@xmath510@xmath6@xmath7 at 200k , which would suggest the solubility limit of boron is reached if boron is indeed incorporated into the lattice . these extrinsic samples have a hall mobility of @xmath312@xmath8/v / s at 300k , and a parabolic band equivalent effective mass of @xmath33.5 times the free electron mass . at 700k , the thermoelectric figure of merit @xmath9 reaches @xmath30.1 . further improvements in thermoelectric performance would require reaching higher carrier densities , as well as a mechanism to further reduce the lattice thermal conductivity , which is @xmath3 5w / m / k at 700k . alloying in ge results in a slight reduction of the thermal conductivity at low temperatures , with little influence observed at higher temperatures . [ [ section ] ] copyright ( 2013 ) american institute of physics . this article may be downloaded for personal use only . any other use requires prior permission of the author and the american institute of physics . the following article appeared in ( j. appl . phys . 113 , 103707 ( 2013 ) ) and may be found at ( http://dx.doi.org/10.1063/1.4794992[this link ] )
the theoretical description of the observed universe acceleration @xcite is one of the challenges of current research . the simplest way to explain this remarkable behavior ( apart form the sole cosmological constant which leads to the corresponding problem ) is to construct various `` field '' models of dark energy , using a canonical scalar field ( quintessence ) @xcite , a phantom field , that is a scalar field with a negative sign of the kinetic term @xcite , or the combination of quintessence and phantom in a unified model named quintom @xcite . however , the arbitrary consideration of additional scalar fields ( which may even have non - conventional kinetic terms inserted by hand ) should be constrained by the fact these extra scalars to be neutral under all the standard model symmetries , and thus not introducing additional fifth forces . this non - trivial requirement led many authors to the alternative direction of modifying gravity itself @xcite , with a promising attempt in these lines being perhaps the recent developments in hoava - lifshitz gravity @xcite , a power - counting renormalizable , ultra - violet ( uv ) complete gravitational theory ( although there may well be problems with the theory due to additional degrees of freedom becoming strongly coupled in the infra - red ) . on the other hand , constructions arising from string theory are hard to result in dark energy phenomenology consistent with observations , purely using the closed string sector . for more details see the review @xcite . however , cosmological dynamics driven by the open string sector through dynamical @xmath2-branes , which is the basic idea of the so - called dirac born infeld ( dbi ) formalism , has led to interesting successes , mainly in inflationary paradigms @xcite . after inflation the universe lives on branes that wrap various cycles within the compact space , and in this sense the gut or electro - weak phase transition can be manifested through a geometric fashion . thus , dark energy does present a dynamical nature , retaining additionally a form of geometric origin . quantitatively , the tight constraints from wmap @xmath3 year data set @xcite on the model parameters have led dbi - models to more complex versions , including multiple fields @xcite , multiple branes @xcite , wrapped branes @xcite or monodromies @xcite . finally , the phase - space analysis of a solitary @xmath4-brane moving through a particular warped compactification of type iib was done @xcite , while the generalization to multiple and partially wrapped branes has been performed in @xcite . in the present work we are interested in constructing a dbi - scenario based on ghost @xmath0-branes , that is @xmath2-branes that have a @xmath5 symmetry acting to flip the signs of the @xmath6-@xmath6 and @xmath7 sectors . such a consideration is more robust than the naive and ambiguous use of the prototypical non - bps @xmath4-brane action , albeit with the wrong sign kinetic term . in addition , although in a typical flux compactification of type ii string theory down to four - dimensions one must introduce negative tension objects called orientifolds ( in order to cancel the @xmath4-brane charge associated with the closed string fluxes , and to project out various string states breaking half of the bulk supersymmetry so that the vacuum retains an @xmath8 structure ) , the existence of ghost branes could possibly negate the need for orientifolds . as we show , in such a ghost @xmath0-brane scenario we can naturally acquire an effective dark energy behaving either as quintessence or as phantom . the plan of the work is as follows : in section [ model ] we present the formalism of @xmath2-branes , used in dark energy scenarios . in section [ ghostd ] we extend it , introducing the concept of ghost d - branes , and we extract the dark - energy equation - of - state parameter . in section [ cosmimpl ] we investigate its general features , examining the conditions for the appearance of quintessence or phantom behavior , while in section [ staticde ] we perform an explicit phase - space analysis of the static , dark - energy dominated , solution sub - class . finally , our results are summarized in section [ conclusions ] . the open string sector of type ii string theory is usually governed by the dbi action , governing the low energy fluctuations of such strings attached to a @xmath2-brane . for @xmath9 coincident branes , the world - volume symmetry is enhanced from @xmath10 to @xmath11 , and the scalar fluctuations are then promoted to matrices obeying a lie structure . this is similar to the induced non - commutative world - volume theory on a single @xmath2-brane when we turn on a non - trivial @xmath12-field . string scattering calculations indicate that there is a particular trace prescription required in order to account for the full string cross - section , which is given by the symmetrized average over all possible orderings of the lie - algebra valued objects @xcite . much like the single @xmath2-brane action , one can sum the relevant terms into non - linear form ( although only valid up to @xmath13 ) to partially reconstruct the non - abelian theory using the effective action @xcite : @xmath14)}\sqrt{\rm{det}q^i_j}\right ) \pm \mu_p \int \rm{str } \mathcal{p}[e^{i\lambda i_{\phi}i_{\phi}}\sum c^{(n ) } e^{b}]e^{\lambda f},\ ] ] where @xmath15 e_{kj}.\end{aligned}\ ] ] in the expressions above , @xmath16 is the inverse of the @xmath17-string tension , and @xmath18 is the square of the string - length - the fundamental length scale in our theory . the scalar @xmath19 is related the space - time embedding @xmath20 and finally @xmath21 is the @xmath6-@xmath6 gauge potential , which we will ignore in the following . moreover @xmath22 denotes the pullback operator acting on the bulk space - time tensor fields , and @xmath19 are the scalar field fluctuations where @xmath23 . in the @xmath7-sector we see the introduction of the interior derivative @xmath24 , whose action on an @xmath25-form is @xmath26c_{ji}^{(n)}.\ ] ] the presence of the interior rather than exterior derivative allows the @xmath2-brane to couple to gauge potentials of higher order , such as the @xmath27 forms . this suggests that there is a transmutation ( or dielectric ) effect where the @xmath2-brane can blow up into a @xmath28-brane through higher order terms in the expansion of the chern - simons action @xcite . a concrete example of this effect is when @xmath9 @xmath4-branes blow up into a solitary @xmath29-brane via the formation of a fuzzy @xmath30 , more commonly referred to as the myers effect . if the scalar fields transform under an appropriate representation of a higher dimensional gauge group , then the @xmath4-branes can be polarized into higher dimensional branes in an analogous fashion through the extended myers effect @xcite . for example , if the scalars transform under irreducible representations of the @xmath25-fold tensor product of @xmath31 , then the branes orient themselves along a fuzzy @xmath32 to form a configuration of @xmath25 @xmath33-branes @xcite . the construction of odd - dimensional fuzzy sphere solutions is actually non - trivial and requires the introduction of spinorial representations @xcite of @xmath34 , where @xmath35 . one important simplification to the above action is when we consider the large-@xmath9 limit , as the str operation reduces to a trace ( up to @xmath36 corrections ) . the reason why this limit is important can be understood when one considers the dual description of the brane configuration . recall that the myers effect describes lower - dimensional branes being polarized into higher - dimensional configurations via a fuzzy sphere . this means that there is a dual description of the myers effect in terms of a higher - dimensional ( spherical ) brane with world - volume flux . more concretely we see that that ( assuming the scalars lying in irreducible representations of @xmath37 ) @xmath38-branes are dual to a single @xmath29-brane wrapped on @xmath39 with @xmath9 units of flux through the @xmath30 @xcite . duality in this sense actually means that the effective actions are identical , provided that the @xmath10 flux on the @xmath29-brane is large . the above statements are all assumed to be true in a curved background , although the required string scattering calculations are difficult to be computed to the necessary order and therefore a direct check is not possible . however given the prevalence of such dualities in string theory , one can be reasonably confident that the statement is correct . the most general cosmological backgrounds in type ii string theory can be written in the following form @xcite @xmath40 where @xmath41 is the warp factor , which is a function of @xmath42 - a warped throat that is fibred over some five - dimensional manifold @xmath43 , and the four - dimensional metric takes the usual frw - form . for concreteness we will specialize to the case of type iib string theory , where the throat can be generated by threading @xmath4-brane flux through a compact three - cycle . moreover , since the dilaton is constant in these backgrounds , the einstein frame and string frames coincide . we will also assume that our theory consists of @xmath38-branes which are oriented parallel to the @xmath44-large dimensions and that the scalars are homogeneous , transforming under irreducible representations of @xmath45 . the resulting action for @xmath9 coincident @xmath4-branes can be written as @xcite - branes we would require the second term to take the positive sign . ] with @xmath46 the warped , positive - definite brane tension . the radius of the fuzzy sphere is defined in terms of the geometric radius @xmath42 via @xmath47 and @xmath48 is the quadratic casimir of @xmath49 , namely @xmath50 . we have also included a scalar potential contribution arising from the interaction of the @xmath4-branes with the closed string background . we remind the reader that at large @xmath9 this action is precisely the same as that arising from a single wrapped @xmath29-brane with @xmath9 units of @xmath10 flux . it is convenient to use the field redefinition @xmath51 , with @xmath42 the induced world - volume scalar coming from the background in the string frame @xcite . however , it has ( mass ) dimension @xmath52 and thus since we desire to write all the fields with canonical mass terms we have to redefine the involved world - volume scalars . following these lines , the action ( [ action0 ] ) can be re - written in the generalized form @xmath53,\ ] ] where @xmath54^{-1/2}$ ] is the usual generalization of the relativistic factor . the cosmological consequences of such an action have been discussed elsewhere @xcite and we refer the interested reader there for more details . the appearance of the positive - definite function @xmath55 , which generalizes the aforementioned action comparing to the usual @xmath56 case , can be theoretically justified @xcite , since if @xmath9 multiple coincident branes are present then the world - volume field theory is a @xmath11 non - abelian gauge theory and this `` potential '' term is simply a reflection of the additional degrees of freedom @xcite . additionally , this configuration is related to a @xmath29-brane , wrapping a two - cycle within the compact space and carrying a non - zero magnetic flux along this cycle . on the other hand , the positive - definite effective potential @xmath57 accounts for the possible open or closed string interactions . its precise form depends upon the number of additional branes and geometric moduli , the number of non - trivial cycles in the compact space , the choice of embedding for branes on these cycles , the coupling of the brane to any background rr - form fields , the contribution from higher dimensional bulk forms @xcite etc . finally , note that using the above generalized form for the action , allows us to interpolate between a single @xmath4-brane ( taking @xmath58 ) and the multi - brane , or wrapped @xmath29-brane ( where @xmath59 ) solutions . it is well established that @xmath2-branes are not the only hypersurfaces within string theory . there are also orientifold @xmath60-planes which have negative tension and reduced charge ( compared to the @xmath2-branes ) @xcite . their role is vital in flux compactifications of type ii string theory , since they cancel global flux tadpoles and also break one half of the residual supersymmetries . there exists another type of extended object , which has been dubbed a ghost - brane @xcite , that we will briefly describe using the boundary state formalism , which is the most appropriate for the cft description of @xmath2-branes . the bosonic sector of a bps @xmath2-brane is represented by a boundary state of the form @xmath61 where @xmath62 represents the full @xmath2-brane state . it was shown in @xcite that one can define an analogous ( bps ) ghost - brane state ( which we will denote by a @xmath1-brane ) through the introduction of an operator @xmath63 : @xmath64 such that the ghost state precisely cancels the combined bps sectors of the @xmath2-brane . since the @xmath1-brane state preserves the overall relative sign of the two different sectors , it must also preserve the same supersymmetries as the @xmath2-brane . this makes it a distinct object , and it should not be confused with the @xmath65-brane - which has a boundary state of the form @xmath66 the formalism implies that the @xmath1-sector exactly cancels the @xmath2-sector . this means that a theory consisting of @xmath9 coincident @xmath2-branes and @xmath67 coincident @xmath1-branes can be described in two equivalent ways ; either by @xmath68 @xmath2-branes or by @xmath69 @xmath1-branes . the corresponding world - volume theory is given by a @xmath70 gauge theory , which is enhanced to @xmath71 as the two groups of branes are brought together . importantly , this means that when @xmath72 the resulting solution is simply space - time with no @xmath2-branes . in this way we see that the ghost brane can screen the @xmath2-brane , and a useful consequence of this screening was employed in ads / cft framework in @xcite . since the @xmath1-brane is simply minus the standard @xmath2-brane state , one sees that the effective world - volume theory for the @xmath1-brane is also of dbi form , albeit with an additional sign change in the definition of the tension . thus , for multiple coincident @xmath73-branes , we expect the effective theory to be well described by the action @xmath74,\ ] ] where we have embedded the branes in the warped background ( [ eq : metric ] ) . note that in the non - relativistic expansion of this action , the kinetic term will have the wrong sign , implying phantom - like behavior for the scalar fluctuations . this suggests that the world - volume theory tends to anti - gravitate , rather than gravitate . we stress that the @xmath73-brane theory is different from previously proposed phantom models based on the dbi - action @xcite , which have been constructed in the light of the non - bps action proposed by sen as an effective description of tachyon condensation @xcite . the models in this class have the wrong sign kinetic term _ inside _ the square - root structure , in contrast to that of the ghost action . furthermore , that sign - change can only be inserted by hand @xcite , since the world - volume metric is induced from the background geometry , and it is unlikely to contain a sub - manifold where the sign changes in a continuous fashion . additionally , the non - bps action can only couple to any of the bulk @xmath7-form fields through terms of the form @xmath75 , which are typically zero according to our assumptions . therefore , it seems unlikely that such boundary states can be stable within the full theory . on the other hand , the ghost branes are supersymmetric and do couple to the bulk form - fields , suggesting that they constitute actually stable states within the theory . let us now focus on the cosmological consequences of the scenario at hand . assuming that the scalar is time - dependent , one reads off the diagonal components of the energy momentum - tensor in the usual fashion : @xmath76 + v(\phi)\\ \label{pphi } & & p_{\phi } = t(\phi)[w(\phi)\gamma^{-1}-1]-v(\phi).\end{aligned}\ ] ] thus , since in dbi - constructions @xmath77 is responsible for dark energy , we can define its equation - of - state parameter as : @xmath78-v(\phi ) } { t(\phi)[1-w(\phi)\gamma ] + v(\phi)}.\ ] ] as can be deduced from expression ( [ wphi ] ) , the dark - energy equation - of - state parameter can present quintessence or phantom behavior , depending on the choice of scalar potential and background . the friedmann equations arising from action ( [ action ] ) write : @xmath79\nonumber\\ & = & -\frac{1}{2 m_p^2 } \left[\rho_m+p_m -\gamma w(\phi ) \dot{\phi}^2 \right],\end{aligned}\ ] ] with @xmath80 the hubble parameter . additionally , variation of the action ( [ action ] ) with respect to @xmath77 leads to the equation of motion for the scalar field , namely : @xmath81 = 0,\\end{aligned}\ ] ] where the subscript @xmath82 denotes differentiation with respect to @xmath77 . this equation is the generalization of the klein - gordon one in the dbi framework , and using ( [ rhophi]),([pphi ] ) it can be written in the usual form @xmath83 . finally , the corresponding equation of motion for matter writes @xmath84 . in the previous section we introduced the concept of ghost d - branes , and we extracted the dark - energy equation - of - state parameter of this ghost version of dbi scenario . thus , we can now investigate the various cosmological possibilities , trying to remain sufficiently general . we mention that we desire to explore the general features of @xmath85 for possible forms of the involved tension and potentials , without examining in detail the equations of motion . as we see , although a full dynamical investigation would be interesting , this basic `` kinematical '' study is sufficient to qualitatively reveal the novel features of the ghost @xmath0-brane model . let us first consider the scenario where no scalar potential is present , that is study solely the brane action . in this case the equation of state reduces to @xmath86.\ ] ] for a general @xmath55 , in the regime where @xmath87 we see that the equation of state is typically zero , unless there are divergences in @xmath88 , which is not the case if we desire our model to be physical . on the other hand , in regions where @xmath55 is dominant we find that @xmath89 and therefore it is negative - definite ( although small ) . similarly , if @xmath90 then @xmath91 which is positive - definite although typically small . note that physical solutions imply @xmath92 , however if we treat the action phenomenologically and assume smaller values for @xmath55 then we find solutions where @xmath93 from above after being initially large . in conclusion , we observe that the possible solution space is quite large even without a scalar potential . this preliminary phenomenology suggests that @xmath94 could cross the @xmath52-bound . in particular , the equation of state would become phantom if @xmath95}{[1-w(\phi)\gamma ] } < 0.\ ] ] however , this condition can not be met physically , and thus we conclude that the brane action alone can not generate phantom dynamics . let us now turn on the scalar potential term @xmath57 . a first simple solution subclass would be to consider @xmath96 , where we obtain @xmath97 recovering the case of pure de - sitter expansion . in the general case of non - zero potential and tension terms , but with @xmath98 , we can expand ( [ wphi ] ) in taylor series acquiring : @xmath99 neglecting higher order terms . therefore , in the relativistic regime ( @xmath100 ) the correction term will be negative - definite , leading to the realization of the phantom phase . we mention that this phantom realization is obtained naturally from a large solution subclass of the model . additionally , it is not the only combination of possibilities which lead to phantom behavior , but just a simple example . these features reveal that the use of ghost d - branes does lead to quintessence and phantom realization , depending on the specific forms of the potential - like terms and of the tension in the effective action . another class of solutions will occur when we have @xmath101 , which upon performing the taylor expansion leads to @xmath102}\left\{1+\frac{v(\phi)}{t(\phi)}\frac{w(\phi)(\gamma^2 - 1)}{[\gamma w(\phi)-1][\gamma - w(\phi)]}\right\}\ ] ] at leading order , and therefore it is highly dependent on the particular background field - parametrization . for initially static configurations ( @xmath103 , i.e @xmath104 ) we recover the usual result @xmath105 , and therefore the static brane mimics the cosmological constant . as the velocity of the brane increases we again find that @xmath93 along the asymptotic branch . on the other hand , if @xmath106 then the equation of state tends to @xmath107 and therefore will relax to zero from below . finally , imposing @xmath108 , that is considering the single brane case , the resulting equation - of - state parameter tends to zero from above as the velocity term increases , as can be seen from ( [ wquint ] ) . note that since @xmath109 , all the cases of the regime @xmath110 present a quintessence behavior with @xmath111 . as we have mentioned , in the present work we are interested in exploring the general qualitative features of the equation - of - state parameter of ghost @xmath0-brane scenario . we have not extracted the equations of motion , studying just @xmath85 as a function of @xmath112 , @xmath55 , @xmath57 and @xmath113 , which is itself a function of @xmath77 and @xmath114 . therefore , for given @xmath112 , @xmath55 , @xmath57 , the dependence of @xmath85 on @xmath113 provides qualitative information about the phase - space structure . a first observation is that ( [ wphi ] ) possesses a singularity at @xmath115.\ ] ] according to the specific choice of @xmath112 , @xmath55 , @xmath57 and of initial conditions , a particular universe evolution ( i.e a particular orbit of @xmath116 in the @xmath117-plane ) can remain either to one or to the other regime , tend asymptotically into the singularity , or even cross it . such singularities are common in field dark energy models , especially in phantom ones , and they correspond to big rip @xcite or to realization of a cosmological bounce @xcite . finally , note that if @xmath118 turns out to be less than @xmath119 , that is unphysical , then the specific model is free of such behaviors , independently of the initial conditions . in order to provide a more transparent picture of the obtained cosmological behavior , in fig . [ fig1 ] we present the solution space for the simple scenario of fixed @xmath120 , imposing @xmath108 ( corresponding to the single brane model ) . as we observe , for @xmath121 there is a singularity at @xmath122 , thus in almost the whole phase - space the equation - of - state parameter is quintessence - like and in particular it is positive - definite and tends to zero asymptotically . for larger values of @xmath120 ( i.e when the scalar potential effect is enhanced comparing to that of the tension ) we observe the singularity at a specific @xmath118 , which in this special subclass ( fixed @xmath120 and @xmath77-independent @xmath88 ) is constant , i.e @xmath77-independent . for values of @xmath123 phantom behavior is realized , while for @xmath124 one finds quintessence - like evolution with @xmath93 from above . let us now consider the same subclass of fixed @xmath120 , but setting @xmath125 . this scenario can be obtained in a class of string theory backgrounds with @xmath73-branes or a @xmath126-brane with flux . in fig . [ fig2 ] we depict the corresponding @xmath85-behavior as a function of @xmath113 . in this case , for small values of @xmath120 the value of @xmath118 is unphysical . therefore , the resulting trajectories are quintessence - like , and the model is singularity - free independently of the initial conditions . for larger values of @xmath120 , @xmath118 becomes physical , with @xmath123 leading to phantom behavior and @xmath127 to quintessence - like one with @xmath128 or @xmath129 . in summary , from this simple solution - subclass we observe an interesting @xmath85-behavior . we mention that in principle , @xmath55 and @xmath112 are determined by the supergravity background and can have various forms , while @xmath57 can be more arbitrary since it arises from the interactions of the open / closed string sector which are difficult to compute in general . clearly , considering more general scenarios , with various @xmath112 and @xmath57 or / and not constant @xmath55 , the resulting cosmological behavior can be significantly richer . having discussed qualitatively the cosmological behavior of the model at hand , it would be interesting to perform a systematic investigation of the various cosmological solution sub - classes . in particular , we desire to study the cosmologically important scenario of static solutions characterized by complete dark energy domination . we examine whether there exist late - time attractor solutions , and if they do exist to determine their observable features , that is the dark - energy equation - of - state parameter and density parameter . furthermore , we want to extract information about the intermediate - time behavior , that is the evolution towards the aforementioned late - time attractors , since such an investigation could also leave imprints in observables related to the cosmological past . as usual , we will first transform the cosmological system into its autonomous form @xcite : @xmath130 , where @xmath131 is the column vector constituted by the ( suitably defined ) dimensionless variables and * f(x ) * the corresponding column vector of the autonomous equations , and we extract its critical points @xmath132 satisfying @xmath133 . then , in order to determine the stability properties of these critical points , we expand around @xmath132 , setting @xmath134 with @xmath135 the perturbations of the variables considered as a column vector . thus , for each critical point we expand the equations for the perturbations up to the first order as @xmath136 where the matrix @xmath137 contains the coefficients of the perturbation equations . thus , for each critical point , the eigenvalues of @xmath138 determine its type and stability . defining the dimensionless variables @xmath139 the equations of motion reduce to the following set of equations @xmath140 \nonumber \\ & - & \frac{w_{\phi}}{w}\frac{\sqrt{t}}{\gamma^2}-\frac{\sqrt{3}y}{\gamma^2 m_p}\sqrt{t(1-w\gamma)+v}\end{aligned}\ ] ] where we have set @xmath141 since we are investigating the complete dark - energy dominated scenario . furthermore , in terms of the dimensionless variables , the dark energy equation - of - state parameter ( [ wphi ] ) writes : @xmath142.\end{aligned}\ ] ] since we are interested in static late - time attractors , that is possessing @xmath103 , the corresponding critical points are of the form @xmath143 . thus , linearized perturbations ( @xmath144 , @xmath145 ) lead to @xmath146 where all the derivative terms on the right hand side are evaluated at @xmath147 , and @xmath148 stands for the value of the hubble parameter ( given by ( [ fr1 ] ) and ( [ rhophi ] ) ) calculated at @xmath147 . thus , the corresponding stability matrix reads @xmath149,\ ] ] and its eigenvalues are @xmath150 . requiring negativity of the eigenvalue real part ( which corresponds to stability of the corresponding fixed point ) we result to the constraint @xmath151 in the following we explore the general features of this stability condition , for various cases of the involved potentials and tension . we first consider a solution where @xmath152 const . , and therefore condition ( [ staticstability ] ) reduces to @xmath153 evaluated at the critical value of @xmath147 . this expression ( together with the potential positivity ) imposes tight restrictions on the form of the potential , if we desire to obtain a late - time attractor . in particular , it requires potentials where the field is initially localized near @xmath154 and rolls to larger values ( analogous to the small - field models of inflation ) . candidates are therefore @xmath155 , @xmath156 and @xmath157 . in order to provide an explicit example of this sub - class , we consider the potential @xmath158 . transforming to the variables @xmath159 , where @xmath160 is a reference field position , we can write the equations of motion ( [ eqsfull ] ) in dimensionless form @xmath161 the usual approach is to depict the phase - space plots in the @xmath162 plane , and show the convergence to a stable fixed point . however , it is more transparent to depict the evolution of @xmath163 in terms of the variable @xmath164 . thus , convergence of the system to a static late - time attractor @xmath143 means convergence of @xmath163 to @xmath52 ( as can be immediately seen from ( [ wphixy ] ) setting @xmath165 ) . in fig . [ tight ] we depict @xmath163-evolution , for @xmath166 , @xmath167 , and for various @xmath160-values . as we can see , the system presents phantom behavior , going asymptotically to the cosmological - constant universe . additionally , we see that the location of the ( global ) minimum is shifted to earlier values of @xmath164 as we increase @xmath160 , and it also is closer to the cosmological - constant equation of state . for completeness , in fig . [ cosh ] we depict another example of this solution sub - class , namely corresponding to @xmath168 , with @xmath166 and @xmath167 . as we can see the solution only appears to converge for @xmath169 , and diverges for larger values . let us now examine a more complicated case , considering @xmath170 and @xmath171 , accompanied with @xmath108 . this choice introduces a new mass - scale , which combined with @xmath160 allows us to write @xmath172 . if we desire to provide a theoretical justification through supergravity solutions then @xmath173 is typically large since it governs the radius of an @xmath174 spacetime , and thus @xmath175 will be small . in fig . [ plot1 ] we present the corresponding @xmath176 . one notices that as @xmath160 is increased , the equation of state tends to @xmath52 from below , however it is never too far away from @xmath52 . finally , numerical investigation reveals that the solution is sensitive to the @xmath175-value , with smaller @xmath175 leading @xmath177 to larger time - scales . in this work we have introduced a novel mechanism for realizing either quintessence or phantom dark - energy - dominated phases , within a string - theoretical context . this mechanism is based upon the existence and subsequent dynamics of ghost @xmath1-branes in a warped compactification of the type iib theory , which cancel the combined bps sectors of the @xmath2-brane , preserving the same supersymmetries as their @xmath2 counterparts . the scenario at hand admits a wide range of cosmological behavior , depending on the various terms arising from the supergravity background . in the simplest case , consisting of a single @xmath73-brane and with @xmath120 being constant , we see that phantom behavior will dominate the phase - space dynamics for sufficiently large @xmath178 , since the phase - space singularity @xmath118 is pushed to larger values . beyond the singularity one finds a quintessence solution , asymptotically tending towards @xmath179 . although these features arise from this particular model - subclass , it is clear that more complicated behavior can be revealed considering more general @xmath55 , @xmath112 , @xmath57 cases , with a natural realization of quintessence and phantom behavior , of the @xmath52-crossing and of a big rip . surprisingly enough , although the corresponding @xmath2-brane scenario experiences only quintessence - type solutions @xcite , the present @xmath1-model may lead to both quintessence or phantom cosmology . one can proceed to a more detailed investigation of the phase - space behavior of @xmath1-brane cosmological scenarios , for various cases of the involved tension and potentials @xcite . alternatively one can impose the desired cosmological evolution , and re - construct the corresponding aforementioned functions . since in this work we desire to remain general , exploring the qualitative kinematic features of the @xmath1-brane model , we do not proceed to such extensions , leaving them for a future investigation @xcite . one remaining issue pertains to the quantum stability of such a construction . as it is typical for phantom models , the energy is unbounded from below leading to potential problems upon quantization . however , since the @xmath1-brane is treated semi - classically , we may hope that quantizing the open - string modes with the appropriate boundary conditions may regularize the theory . in particular , since we are dealing with a phantom scalar field , all of the relevant energy conditions are violated . this feature suggests that the phantom may be an unstable mode . to verify the stability one must resort to a quantum field - theoretical analysis . however , developing a quantum theory of world - volume open - string modes using the dbi - action has proven to be difficult , since the @xmath0-brane itself is a non - perturbative state with regards to the string coupling . a boundary state analysis may be possible , but it is beyond the scope of the current work . finally , we mention that since the usual phantom models are robust only for small momenta ( because at larger momenta the higher - derivative terms dominate ) , one could estimate the quasi - stable lifetime of the phantom field , provided the momentum cut - off is fine - tuned and the phantom decays solely into gravitons . similarly , we could follow this line of reasoning for the model at hand , although this means that the field should decay to the closed - string vacuum in a way that the open - string modes give rise to gravitons . for the static case , where the @xmath73-branes are localised , they play a similar role to orientifold planes . in conclusion , we stress that the quantum stability of such negatively - charged objects within string theory is still an open question , but one that is ripe for future exploration . we end this work referring to an additional advantage of the model at hand , namely that it possesses a concrete uv - completion . therefore , it would be interesting to try to embed such branes into full stringy compactifications , particularly if they could serve as replacements for orientifold planes . obviously , the underlying theory would then be @xmath180 , which is phenomenologically unfavored , however there might be another mechanism in the bulk which breaks half of this residual supersymmetry . exploring the nature of such scenarios is something we leave for future endeavor . j.w is partly supported by nserc of canada . e.n.s wishes to thank institut de physique thorique , cea , for the hospitality during the preparation of the present work . b. ratra and p. j. e. peebles , phys . d * 37 * , 3406 ( 1988 ) ; c. wetterich , nucl . b * 302 * , 668 ( 1988 ) ; a. r. liddle and r. j. scherrer , phys . d * 59 * , 023509 ( 1999 ) ; i. zlatev , l. m. wang and p. j. steinhardt , phys . rev.lett . * 82 * , 896 ( 1999 ) . r. r. caldwell , phys . b * 545 * , 23 ( 2002 ) ; s. nojiri and s. d. odintsov , phys . b * 562 * , 147 ( 2003 ) ; v. k. onemli and r. p. woodard , phys . d * 70 * , 107301 ( 2004 ) ; x. m. chen , y. g. gong and e. n. saridakis , jcap * 0904 * , 001 ( 2009 ) ; e. n. saridakis , nucl . phys . b * 819 * , 116 ( 2009 ) . b. feng , x. l. wang and x. m. zhang , phys . b * 607 * , 35 ( 2005 ) ; z. k. guo , _ et al . b * 608 * , 177 ( 2005 ) ; m .- z li , b. feng , x .- m zhang , jcap , 0512 , 002 ( 2005 ) ; b. feng , m. li , y .- s . piao and x. zhang , phys . b * 634 * , 101 ( 2006 ) ; w. zhao and y. zhang , phys . d * 73 * , 123509 ( 2006 ) ; m. r. setare and e. n. saridakis , phys . b * 668 * , 177 ( 2008 ) ; m. r. setare and e. n. saridakis , jcap * 0809 * , 026 ( 2008 ) . p. bintruy , c. deffayet , d. langlois , nucl . b * 565 * , 269 ( 2000 ) ; g.r . dvali , g. gabadadze , m. porrati , phys . b * 485 * , 208 ( 2000 ) ; s. capozziello , int . d * 11 * , 483 ( 2002 ) ; s.nojiri and s. d. odintsov , phys . d * 68 * , 123512 ( 2003 ) ; a. lue and g. starkman , phys . d * 67 * , 064002 ( 2003 ) ; p. s. apostolopoulos , n. brouzakis , e. n. saridakis and n. tetradis , phys . d * 72 * , 044013 ( 2005 ) ; g. calcagni , s. tsujikawa and m. sami , class . quant . grav . * 22 * , 3977 ( 2005 ) . p. horava , phys . d * 79 * , 084008 ( 2009 ) ; g. calcagni , arxiv:0904.0829 [ hep - th ] ; e. kiritsis and g. kofinas , nucl . b * 821 * , 467 ( 2009 ) ; h. lu , j. mei and c. n. pope , arxiv:0904.1595 [ hep - th ] ; c. charmousis , g. niz , a. padilla and p. m. saffin , arxiv:0905.2579 [ hep - th ] ; e. o. colgain and h. yavartanoo , arxiv:0904.4357 [ hep - th ] ; e. n. saridakis , arxiv:0905.3532 [ hep - th ] ; x. gao , y. wang , r. brandenberger and a. riotto , arxiv:0905.3821 [ hep - th ] ; m. i. park , arxiv:0905.4480 [ hep - th ] ; m. botta - cantcheff , n. grandi and m. sturla , arxiv:0906.0582 [ hep - th ] ; m. i. park , arxiv:0906.4275 [ hep - th ] ; c. bogdanos and e. n. saridakis , arxiv:0907.1636 [ hep - th ] . m. grana , phys . rept . * 423 * , 91 ( 2006 ) . m. alishahiha , e. silverstein and d. tong , phys . d * 70 * , 123505 ( 2004 ) ; e. silverstein and d. tong , phys . d * 70 * , 103505 ( 2004 ) . s. kecskemeti , j. maiden , g. shiu and b. underwood , jhep * 0609 * , 076 ( 2006 ) ; x. chen , phys . d * 71 * , 063506 ( 2005 ) ; b. gumjudpai , t. naskar and j. ward , jcap * 0611 * , 006 ( 2006 ) ; j. ward , jhep * 0712 * , 045 ( 2007 ) ; x. chen , jcap * 0812 * , 009 ( 2008 ) ; f. gmeiner and c. d. white , jcap * 0802 * , 012 ( 2008 ) . j. e. lidsey and i. huston , jcap * 0707 * , 002 ( 2007 ) ; d. baumann and l. mcallister , phys . d * 75 * , 123508 ( 2007 ) . e. komatsu _ et al . _ [ wmap collaboration ] , astrophys . j. suppl . * 180 * , 330 ( 2009 ) . d. a. easson , r. gregory , d. f. mota , g. tasinato and i. zavala , jcap * 0802 * , 010 ( 2008 ) . y. f. cai and w. xue , arxiv:0809.4134 [ hep - th ] . y. f. cai and h. y. xia , arxiv:0904.0062 [ hep - th ] . s. thomas and j. ward , phys . d * 76 * , 023509 ( 2007 ) ; s. thomas and j. ward , jhep * 0610 * , 039 ( 2006 ) ; s. thomas and j. ward , jhep * 0611 * , 019 ( 2006 ) . n. r. constable , r. c. myers and o. tafjord , phys . rev . d * 61 * , 106009 ( 2000 ) . y. hyakutake , phys . d * 71 * , 046007 ( 2005 ) . a. a. tseytlin , nucl . b * 501 * , 41 ( 1997 ) ; p. bordalo , l. cornalba and r. schiappa , nucl . b * 710 * , 189 ( 2005 ) . n. r. constable , r. c. myers and o. tafjord , jhep * 0106 * , 023 ( 2001 ) ; p. l. h. cook , r. de mello koch and j. murugan , phys . d * 68 * , 126007 ( 2003 ) . c. papageorgakis and s. ramgoolam , int . . phys . a * 21 * , 6055 ( 2006 ) . s. thomas and j. ward , jhep * 0611 * , 019 ( 2006 ) . l. mcallister and e. silverstein , gen . * 40 * , 565 ( 2008 ) . d. langlois , s. renaux - petel and d. a. steer , arxiv:0902.2941 [ hep - th ] . a. dabholkar , arxiv : hep - th/9804208 . t. okuda and t. takayanagi , jhep * 0603 * , 062 ( 2006 ) . s. terashima , jhep * 0605 * , 067 ( 2006 ) . n. evans , t. r. morris and o. j. rosten , phys . b * 635 * , 148 ( 2006 ) . j. g. hao and x. z. li , phys . d * 68 * , 043501 ( 2003 ) ; x. z. li and j. g. hao , phys . d * 69 * , 107303 ( 2004 ) . a. sen , jhep * 9910 * , 008 ( 1999 ) ; j. kluson , phys . rev . d * 62 * , 126003 ( 2000 ) ; e. a. bergshoeff , m. de roo , t. c. de wit , e. eyras and s. panda , jhep * 0005 * , 009 ( 2000 ) ; m. r. garousi , nucl . b * 584 * , 284 ( 2000 ) . a. vikman , phys . d * 71 * , 023515 ( 2005 ) . h. p. nilles , phys . 110 , * 1 * ( 1984 ) ; c. m. hull , jhep 0111 , * 012 * ( 2001 ) . r. brandenberger , arxiv:0904.2835 [ hep - th ] ; r. h. brandenberger , arxiv:0905.1514 [ hep - th ] ; y. f. cai and e. n. saridakis , arxiv:0906.1789 [ hep - th ] . ferreira , m. joyce , phys . lett . * 79 * , 4740 ( 1997 ) ; e. j. copeland , a. r. liddle and d. wands , phys . rev . d * 57 * , 4686 ( 1998 ) .	we present a novel dark energy candidate , based upon the existence and dynamics of ghost @xmath0-branes in a warped compactification of type iib string theory . @xmath1-branes cancel the combined bps sectors of the @xmath2-branes , while they preserve the same supersymmetries . we show that this scenario can naturally lead to either quintessence or phantom - like behaviors , depending on the form of the involved potentials and brane tension . as a specific example we investigate the static , dark - energy dominated solution sub - class .
deep near - infrared ( nir ) imaging is essential for understanding galaxy evolution at high redshifts . one of its key applications is the determination of galactic mass . nir luminosity is the best tracer of galactic stellar mass because it is less affected by extinction and young stars . this role is traditionally played by deep ground - based @xmath8-band imaging ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? an important reason for this is that the @xmath8-correction at @xmath8-band is small across a broad range of redshifts and is relatively invariant against galaxy types compared to the other two nir bands ( @xmath9 and @xmath10 , see , e.g. , @xcite ) . recently , the high sensitivity of the _ spitzer _ infrared array camera ( irac , @xcite ) adds additional powerful wavebands for this purpose ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? however , ground - based @xmath8-band imaging remains important because of the ease of observations . in order to study galaxy formation and evolution , we have been carrying out deep nir imaging in the great observatories origins deep survey - north ( goods - n , * ? ? ? * ) . in ( * ? ? * hereafter b08 ) we presented a deep @xmath0-band catalog in the goods - n obtained with the wide - field infrared camera ( wircam ) on the 3.6 m canada - france - hawaii telescope ( cfht ) . the wircam image was used by @xcite to study an optically faint submillimeter galaxy . the catalog was used by @xcite to construct a mass - selected sample of galaxies at @xmath11 with an ab magnitude limit of 23.4 . we have since added more data to the wircam @xmath0 imaging and slightly improved the reduction . in this paper we present the new image and a new @xmath0 catalog . in order to achieve a more complete sampling of galaxy mass at high redshifts , we have also constructed a new 3.6 to 8.0 @xmath5 m catalog based on _ spitzer _ irac images in the goods - n ( m. dickinson et al . , in preparation ) , which we also present here . colors in the _ spitzer _ irac bands are used to study the redshifts of galaxies , especially optically faint ones ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? this is largely based on the rest - frame 1.6 @xmath5 m bump in galactic spectral energy distributions ( seds ) , which is caused by the h@xmath12 opacity minimum in the stellar photosphere @xcite . irac colors are also used for separating active galactic nuclei ( agns ) from galaxies ( @xcite , see also @xcite ) . in b08 we studied the effectiveness of using irac colors to select agns and galaxies at various redshifts based on a large spectroscopic sample , and we pointed out the limitations in these selections . with our improved color measurements in the irac and @xmath0 bands , we revisit and discuss these issues . we describe the observations in [ sec_observation ] , the data reduction in [ sec_reduction ] , the reduction quality in [ sec_quality ] , and a @xmath0 selected catalog in [ k_catalog ] . we present in [ sec_color ] a new method of constructing irac catalogs based on the @xmath0-band image and catalog , and we evaluate the performance of our new method . in [ sec_properties ] we describe the general properties of the @xmath0 and irac catalog , including its overlap with other multiwavelength catalogs , the @xmath0 and irac color - magnitude diagrams and color - color diagrams . we summarize our results in [ sec_summary ] . the images and catalog data from this work will be available online . all fluxes in this paper are @xmath13 . all magnitudes are in ab system , unless otherwise stated , where an ab magnitude is defined as @xmath14 . ultradeep @xmath0-band imaging observations of the hubble deep field - north ( hdf - n ) , including the goods - n and flanking fields , were carried out with wircam on the cfht . wircam consists of four @xmath15 hawaii2-rg detectors covering a field of view of @xmath16 with a @xmath17 pixel scale . the observations were made in semesters 2006a , 2007a , and 2008a by two groups . a canadian group led by luc simard obtained 11.5 hr of integration in 2006a . our group based in hawaii obtained 37.9 hr of integration between 2006a and 2008a , bringing the total integration to 49.4 hr . to recover the gap between the sensors and bad pixels , all exposures were dithered with various offsets ranging from @xmath18 to @xmath19 along r.a . and the dither patterns were changed after each semester in order to minimize artifacts caused by flat fielding or by sky subtraction . after each 0.5 to 1 hr of dithered observations , the dither centers were offset by up to @xmath20 to further broaden the area coverage and smooth out the distribution of integration time . the observations were carried out by the cfht in queue mode with weather monitoring . nearly all of the data were taken under photometric conditions . a small fraction ( @xmath21 ) of the data were taken under thin cirrus , so we have down - weighted them ( see next section ) . typical seeing for the observations is between @xmath22 and @xmath23 ( fwhm ) . images taken under @xmath24 seeing were not used in the reduction . in order to reduce imaging data taken with wircam and similar cameras , we have developed an interactive data language ( idl ) based reduction pipeline simple imaging and mosaicking pipeline ( simple ) . here we describe the simple reduction of the wircam data . the wircam images were reduced chip by chip . because of the rapidly varying sky color in the nir , only dithered images taken within @xmath25 hr were grouped and reduced together . a group of raw images were first corrected for nonlinearity in the instrumental response with correction coefficients provided by the cfht ( c .- h . yan , 2008 , priv . the images were then normalized , median - combined to form a sky flat field , and flattened with that . in the flattened images , objects were detected and masked back in the raw images . the masked raw images were normalized again by taking into account the locations of the masked pixels and then median - combined to form a second sky flat . this was used as the final flat field for flattening the raw images again . because there was still some variation in sky color within the @xmath25 hr period , the sky flat did not always perfectly flatten the images . to remove the residual gradients in the flattened images , objects were detected again and masked . a 5th - degree polynomial surface was fitted to each masked image and subtracted from the unmasked image . on each wircam detector , there is crosstalk between each of the 32 readout channels ( @xmath26 pixels each ) , as well as within each of the four video boards ( 8 channels each ) . for every image , simple median - combined the 32 @xmath26 stripes and subtracted the combined stripe from each of the original ones to remove the 32-channel crosstalk features . the 8-channel crosstalk features are relatively weak and only appear around bright objects after combining several hours of images . simple repeated the median combination around only bright objects and only within the affected 8 channels , and then subtracted the median image to remove the 8-channel features . all the median combinations here were performed after detected objects were masked . the above procedures cleanly removed most of the 32 and 8-channel crosstalk . very low level 8-channel crosstalk features are still visible around the brightest @xmath27 stars in our field - of - view after combining the 39 hr of data taken in 2006 and 2007 . this is probably caused by the coupling between crosstalk , flat field pattern , and sky background structures in the images . there is no trivial way to solve this , and we did not attempt to further remove this very low level effect . thanks to the wircam instrument team , the crosstalk in 2008 had improved behavior and no crosstalk residual effect is observed after the above 32 and 8-channel removal procedure . the flattened , sky subtracted , and crosstalk removed images were then used for deriving the pointing offsets between the dithered images and the distortion function of the optics . in order to obtain accurate offsets , simple uses the package sextractor @xcite to detect objects and to determine their pixel coordinates . then offsets are derived by averaging the displacements of the same objects that appear in different images . because of the optical distortion , objects in different parts of the images have slightly different dither displacements . this differential displacement is used by simple to derive the distortion function , following the technique developed by @xcite . simple uses cubic polynomials to fit the distortion functions , i.e. , @xmath28 where @xmath29 and @xmath30 are the undistorted coordinates of objects , and @xmath31 and @xmath32 are the distorted coordinates , @xmath33 and @xmath34 are the distortion functions , and @xmath35 . to first order , the displacements of objects between dithered images can then be expressed as the expansions of the distortion functions : @xmath36 where @xmath37 and @xmath38 are the measured dither displacements of objects , which are functions of positions , and @xmath39 and @xmath40 are the pointing offsets of the dithering . under the above approximation , the measured displacements of objects are the first order derivatives of the distortion functions . for @xmath41 objects in @xmath42 dithered images , there are approximately @xmath43 sets of linear equations for each of @xmath37 and @xmath38 . the systems of linear equations were solved for the coefficients of @xmath33 and @xmath34 with a least - squares method . initially , the mean values of @xmath37 and @xmath38 were used as rough estimates of @xmath39 and @xmath40 . after @xmath33 and @xmath34 were integrated , they were used to correct for @xmath31 and @xmath32 to derive better estimates of the dither offsets @xmath39 and @xmath40 . then improved coefficients of @xmath33 and @xmath34 were solved with the new @xmath39 and @xmath40 . for wircam , one such iteration seems to provide sufficiently good distortion functions . since the optical distortion was derived within a @xmath25 hr sequence of dithered images , the above approach has the advantage of accounting for the long - term time - dependent flexure of the telescope . it also does not require any external information and thus can easily be applied to any target field . to determine the absolute astrometry , we used several different catalogs in the hdf - n region . we first combined the catalogs of the sloan digital sky survey(sdss ) data release 6 around the hdf - n region and the v1.0 goods - n advanced camera for surveys ( acs ) catalog @xcite , after correcting for the well - known @xmath44 offset between the goods - n acs frame and the radio frame . the combined catalog was cross - checked with the source positions in the very large array ( vla ) 1.4 ghz catalog of @xcite to ensure consistency with the radio frame . the comparison between the acs and the vla astrometry is only good to several tens of milli - arcseconds , because of the differences in optical and radio morphologies of distant galaxies . this sets the limit for our absolute astrometry . we then reduced several hours of the best quality wircam @xmath0-band data and matched the wircam object positions to the above sdss / acs astrometric catalog . we added bright and compact wircam @xmath0 objects to the sdss / acs catalog to form a dense astrometric catalog for reducing the full set of wircam @xmath0 images . the final catalog has @xmath45 objects over a diameter of 1.3 degree , approximately 5700 of which are concentrated in the central 0.6 deg@xmath2 region . in all the reductions , when individual images were warped to correct for distortion and projected onto a common sky plane , the objects detected in the images were forced to match the positions in the astrometric catalog . the registration of the images onto the catalog was performed at the sub - pixel level with a bilinear resampling . there was only one resampling for each exposure in the entire reduction , which was a fifth degree polynomial including both the distortion correction and a tangential sky projection . this one - step resampling minimized the image smearing . before the distortion - corrected images were combined to form a deep mosaic , the images were calibrated for extinction variation , cosmic rays were removed , and the images were weighted . the relative extinction variation was corrected using fluxes measured by sextractor on high s / n objects in the images . then pixels in different images that have the same sky coordinates were compared for rejecting cosmic rays with sigma clipping . the clipping rate was kept to be approximately 1% so only an insignificant amount of useful data were discarded . the sigma clipping allows weighted mean combinations of the images , which produces 25% higher s / n compared to median combinations when all images have equal weights . the images were optimally weighted according to the atmospheric transparency ( @xmath46 , derived from the flux calibration , see below ) , integration time ( @xmath47 ) , and background brightness ( @xmath48 ) . furthermore , each pixel was weighted by its relative efficiency ( @xmath49 , proportional to flat field ) . thus the final weight applied to a pixel is : @xmath50[weight_eq ] since objects in the images have a wide range of sizes , we did not weight the images with seeing . the relative extinction corrected , cosmic ray removed , and weighted images were then mean - combined to form a deep mosaic . on the mosaicked images , object fluxes were measured again with sextractor using @xmath51 diameter apertures , and the results were cross - correlated with the two micron all sky survey ( 2mass , @xcite ) point - source catalog . we used 2mass point sources with @xmath5216.34 to calibrate the absolute flux of the wircam image . objects brighter than this range are affected by wircam nonlinearity ( despite the correction for linearity ) and objects fainter than this range show a significant selection effect in the 2mass fluxes . we adopted the 2mass `` default magnitudes , '' which attempt to account for the total fluxes of the point sources @xcite . because the 2mass @xmath0 and wircam @xmath0 filter transmission curves are very similar ( figure [ fig_filter ] ) , we directly adopted the 2mass @xmath0 magnitudes and did not attempt to account for the slight difference between the two systems . we adopted the 666.7 jy @xmath0 vega zero - magnitude provided by 2mass @xcite . calibrated mosaic images from all the dither sets and from the four chips were combined into a final wide - field , ultradeep image . all the weights described above were carried into the final image in the form of effective integration time . we found that despite the large variation in sky background brightness , noise in the final image scales nicely with the inverse square - root of the effective integration time in intermediate and shallow regions where confusion noise is less important . this means that our weighting scheme provides excellent s / n optimization . figure [ fig_layout ] shows the integration time distribution and layout of our survey field . figure [ fig_area ] shows the cumulative area in the final mosaic as a function of effective integration time . the final combined mosaic covers a square area of @xmath53 , centered at @xmath54 , @xmath55 ( j2000.0 ) . the central @xmath56 region has more than 20 hr of integration at each position and fully covers the goods - n acs area . the image has uniform image quality across the entire field , with a fwhm of @xmath57 . the simple reduction of the wircam @xmath0 image forced the positions of detected objects to match those in the astrometric catalog formed with the sdss and the goods - n acs catalogs . the resultant astrometry is excellently constrained by the acs catalog in the central region and reasonably constrained by the sdss catalog in the outer region . to quantify the ( relative ) astrometric accuracy , we selected well - detected compact sources ( s / n @xmath58 and fwhm @xmath59 ) from the sextractor catalog ( section [ k_catalog ] ) . nearly 4000 such sources have counterparts in the sdss or goods - n acs catalog . figure [ fig_astrometry_1 ] and figure [ fig_astrometry_2 ] show the offsets between the wircam and the sdss / acs positions . there are no observable systematic offsets between the wircam positions and both the sdss and acs catalogs . however , the astrometry is clearly better in the central part of the image where acs sources were used to constrain the astrometric solution . the rms dispersions of the offsets are @xmath60 between wircam and acs along both the r.a . and axes on nearly 4000 sources . the rms offsets between wircam and sdss are larger , @xmath61 for ra and @xmath62 for dec , on nearly 460 sources . we found that the sdss and acs catalogs have similar @xmath63 rms offsets between each other . this suggests the larger observed offsets between the wircam and sdss positions may primarily be a consequence of an uncertainty internal to the sdss catalog . the above measured rms offsets include both systematic and random errors . for example , the measured @xmath60 rms offset between wircam and acs is slightly larger than the nominal positional uncertainty for fwhm @xmath57 and s / n @xmath58 , which should be @xmath64 . to estimate the systematic error in astrometry , we repeated the same measurements with higher s / n cuts and lower fwhm cuts . the rms offset decreases initially but stops decreasing when it approaches @xmath65 , even when the selected sources have extremely high s / n ( e.g. , @xmath66 ) and have fwhm comparable to the seeing . we conclude that there is an @xmath67 systematic uncertainty in our wircam astrometry in the goods - n acs region . such an error includes both the internal error in the acs catalog and the error introduced by the image registration of simple . on the other hand , the rms offset between wircam and sdss is @xmath63 . since all sdss sources are detected in the wircam image with s / n @xmath68 ( median s / n is 2300 ) , random noise is negligible here . we conclude that the systematic error is @xmath63 in outer regions where acs positions are unavailable . the absolute flux scale of our wircam @xmath0 image was calibrated with 2mass point sources in the image . figure [ fig_photometry_1 ] shows the comparison between the 2mass fluxes and the calibrated wircam fluxes derived with @xmath51 diameter apertures . the ratio between the two is only flat in a narrow flux range . we chose @xmath6916.34 ( the two vertical lines in figure [ fig_photometry_1 ] ) for the calibration of the wircam image . outside this range , there are nonlinearity problems with the wircam sensors ( bright end ) and selection effects in 2mass ( faint end ) . in the chosen range for calibration , the error - weighted mean of the flux ratios ( @xmath70 ) for 70 sources is @xmath71 , indicating that the overall absolute calibration of the wircam @xmath0 image is good to @xmath72 . we note that the error bars in figure [ fig_photometry_1 ] are dominated by 2mass flux errors . thus , the calibration of our wircam image is limited by the photometric quality of 2mass . it is not possible to achieve better results unless we make frequent monitoring of standard stars during our observations , which is quite expensive in terms of observing time . we also investigated the photometric uniformity in our wircam @xmath0 image . the above value of 0.5% is the mean calibration error over the entire image . it is possible that the calibration of certain smaller regions is worse than this value . we subdivided the wircam image into four quadrants , roughly corresponding to data obtained from the four detector chips , and then compared the wircam fluxes with the 2mass fluxes as we did for the entire image . for quadrants 1 through 4 , the error - weighted flux ratios are @xmath73 ( 16 sources ) , @xmath74 ( 19 sources ) , @xmath75 ( 16 sources ) , and @xmath76 ( 22 sources ) , respectively . in all of these cases , based on the numbers of available 2mass sources and their 2mass flux errors , the expected calibration errors are roughly 1% . the standard deviation of the four measured offsets is 1.3% , which compares well with the expected calibration error from 2mass . this suggests that there is not an observable photometric gradient over size scales of @xmath77 . the absolute fluxes of objects are still limited by the photometric quality of 2mass at this scale . the measured 1.3% variation is indeed excellent when compared to other @xmath8-band extragalactic deep surveys over similar size scales ( e.g. , 2% in the cosmos survey , * ? ? ? we used sextractor ( v2.5.0 ) to produce a @xmath0 source catalog . the most important source extraction parameters in sextractor are summarized in table [ tab_sex ] . we provide our raw sextractor @xmath0 catalog in this data release . we caution that the detection parameters ( the first three in the table ) that we adopted are somewhat aggressive in detecting faint sources , which may lead to more spurious sources ( see [ sec_spurious ] ) . the users of our data product may want to regenerate catalogs with more conservative values , based on the nature of their work . based on the raw catalog , we compiled a @xmath0 selected catalog , which is the primary product of this paper . we first estimated the true noise level in the wircam @xmath0 image . we masked the detected objects and convolved the image with circular kernels of amplitude 1.0 ( i.e. , flux apertures ) . the local ( within @xmath78 ) rms fluctuations in the convolved image were considered as errors in aperture photometry that include all possible sources of noise : photon noise , correlated noise between pixels , confusion noise from faint , undetected objects , and uncertainty in the estimate of the small - scale background . we compared the noise measured in this way with the photometric errors derived by sextractor . we found that for all aperture sizes between @xmath79 and @xmath51 in diameter , the ratio between the two errors is fairly insensitive to the aperture size and has a value of 1.28 . thus , we multiplied all of the flux errors in the sextractor catalog by this factor to approximate the total noise . this does not affect the flux measurements . lr detect_minarea & 2 + detect_thresh & 1.25 + analysis_thresh & 1.25 + filter & y + filter_name & gauss_1.5_3x3.conv + deblend_nthresh & 64 + deblend_mincont & 0.00001 + clean & y + clean_param & 0.1 + seeing_fwhm & 0.8 + back_size & 24 + back_filtersize & 6 + back_type & auto + backphoto_type & local + backphoto_thick & 24 + weight_type & map_weight + weight_thresh & 10 the next step is to estimate aperture corrections . in our data reduction we calibrated the flux on 2mass point sources using fixed @xmath51 diameter apertures . thus , sextractor fluxes measured with @xmath51 apertures do not need any aperture corrections and we consider @xmath51 aperture fluxes to be the best estimate of the total fluxes of the sources . this clearly does not apply to the several most extended objects in the image , since their aperture corrections should be much larger than that for smaller objects ( i.e. , 2mass point sources used in the calibration ) . since they are not the major targets of our work , we did not attempt to obtain highly accurate fluxes for these very few sources . despite that @xmath51 aperture fluxes are the best estimate of total fluxes , their s / n is relatively low . we thus looked for suitable fluxes measured with smaller apertures , and corrected the small aperture fluxes to @xmath51 when it is necessary . we compared all other sextractor flux options with the fixed @xmath51 diameter aperture fluxes and found that the sextractor `` auto - aperture '' fluxes agree excellently with them at all flux levels . in addition , the auto - aperture fluxes have the best optimized s / n for sources with various morphologies and fluxes . we thus decided to adopt directly the sextractor auto - aperture fluxes and errors in our final catalog . we did not apply any additional corrections except for the aforementioned factor of 1.28 to the errors . the adoption of sextractor auto - aperture fluxes works excellently for nearly all sources . however , a very small number of sources ( 0.7% of total ) are undetected in auto - apertures but detected in fixed apertures . for such sources , we adopted their @xmath80 fluxes and corrected the fluxes to @xmath51 with a median correction factor from all other detected sources . in addition , after all the above corrections to the fluxes and flux errors , some sources in the sextractor raw catalog are still undetected . given that our error estimate is realistic and the flux is well optimized , we considered sources with s / n @xmath81 as undetected ( 1.9% of total ) and only included their flux errors in the final catalog for upper limits . to keep the source number and order in our final @xmath0 catalog identical to those in the sextractor raw catalog , we did not remove such undetected sources . this is for easy comparisons with the sextractor catalog , in case the users of our data need to carry out more consistency or quality check . we next evaluated the depth and completeness of our @xmath0 wircam catalog . in the central @xmath82 of the image where the sensitivity distribution is more uniform , the median 1@xmath3 error among 3@xmath3 sources is 0.15 @xmath5jy , corresponding to a 3@xmath3 limiting magnitude of 24.77 . in the goods - n region , where the image is the deepest , the median 1@xmath3 error is 0.12 @xmath5jy , corresponding to a 3@xmath3 limiting magnitude of 25.10 . this is as deep as the irac 3.6 @xmath5 m image in the same region . we estimated the completeness of our @xmath0 catalog with monte carlo simulations . we note that completeness is a complex function of flux , morphology , crowdedness , and source extraction . the results here are meant only to provide a guide . exactly how one should carry out the completeness computation depends on the purpose of the study , and we shall leave this to the users of our data . here , we first randomly selected well - detected ( s / n @xmath58 ) sources from the sextractor catalog . the selected sources span a wide range of flux , size , and shape . we extracted these sources from the wircam image and artificially dimmed or brightened them to the flux range of interest . then we randomly placed them in the wircam image and ran sextractor to detect them . each time we only placed a small number of sources in the image so the overall source density did not change by @xmath83 . a total of @xmath84 sources were randomly placed in the image . the sextractor catalogs for simulated images were processed identically to the real catalog . we considered a source recovered if it is detected at @xmath85 . we divided sources into four groups according to their half - light radii ( @xmath86 ) and calculated the recovery rate in each 0.1 dex flux bin and @xmath86 group . the results in the central @xmath82 region are shown in figure [ fig_completeness]a . the 90% completeness limits for the two compact groups ( @xmath87 and @xmath88 , comprising @xmath89 of the total sample ) are 1.10 @xmath5jy ( @xmath90 ) and 1.32 @xmath5jy ( @xmath91 ) , respectively , equivalent to 23.80 and 23.60 magnitude . for the entire sample the mean 90% completeness weighted by the size distribution is 1.38 @xmath5jy ( @xmath92 , which is equivalent to 23.55 magnitude . if we restrict this to the deeper 0.06 deg@xmath2 goods - n irac region ( figure [ fig_completeness]b ) , then the mean 90% completeness limit becomes 0.94 @xmath5jy ( @xmath93 ) , or 23.96 magnitude . the above results show that the present wircam imaging is the deepest in the @xmath8-band over a substantial area . for example , the very large telescope ( vlt ) nir imaging in the goods - s is 90% complete at @xmath94 in a small area of @xmath95 arcmin@xmath2 @xcite . on the other hand , our team @xcite and japanese groups @xcite have carried out extremely deep @xmath0 band imaging in the goods - n at the 8.2 m subaru telescope with a smaller @xmath96 imager . @xcite combined all archived data to make a deep @xmath0 image of the goods - n . they achieved 3@xmath3 limiting magnitudes of 25.5 and 26.5 in 103 arcmin@xmath2 and 28 arcmin@xmath2 , respectively . this subaru image is deeper than our wircam image ( 3@xmath3 of 25.1 magnitude in the @xmath97 arcmin@xmath2 goods - n region ) , but the area coverage is much smaller . we took two approaches to estimate the spurious fraction , one based on an inverted image and the other based on multi - wavelength counterpart identification . first , we constructed an image with inverted background , and then ran through the same source detection and catalog compilation described above . this infers a spurious fraction of @xmath98 on sources with s / n @xmath99 , which is unreasonably high . after examining the image and the `` sources , '' we believe this is because the crosstalk of wircam produces negative holes near bright objects , making the noise non - gaussian and greatly overestimating the spurious fraction . one evidence for this is that , in relatively crosstalk - free regions , the spurious fraction inferred from the inverted image dramatically decreases to @xmath100 . this value of 5% is an upper limit since there are still low - level crosstalk effects . we then looked for @xmath0 sources that are not detected in other images . in the region where the goods - n acs and irac images overlap , there are 15112 @xmath0 sources detected at @xmath101 . among them , 489 ( 3.2% ) sources do not have counterparts in the acs catalog and are not detected at 3.6 and 4.5 @xmath5 m at @xmath101 in the irac catalog described in the next section . ( see [ sec_cat_overlap ] for a slightly different comparison . ) the fraction 3.2% is also an upper limit since we can not rule out real @xmath0 sources being not detected in the acs and irac images , especially given that the irac images are not substantially deeper than the @xmath0 image . this 3.2% upper limit is tighter than the 5% limit set by the inverted image , consistent with the above suggested non - gaussian noise . because of this limitation , it is difficult to further constrain the spurious fraction of our @xmath0 catalog . we suggest the users of our data to carry out independent source extraction if more conservative extractions are desired . in table [ tab_prop ] we summarize the basic properties of our wircam @xmath0 image described in this and previous sections . lc semesters & 2006a , 2007a , 2008a + p.i . & l. cowie , l. simard + total integration & 49.4 hr + field center ( j2000.0 ) & + r.a . & @xmath102 + dec . & @xmath103 + area ( @xmath104 hr ) & 0.25 deg@xmath2 + seeing fwhm & @xmath23 + calibration uncertainty & 1.3% + astrometric uncertainty & @xmath65@xmath105 + 90% completeness limit & 1.38 @xmath5jy ( 23.55 mag . ) + spurious fraction & @xmath106 we next used the wircam @xmath0 image and catalog as priors to analyze the fluxes of galaxies in the goods - n _ spitzer _ irac images at 3.68.0 @xmath5 m ( m. dickinson et al . , in preparation ) . similar work has been done previously . in both the goods - n @xcite and the goods - s @xcite , acs @xmath107-band images were used as priors to determine fluxes in the irac bands . while our @xmath0 image has a lower resolution than the acs @xmath107 images , the waveband is much closer to the irac bands , making it a better choice for this type of analysis . the primary goal is to better measure the colors of objects between the irac and @xmath0 bands . the nearly @xmath108 resolution of irac is significantly worse than the typically sub - arcsecond resolution in the optical and nir , but it is not much larger than the sizes of galaxies . this makes color measurements extremely uncertain . normally one would need to sacrifice resolution in the optical and nir bands to achieve this the approach we describe below does not have this disadvantage . a secondary goal is to better separate irac fluxes of objects that are close to each other . there are nearly 20,000 @xmath0 detected source in the goods - n irac area , but sextractor can only directly detect @xmath109 sources in the 3.6 @xmath5 m irac image . one reason for the smaller number of sources in the irac image is that many objects are blended under the irac resolution . we will demonstrate that our method provides reasonable estimates of the irac fluxes of lightly blended objects . the basic idea is to convolve the @xmath0 image of each galaxy individually with a scaled convolution kernel to construct a model image for an irac image . the scaling of the kernel varies between galaxies because galaxies have different irac - to-@xmath0 flux ratios ( i.e. , colors ) . the set of scaling factors that produce the best model image are then the best estimates of the flux ratios . it is immediately obvious that the results would be sensitive to the evaluation of whether a model best approximates the irac image , as well as whether the chosen convolution kernel well represents the psf . in @xcite and @xcite , the model that produced the minimum @xmath6 was considered as the best model . in such analyses , @xmath6 is very sensitive to the mismatch of the psf , and unfortunately , there is a huge uncertainty in the psfs in the irac bands . this was pointed out by grazian et al . and is also clearly shown in figure [ fig_subtraction ] . it is thus unclear to us whether or not @xmath6 analyses are the best approach . here we present an alternative approach based on the widely used image deconvolution technique in radio astronomy clean @xcite . instead of minimizing the @xmath6 of the pixels , clean determines the scaling of the kernels based on the fluxes of the image components . this is very attractive to us since our true goal is to be able to account for the fluxes detected in the irac images . as long as the fluxes in an irac image can be cleanly removed by subtracting a model , we consider the model to be the best . below we describe the basic steps in our analysis of the galaxy colors . we hereafter refer to our technique as realclean , since it is done in real space instead of fourier space . the goods - n irac images have pixel scales of @xmath110 , so we background subtracted and resampled them to the wircam @xmath17 pixel scale . during the resampling , the irac images were also slightly warped such that the positions of objects precisely match those in the wircam image . to construct psfs for the irac and wircam images , we selected @xmath111 bright but unsaturated , point - like , and isolated objects in these images . the images of these objects were aligned , stacked , and normalized to form the psfs . the irac psfs were then deconvolved by the wircam psf with a maximum likelihood deconvolution of a few iterations . the deconvolved irac psfs were used as the convolution kernels to convolve the wircam image . one may worry about the uncertainty in the deconvolution of the psfs . we also tried an alternative approach that does not require the deconvolution of the irac psf , in which we measured colors in the irac images that are convolved with the wircam psf and in the @xmath0 image that are convolved with the irac psfs ( see [ xconv ] ) . we did not see a systematic difference between the two approaches . we thus adopted the deconvolved irac psfs as the convolution kernels . in the background - subtracted wircam @xmath0 image , for each detected object we identified pixels associated with that object . this was done with the `` segmentation map '' provided by sextractor . this allowed us to apply different scaling factors to the convolution kernels on different objects . a minor drawback here is that very faint pixels in the outer parts of galaxies would be missed . however , as long as these faint pixels do not have dramatically different colors from the brighter cores of the galaxies , our measured colors can still be considered as good estimates of the flux - weighted mean colors of galaxies . another minor drawback is pixels that contain emission from more than one object . this would only become problematic when objects are not well resolved in the wircam image . since this is the fundamental limit of the wircam observations , we did not attempt to resolve this issue . to obtain model images that best approximate the observed irac images , we use our realclean method , which is especially useful when determining fluxes of blended galaxies . in a realclean cycle , we first identify the brightest galaxy in the irac image . then we remove a small fraction ( 25% at 3.6 @xmath5 m down to 10% at 8.0 @xmath5 m ; this factor is often called the `` gain '' ) of its flux from the irac image by subtracting a model image constructed by convolving the wircam image with the above mentioned kernel . the flux is determined by placing a circular aperture at the object . we used aperture diameters of @xmath112 , @xmath112 , @xmath51 , and @xmath113 for the irac 3.6 , 4.5 , 5.8 and 8.0 @xmath5 m bands , respectively . despite the fact that such aperture photometry is affected by nearby objects , because we only look at the brightest object and because the gain is @xmath114 , we guarantee that we do not over - subtract the object . to speed up the realclean process , we subtract fractional fluxes from many objects in one realclean cycle . however , in each realclean cycle we only subtract objects that are sufficiently far away from each other that locally each subtracted object is the brightest . in the subsequent realclean cycle we look for the next brightest objects in the subtracted image and subtract small fractions of fluxes from them . we repeat this iterative procedure until no objects in the residual irac image have fluxes brighter than a few @xmath3 . it is interesting to note that the above procedure is similar to the clean deconvolution of radio interferometric data ( see a summary in @xcite ) in many ways . to limit the degrees of freedom , clean is often limited to clean windows or boxes within which emission is known or guessed to be present . clean also sometimes uses priors to improve the speed of convergence of the solution . these are exactly what our procedures do : we only subtract irac fluxes at the locations of @xmath0 detected galaxies , and we use the wircam @xmath0 images of the galaxies as the model . once realclean reaches a few @xmath3 , all objects have similar brightness and realclean is no longer efficient and necessary . thus , we measured the residual irac fluxes of all of the wircam objects and subtracted them from the irac image all at once . this global subtraction unavoidably over - subtracts blended objects , especially those whose separations are less than the flux aperture sizes . since the global subtraction is carried out only after the objects are realcleaned to a few @xmath3 , the amount of over - subtraction is small . nevertheless , the over - subtraction can be further minimized . we estimated the over - subtraction by measuring the fluxes again in the residual image . then we re - iterated the global subtraction by adjusting the subtracted fluxes and again measured the over - subtraction in the next residual image . in each iteration , the adjustment of fluxes were small fractions ( equal to the realclean gains ) of the estimated over - subtraction . this slowly brought the residual fluxes of galaxies in the subtracted image from negative ( after the first global subtraction following the realclean cycles ) to nearly zero after tens of iterations . in the iterations , we closely watched the residual fluxes of galaxies to see if they converged to zero . when the residual fluxes no longer converged ( typically after 4060 iterations ) , we stopped the iterations . fluxes subtracted in the realclean stage and in this minimization stage were added to make the total irac fluxes of the wircam @xmath0 detected galaxies . for each irac band , there are two epochs of observations covering different areas in the goods - n but with some overlap . all the procedures described above , including the psf construction , were carried out independently on each single - epoch image . this is because the two epochs have different orientations of pixels and psfs . for sources in the overlap region , we adopted the error weighted fluxes from the two epochs . as mentioned , we used @xmath112@xmath113 diameter apertures to measure fluxes , or in other words , to evaluate whether the model best removes fluxes of galaxies . such small aperture sizes nicely cover the cores of the irac psfs and minimize influences from nearby sources . we also derived aperture corrections from the convolution kernels . however , the proper correction is a function of the sizes of the galaxies , which are not exactly point sources under the irac resolution . this biases the measured fluxes , especially when the apertures are small . our small apertures also miss the information in the outer wings of the psfs . to account for these , we repeated the realclean procedures with 50% larger apertures and compared the results with those from small - aperture realcleans . such aperture sizes are comparable to the sizes of the first diffraction rings in the irac psf and are sufficiently large for the majority of galaxies in the goods - n . we applied a constant correction factor ( 2%6% ) to all the small - aperture fluxes , derived from the large - aperture realclean on high s / n and isolated sources . we estimated flux errors in a similar way to those for the wircam fluxes described in [ k_catalog ] . we convolved the residual irac images with circular uniform kernels of amplitude 1.0 and measured the rms fluctuation locally . this way , the measured noise included not only photon noise , confusion noise , and correlated noise , but also rough estimates of the uncertainties caused by imperfect object removal in the entire realclean and flux minimization processes . irac fluxes and errors measured in this way are included in the @xmath0 selected source catalog in this data release ( table [ master_cat ] ) . to evaluate the performance of the above realclean - based procedures , we first visually inspected the residual images after subtracting the best models . figure [ fig_subtraction ] shows the hdf - n region of the irac images before and after the subtraction . overall , the subtraction of irac objects with our realclean method appears reasonably good in all irac channels . however , minor residual effects can be seen . there are faint halos around many objects in all channels after the subtraction , likely a consequence of the fact that our small flux apertures did not fully account for fluxes from slightly extended objects . on the other hand , on compact objects there are residual patterns around the cores of the objects . in the first two irac channels , this indicates a @xmath115 uncertainty in the adopted irac psf . because this occurs across the entire field , this is unlikely to be an effect of psf variation or position - dependent mis - registration . this is the most severe in the first two irac channels , suggesting that under - sampling of the psf by the large irac pixels may have played a role . however , the exact reason remains unclear to us . the 5.8 @xmath5 m channel has the cleanest subtraction of irac objects . it gets slightly worse again in the 8.0 @xmath5 m channel , likely because the 8.0 @xmath5 m morphology becomes sufficiently different than that in the @xmath0-band . 0.5 cm the above visual inspection of the residual images qualitatively shows that _ overall _ our realclean approach can account for the irac fluxes of @xmath0-band detected objects and that there are no catastrophic failures in subtracting off the irac objects . however , this does not guarantee that fluxes are correctly assigned to objects when multiple objects are blended by the irac psf . to test this , we carried out simple monte carlo simulations . we simulated objects of high s / n ( @xmath111 ) blended by nearby objects . the flux ratios between the blending bright sources and the targets are from 1.0 to 100 . the separations between the targets and the nearby sources are from @xmath116 to @xmath112 , with various position angles . all objects are point - like and have flat spectra . psfs derived from the two epochs of the real irac images and the real wircam image were used to create the associated simulated images . all the images were processed identically to the above realclean and catalog making procedures . we shown the results of the simulations in figure [ fig_sim_clean ] . a few trends can be observed and are indeed as expected . realclean performs the best when the fluxes ratios are @xmath95 and separations are @xmath117 . at very small separations of @xmath116 , realclean can also correctly recover the fluxes if the nearby objects are less than @xmath118 brighter then the targets . among the @xmath119 real sources detected at 3.6 @xmath5 m in our final catalog ( [ sec_combined_catalog ] ) , 90% of their brightest neighbors fall in the above ranges of flux ratio and angular separation . there no data points in regions of large flux ratios and small separations ( including flux ratios of greater than 31.6 , which are not shown ) . these are the cases in which even sextractor can not resolve the blended sources in the wircam images . realclean can not determine the irac fluxes if the sources are not resolved by sextractor in the wircam images . at the boundary when the sextractor can barely resolve the blended pairs , realclean either over - estimates ( large separation ) or under - estimates ( small separation ) the fluxes of the targets . the amount of over and under - estimate can be up to 40% . here we are pushing the limits of both realclean and sextractor . in our catalog , the fractions of sources with @xmath120 neighbors that are @xmath121 and @xmath122 brighter are 5.7% and 3.6% , respectively . the above results imply good realclean fluxes for the majority of sources in our catalog . a small number of sources may have irac fluxes affected by very bright neighbors . the effect is generally not @xmath123 before sextractor becomes unable to detect the faint sources . the caveat here is that the above simulations are very simple . one may conduct much larger simulations with variable object morphology , s / n level , and number of blended sources . we do not think these are particularly useful given the large uncertainty in the irac psf discussed in [ sec_residual ] . realistic simulations should take into account such uncertainties , folding in how the psf is under - sampled by the irac pixels and how the observations were carried out and reduced to better sample the images ( e.g. , dither and drizzle ) , which is out of the scope of this paper . without good knowledge about the psf , going further beyond the above simulations does not likely help on understanding the performance of realclean . in what follows , we will rely on simple consistency checks within the available data , paying special attention to blended galaxies . the goods - n irac observations consist of data from two epochs . both the psfs and the pixels in these two epochs have different orientations . the two epochs of imaging overlap in approximately 35 arcmin@xmath2 of area at the center of the goods - n . since we ran our realclean procedures on the two epochs separately , we compared our results in the overlap region to see if they are consistent . figure [ fig_epoch1vs2 ] shows the ratios of the fluxes derived from epochs 1 and 2 . the _ dashed _ curves in each plot show the dispersion of the flux ratios , and the _ solid _ curves show the expected dispersion based on flux errors estimated in [ sec_irac_flux ] . the fact that the two curves are very close to each other is a strong indication that our error estimate is realistic and that the uncertainty of the fluxes derived from our realclean procedure is limited by the noise in the data . we examined the irac sources that have nearby @xmath0-band detected objects with separations of @xmath124@xmath108 ( _ diamonds _ and @xmath108@xmath80 ( _ triangles _ ) . in the @xmath124@xmath108 category , the distributions of the flux ratios are not significantly different from the main sample , except that at 8.0 @xmath5 m . at 8.0 @xmath5 m the @xmath108@xmath80 category has a distribution similar to the main sample . these results suggest that our procedures produce consistent results against slight changes in the psf for blended sources with separations greater than @xmath124 at 3.6 , 4.5 , and 5.8 @xmath5 m and @xmath108 at 8.0 @xmath5 m . in the @xmath0 catalog for sources with s / n @xmath125 , the median separation between galaxies and their nearest neighbors is @xmath126 , and 99.5% of sources do not have a nearest neighbor within @xmath124 . this suggests that nearly all of the fluxes we derived for the first three irac channels are reliable . on the other hand , among the @xmath127 8.0 @xmath5 m detected sources , approximately 10% of the sources have nearest neighbors at @xmath128 . their fluxes are more problematic , if figure [ fig_epoch1vs2 ] is indicative . however , the percentage would be lower if we only considered sources whose nearest neighbors were significantly brighter than themselves . one can imagine that if sources with similar fluxes are blended , realclean naturally assigns similar fluxes to both if the gain is sufficiently small ( which is the case for our 8.0 @xmath5 m realclean ) . on the other hand , if sources with different fluxes are blended , the procedure might fail , especially on the fainter member of the pair . these are also indicated in our monte carlo simulations . the key goal of our analyses is to obtain accurate colors for the wircam and irac sources . we selected sources that have redshifts in the spectroscopic survey of b08 and compared their colors with model seds . here we used two seds with dust and polycyclic aromatic hydrocarbon ( pah ) features from @xcite , m 82 ( a dusty starburst ) and m 100 ( a spiral ) , and a dustless starburst template from @xcite that has a blue stellar continuum and strong recombination lines . we also reddened the m 82 sed with @xmath129 using the @xcite extinction law . these four seds represent a broad range of galaxy colors . figure [ fig_z_color1 ] shows the comparison between the data and the model seds . as in figure [ fig_epoch1vs2 ] , galaxies with nearest neighbors of @xmath124@xmath108 separations are shown with _ overall the derived irac and @xmath0 colors match the sed models excellently . at @xmath130 and 3.4 in the @xmath131 ratio and @xmath132 in the @xmath133 ratio , there are substantial numbers of observed colors significantly bluer than the starburst model . this is true for both isolated galaxies and galaxies with close neighbors . we believe these blue `` outliers '' are not artifacts of our realclean procedures but rather objects with emission lines even stronger than those in the starburst template . the lines are likely h@xmath134 and [ ] ( entering @xmath0 at @xmath130 ) , h@xmath135 and [ ] ( entering @xmath0 at @xmath136 ) , and [ ] ( entering the 3.6 @xmath5 m band at @xmath132 ) , all present in the starburst template . we attempted to verify this with our spectroscopic data . unfortunately , we only have optical spectra on three of these galaxies , as most of the redshifts are adopted from @xcite . the three sources have standard lyman break galaxy spectra with strong absorption . in one case there is also ly@xmath134 emission . in the other cases we do not cover ly@xmath134 . galaxies with strong rest - frame optical ( corresponding to @xmath0 and irac bands at these redshifts ) emission lines have absorption lines in the rest - frame ultraviolet ( corresponding to optical at these redshifts ) . given their rest - frame ultraviolet continua , these three galaxies are inevitably strong emission - line galaxies in the rest - frame optical . this is consistent with strong emission lines causing the blue observed colors at @xmath0 and irac . there are a few unexplained blue outliers at @xmath1372 in @xmath131 and at @xmath138 in @xmath133 . they might be fluxes mistakenly assigned by our realclean procedure . however , if this is the case , it would be very surprising if such outliers are not observed at all at @xmath11 , where most of the spectroscopic samples lie . an alternative explanation is that they are galaxies with misidentified redshifts . we have spectra for these galaxies , except for two whose redshifts were adopted from @xcite . we inspected the available spectra and did not see obvious signs of misidentification . thus , the nature of these blue outliers remains unclear . nevertheless , we emphasize that the observed irac colors of the majority of the galaxies match excellently with the models at nearly all redshifts . combining this with our observations in [ sec_overlap_region ] , we are confident that our realclean procedures correctly assigned fluxes to nearly all of the irac sources . the last performance test is a comparison with _ absolute flux calibrators in @xcite . figure [ fig_star_color ] shows the measured colors of the spectroscopically identified stars brighter than @xmath139 and the colors of the irac primary absolute calibrators in reach et al . among the calibrators , a dwarfs are shown with _ filled squares _ and k giants are shown with _ open diamonds_. reach et al . found a displacement between the calibration derived from a and k stars , especially in the first two irac channels . the reason for this is unclear and the authors decided to adopt an absolute calibration based on only a stars . in figure [ fig_star_color ] we see a gradual drift of the irac colors for our data with respect to the irac calibrators , from blue to red in the @xmath131 color . however , since the irac absolute calibration is based on a stars , it is more appropriate for us to compare a stars only . at the very blue end where we find three of the four a - type irac calibrators , our observed colors for the stars agree excellently with the irac calibrators . the largest offset is observed in the @xmath133 color ( _ top _ panel ) , where the stars in the left appear to be @xmath140 magnitude bluer than the mean color of the a - type calibrators ( measured from stars with @xmath29-axis values @xmath141 ) . we thus conclude that our colors are well calibrated , with systematic errors no larger than @xmath115 . despite all the tests above , the irac fluxes derived by our realclean procedures are not perfect . the psfs we used for realclean can only effectively remove fluxes within @xmath142 the fwhm . there are about 10 or so very bright stars with psf wings , which are much broader than this . objects close to these bright stars , especially those on the diffraction spikes and crosstalk features , have systematically higher irac fluxes . a more subtle effect is seen in a few very crowded regions where more than three objects heavily overlap . in such configurations objects that are close to the centers of the groups of objects can have systematically higher fluxes . visual inspections identify only a handful of such examples in each irac band . lrrrrrrrrrrrr 54739 & 189.104027 & 62.278061 & 6032.930 & 0.229 & 2741.866 & 0.936 & 1662.465 & 0.632 & 1160.937 & 1.426 & 611.762 & 0.790 + 54740 & 189.054105 & 62.283076 & 94.032 & 0.235 & 82.640 & 0.158 & 53.823 & 0.203 & 56.101 & 0.860 & 38.938 & 0.841 + 54741 & 188.919686 & 62.287091 & 0.286 & 0.066 & 0.000 & 0.000 & 0.000 & 0.000 & 0.000 & 0.000 & 0.000 & 0.000 + 54742 & 189.482580 & 62.282344 & 54.345 & 0.233 & 31.454 & 0.173 & 23.463 & 0.171 & 17.371 & 0.468 & 9.386 & 0.356 + 54743 & 189.139434 & 62.284211 & 35.129 & 0.165 & 21.971 & 0.136 & 16.208 & 0.214 & 13.119 & 0.402 & 6.451 & 0.718 + 54744 & 188.774407 & 62.286231 & 1.029 & 0.213 & 0.000 & 0.000 & 0.000 & 0.000 & 0.000 & 0.000 & 0.000 & 0.000 + 54745 & 189.178137 & 62.286029 & 2.824 & 0.173 & 2.823 & 0.110 & 2.159 & 0.130 & 2.149 & 0.361 & 2.311 & 0.507 + 54746 & 189.747513 & 62.286519 & 0.184 & 0.075 & 0.000 & 0.000 & 0.000 & 0.000 & 0.000 & 0.000 & 0.000 & 0.000 + 54747 & 188.729697 & 62.286591 & 0.335 & 0.142 & 0.000 & 0.000 & 0.000 & 0.000 & 0.000 & 0.000 & 0.000 & 0.000 + 54748 & 189.369137 & 62.286251 & 2.069 & 0.103 & 2.484 & 0.121 & 1.975 & 0.133 & 1.891 & 0.432 & 1.085 & 0.262 + 54749 & 189.193693 & 62.286347 & 0.830 & 0.117 & 0.985 & 0.043 & 0.762 & 0.051 & 0.000 & 0.303 & 0.000 & 0.000 + 54750 & 189.071498 & 62.286398 & 1.692 & 0.110 & 2.196 & 0.094 & 1.674 & 0.117 & 0.894 & 0.353 & 0.752 & 0.254 we combined the irac fluxes derived with the realclean procedures with the @xmath0 source catalog to form the final multiband catalog . the catalog is presented in table [ master_cat ] . the total number of sources is identical to the sextractor @xmath0 catalog , but not every @xmath0 source contained a detection in the irac bands . as is the case for the @xmath0 fluxes , our irac flux errors are quite realistic ( see [ sec_overlap_region ] and figure [ fig_z_color1 ] ) , and we consider the irac sources with s / n @xmath81 to be undetected . table [ tab_source_number ] summarizes the number of detections and upper limits in the multiband catalog . we have searched for missing sources in the realcleaned irac images . there may exist sources that are so red that they are detected in the irac channels but not in the @xmath0 image . these sources are mostly too faint to be detected in the wircam @xmath0 image . a small fraction of them are blended with nearby sources in the @xmath0 image and not extracted by sextractor . to search for such sources , we combined the realcleaned ( those in the right hand side of figure [ fig_subtraction ] ) 3.6 and 4.5 @xmath5 m irac images and masked the @xmath0 detected objects . this naturally excludes sources that are too close to bright objects . we visually inspected the realcleaned images and look for rermaining sources there . we excluded sources that are likely cosmic rays in the irac images , or other kinds of artifacts . we also examined the 5.6 and 8.0 @xmath5 m realcleaned images , and we found that all the sources seen in these two images can be extracted from the 3.6 + 4.5 @xmath5 m realcleaned images . thus , the positions of such sources were measured in the 3.6 + 4.5 @xmath5 m image and their fluxes were measured in all the individual irac images . the @xmath0-band fluxes and flux errors are measured in the same way as described in [ k_catalog ] . the positions , @xmath0 fluxes , and irac fluxes of a total of 358 such sources are listed in table [ remain_cat ] . all of them have detections in at least two irac bands , and 112 of them are detected in all four bands . as can be seen in figure [ fig_subtraction ] , source extractions in the realcleaned images are difficult and can be severely affected by the imperfect source subtraction described in [ sec_residual ] . we warn the readers that table [ remain_cat ] is very likely an _ incomplete _ list of @xmath0-faint irac sources . lcccc @xmath0 & 94951 & 75187 & 1784 & 0.18 + 3.6 @xmath5 m & 16950 & 15018 & 731 & 0.11 + 4.5 @xmath5 m & 16631 & 14015 & 1000 & 0.12 + 5.8 @xmath5 m & 16870 & 8481 & 4551 & 0.42 + 8.0 @xmath5 m & 15184 & 7245 & 4118 & 0.42 rrrrrrrrrrrrr 189.049062 & 62.133903 & 0.227 & 0.091 & 0.945 & 0.128 & 0.894 & 0.160 & 1.597 & 0.625 & 1.049 & 0.385 + 189.086443 & 62.134109 & 0.000 & 0.108 & 0.400 & 0.127 & 0.480 & 0.151 & 0.000 & 0.247 & 0.000 & 0.614 + 189.049452 & 62.138386 & 0.000 & 0.074 & 0.377 & 0.078 & 0.110 & 0.080 & 1.870 & 0.560 & 0.741 & 0.377 + 189.209527 & 62.140676 & 0.000 & 0.114 & 0.908 & 0.181 & 1.037 & 0.191 & 1.419 & 0.731 & 1.270 & 0.541 + 189.215522 & 62.144264 & 0.000 & 0.094 & 0.638 & 0.212 & 1.049 & 0.190 & 0.000 & 0.284 & 1.507 & 0.590 + 189.241636 & 62.145789 & 0.250 & 0.082 & 0.423 & 0.169 & 0.589 & 0.118 & 0.590 & 0.402 & 0.000 & 0.282 + 189.246272 & 62.146429 & 0.217 & 0.105 & 0.468 & 0.151 & 0.623 & 0.122 & 0.377 & 0.324 & 0.000 & 0.258 + 189.087076 & 62.150220 & 0.000 & 0.087 & 0.756 & 0.232 & 0.366 & 0.211 & 0.696 & 0.300 & 0.000 & 0.425 + 188.998232 & 62.150675 & 0.000 & 0.095 & 0.269 & 0.117 & 0.581 & 0.159 & 0.000 & 0.337 & 0.426 & 0.335 + 189.018589 & 62.153052 & 0.192 & 0.101 & 0.345 & 0.112 & 0.277 & 0.129 & 0.000 & 0.496 & 0.769 & 0.557 + 189.045539 & 62.154243 & 0.000 & 0.093 & 0.572 & 0.129 & 0.511 & 0.146 & 1.249 & 0.549 & 0.000 & 0.438 + 188.981101 & 62.154402 & 0.254 & 0.081 & 1.280 & 0.198 & 2.046 & 0.210 & 3.184 & 0.922 & 2.433 & 1.231 rrrrr 54739 & -0.839 & -1.434 & -1.807 & -2.430 + 54740 & -0.135 & -0.628 & -0.576 & -0.884 + 54741 & 99.000 & 99.000 & 99.000 & 99.000 + 54742 & -0.579 & -0.928 & -1.283 & -1.879 + 54743 & -0.517 & -0.862 & -1.095 & -1.537 + 54744 & 99.000 & 99.000 & 99.000 & 99.000 + 54745 & 0.154 & -0.067 & 0.166 & 0.174 + 54746 & 99.000 & 99.000 & 99.000 & 99.000 + 54747 & 99.000 & 99.000 & 99.000 & 99.000 + 54748 & 0.365 & 0.190 & -0.273 & -0.473 + 54749 & -0.659 & -0.817 & 99.000 & 99.000 + 54750 & 0.283 & -0.006 & -0.249 & -0.316 in the previous subsections , we described our realclean method of obtaining robust color measurements , for overcoming the issues of the very different psfs and the blending of sources . here we describe an alternative method that can better handle the former issue but is more vulnerable to the latter . we started with the background subtracted and registered images used for the realclean procedure . for each single - epoch irac image , we convolved the wircam @xmath0 image with the irac psf and the irac image with the wircam psf . we refer to this procedure as cross - convolution ( xconv ) . we then used the 2-image mode of sextractor to detect sources in the unconvolved wircam image and to measure the wircam and irac fluxes in the xconved images at the locations of the detected wircam sources . the psfs in the xconved images have fwhms of slightly larger than @xmath108 . after experimenting with various photometry apertures , we found that a @xmath80 aperture gives the best compromise for s / n , inclusion of most fluxes from a wide range of source structure , and blending of sources . here we focus on the @xmath0-to - irac colors measured this way , and we do not attempt to measure total fluxes . the advantage of the xconv method is clear : the psfs are ideally matched and deconvolution of the irac psfs is not required ( unlike in the realclean method ) . the disadvantage is the severely degraded @xmath0 image resolution , which prevents the extraction of information on both blended sources and faint sources . because of these , we consider the xconv colors of relatively bright and isolated sources to be more robust than the realclean colors . however , as shown in figure [ fig_compare_colors ] , the xconv colors and the realclean colors turn out to be remarkably consistent . the systematic offsets in figure [ fig_compare_colors ] between these two colors for @xmath143 @xmath144 @xmath5jy objects are 0.002 , -0.024 , 0.031 , and 0.030 ( from 3.6 to 8.0 @xmath5 m ) . we examined the small number of objects in the figure that have offsets of @xmath145 at @xmath144 @xmath5jy . they are either close to bright objects ( a weakness of the xconv method ) or are in regions with multiple intermediate brightness objects ( a weakness of the realclean method , as mentioned in [ issues ] ) . the excellent consistency between the xconv colors and the realclean colors further shows that our realclean method is robust . we thus consider the realclean catalog as the primary product of this work , and we recommend it for most applications . on the other hand , since the xconv colors have unique advantages in their own right , we also provide the xconv colors in table [ xconv_cat ] . we remind the reader that the xconv colors are less reliable on faint or blended sources . such colors should be adopted with caution . here we provide general descriptions of the properties of our @xmath0 and irac realclean catalogs . in subsequent papers we will present scientific analyses on high - redshift galaxies based on this catalog and other multiwavelength data . the goods - n has extremely rich multiwavelength data and several catalogs have been published of sources in this region . here we summarize the overlap between these catalogs and our @xmath0 selected catalog . in the goods - n and its flanking fields , the two most important optical observations are the _ hst _ acs imaging of the goods - n @xcite and the subaru suprimecam imaging of the @xmath146 deg field around the goods - n @xcite . in the 184 arcmin@xmath2 _ hst _ acs region , there are 16326 @xmath0 selected sources with s / n @xmath99 in our catalog and 39432 @xmath107 ( f850lp ) selected acs sources in the goods v2.0 catalog . with @xmath147 search radii ( @xmath148 seeing fwhm ) , 14291 ( 87.5% ) @xmath0 sources have at least one acs counterpart . the number increases to 14883 ( 91.2% ) when the search radii are @xmath23 . it increases much more slowly beyond this . among the 1443 optically faint @xmath0 sources , 547 of them are detected at and 3.6 @xmath5 m at @xmath149 . these are extremely unlikely spurious sources and appear to be truly optically faint nir objects . the rest of the optically faint @xmath0 sources are either outside the irac region ( 407 sources ) or do not have irac counterparts ( 489 sources ) . the later case may be an indication of the existence of spurious sources in our @xmath0 catalog ( see [ sec_spurious ] ) . we can also reverse the direction of the counterpart search . among the 39432 acs sources , 16982 ( 43.1% ) have at least one @xmath0 counterpart within @xmath23 . the number decreases to 14853 ( 37.7% ) for @xmath147 search radii . these show not only that the @xmath0 depth is insufficient for detecting most of the acs sources , but also that the @xmath0 resolution is too low for separating some close pairs of galaxies in the acs image . @xcite released a subaru suprimecam catalog of the goods - n region covering an area of 713 arcmin@xmath2 with 48858 @xmath150+@xmath107 selected sources . in the same region , there are 51649 @xmath0 selected sources with s / n @xmath99 , among which 73.2% have counterparts in the subaru catalog within @xmath151 . the catalog of @xcite adopted a tighter s / n cut than our @xmath0 catalog . this partially explains why the fraction with optical counterparts here is much lower than in the acs case , despite the fact that the subaru images are very deep . the large scattering halos around bright stars in the subaru image create `` holes '' in the catalog , and @xmath0 sources do not have optical counterparts in such holes . the x - ray _ chandra _ deep field - north ( cdf - n ) is fully covered by our @xmath0 imaging . 2 ms imaging @xcite covers an area of approximately 420 arcmin@xmath2 with an uneven sensitivity distribution . there are 503 x - ray sources in the 2 ms catalog , 493 ( 98.0% ) of which have @xmath0 source counterparts within @xmath152 search radii . among the 10 @xmath0-faint sources , four are close to the edge of relatively bright objects . the brighter @xmath0 objects may have affected the detection of the x - ray counterparts at @xmath0 , or themselves may be the correct counterparts . to check whether the rest 6 are spurious x - ray sources , we look for their counterparts at other wavelengths . two @xmath0-faint x - ray source ( source # 151 and # 183 in the catalog of @xcite ) have faint irac counterparts and are extremely faint in the optical . they are even close to two submillimeter sources detected at 1.1 mm ( azgn10 amd azgn12 in @xcite ) . another ( # 88 ) has a faint optical counterpart but is outside the irac region . these three are therefore genuinely @xmath0-faint x - ray sources . two @xmath0-faint x - ray sources ( # 248 and # 371 ) are undetected in the subaru optical , _ spitzer _ 3.6 to 24 @xmath5 m , and vla 1.4 ghz images . even in the x - ray images they appear to be quite faint . they are more likely spurious x - rayt sources . finally , one x - ray source ( # 427 ) is not detected in any of the optical to radio images we have , but its detection appears quite convincing in the hard x - ray band . its nature is undetermined . the _ spitzer _ goods legacy program also imaged the goods - n field with multi - band imaging photometer for _ spitzer _ ( mips , @xcite ) at 24 @xmath5 m . the goods data release 1 + provided a 24 @xmath5 m catalog that is highly complete at 80 @xmath5jy , containing 1199 sources . with search radii of @xmath152 , 1182 mips sources have counterparts in our @xmath0 catalog . the rest are all in low s / n regions in the mips images and appear to be spurious . thus , our @xmath0 image seems to detect all @xmath153 @xmath5jy 24 @xmath5 m sources , bearing in mind the possibility that some sources could be misidentified by the large search radii . finally , at the long wavelength side , there exist deep vla 1.4 ghz imaging in the goods - n @xcite . @xcite re - reduced the vla data and provided improved images and catalog . the catalog of biggs & ivison contains 537 @xmath154 @xmath5jy ( @xmath155 ) 1.4 ghz sources within a radius of @xmath156 centered at the hdf - n with a non - uniform sensitivity distribution . a total of 450 radio sources are within the @xmath157 region of our @xmath0 image , 439 ( 97.5% ) of which have @xmath0 counterparts within @xmath79 of search radii . the 11 @xmath0-faint radio sources do not appear to be spurious in the radio image , although some of them are very close to the @xmath158 limit at 1.4 ghz . they are all in the outer shallower region of the @xmath0 image . we note that , in the outer part of the image , the @xmath0 sensitivity decreases dramatically with exposure time ( figure [ fig_layout ] ) , while the radio sensitivity decreases slowly with the primary beam of the vla . we therefore conclude that the central part of our @xmath0 image is sufficiently deep for detecting nearly all of the 30 @xmath5jy radio sources . in figure [ fig_color_mag-1 ] we present color - magnitude diagrams for sources selected in each of the four irac bands . the most apparent feature is the color bimodality in the @xmath131 and @xmath133 colors . there are two distinct , nearly horizontal branches . the blue branches contain most of the stars and low - redshift galaxies . the redshift divisions in @xmath131 and @xmath133 are @xmath159 and @xmath160 , respectively . at these two redshifts , galaxies quickly move across the two branches , because of the rest - frame 1.6 @xmath5 m sed bump in galaxies . this can also be seen in figure [ fig_z_color1 ] , which is based on our spectroscopic redshifts . the division at @xmath160 in the @xmath133 color is much more apparent in figure [ fig_color_mag-1 ] because figure [ fig_z_color1 ] suffers a bit from the spectroscopic redshift desert at @xmath161 . the bimodality nearly disappears in the @xmath162 color and entirely disappears in the @xmath163 color . the main reason is that strong pah emission from low - redshift starbursts significantly reddens the colors . a minor reason is the relatively lower s / n at these longer irac bands . in figure [ fig_color_mag-2 ] we show the color - magnitude diagrams for the x - ray , 24 @xmath5 m , and 1.4 ghz selected sources discussed in the previous subsections . the @xmath131 colors of all three populations lie in a well - defined area of the color - magnitude space . the distributions are determined by the sensitivities at x - ray , 24 @xmath5 m and 1.4 ghz , plus loose correlations between luminosities at these wavebands ( tracing star formation and black hole accretion ) and the rest - frame 14 @xmath5 m luminosities ( tracing stellar mass ) . at bright magnitudes the sources are at low redshifts ( the blue branch ) , while at the fainter fluxes the sources are at @xmath164 and the sources move to the red branch . the trend is the weakest for x - ray agns ( open squares in figure [ fig_color_mag-2]a ) , but it is still observable . the distributions in the @xmath133 and @xmath162 colors of these three populations are more or less similar , and we do not show them here . however , the @xmath163 color - magnitude diagrams become significantly different . a group of red @xmath163 sources appear at the bright end . these colors are produced by strong pah emission from low - redshift starbursting galaxies . they are also seen in the general 8 @xmath5 m selected diagram ( figure [ fig_color_mag-1]d ) . it is interesting to note the difference between the low - luminosity x - ray sources ( x - ray luminosity @xmath165 erg s@xmath166 , dots in figure [ fig_color_mag-2]b ) and the x - ray agns ( @xmath167 erg s@xmath166 , squares in figure [ fig_color_mag-2]b ) . in the _ chandra _ 2 ms sample , none of the x - ray agns have red @xmath163 colors of @xmath168 , because the strong radiation field of agns can destroy pah molecules . in b08 we presented @xmath133 vs. @xmath163 and @xmath169 vs. @xmath170 color - color diagrams along with our spectroscopic sample . with these two irac color - color diagrams , we analyzed efficiencies and contaminations of selections of @xmath171 and @xmath172 galaxies as well as agns . since the quality of the irac photometry is improved here , it is interesting to revisit these issues . here we focus on the selections of high - redshift galaxies and x - ray agns . ( b08 also included the most powerful radio sources , but that sample size is very small . ) figure [ fig_color_color1 ] shows @xmath133 vs. @xmath163 color - color diagrams of 8.0 @xmath5 m and x - ray selected sources . the color bimodality shown in figure [ fig_color_mag-1]b is clearly seen here . the _ dashed lines _ are the @xmath171 selection proposed by b08 . we reported in b08 that this selection can pick up 90% of @xmath171 galaxies , but 39% of the galaxies that satisfy this criterion are @xmath173 contaminants . in the 3@xmath3 8.0 @xmath5 m sample here , there are 2096 spectroscopic redshifts , 352 of which are at @xmath171 . the above selection picks up 241 of them , corresponding to a selection completeness of 68% , much lower than that in b08 . on the other hand , the selection only picks up 54 @xmath173 galaxies , corresponding to a selection contamination of 18% , also much lower than that in b08 . the results do not change much if we limit to higher s / n 8 @xmath5 m sources where the measurements of colors are more robust . the inconsistency between the results here and our previous results is most likely a systematic difference in the calibration of the irac fluxes . for example , if we make the galaxies here @xmath174 magnitude redder in the @xmath133 color , then the completeness and contamination both become similar to those in b08 . the above comparison shows that using irac colors to select high - redshift galaxies is extremely sensitive to how galactic colors are measured . this also emphasizes that simply adopting existing irac color selection schemes may be problematic unless one is very careful about the fine details of the color measurements . bearing this in mind , the selection completeness and contamination are then just a matter of fine tuning the selection criteria at the expense of one another . such fine tuning also requires a relatively large sample of spectroscopic redshifts . in addition to the very likely difference in the colors , the basic conclusions about irac color selections on high - redshift galaxies remain unchanged here and in b08 . there are also other high - redshift galaxy selections . the _ dotted lines _ in figure [ fig_color_color1 ] show the @xmath175 selection adopted by @xcite and @xcite for studies of submillimeter galaxies . we found that all of the above conclusions apply to this selection as well . this selection has either incompleteness at high redshift or contaminations at low redshift . figure [ fig_color_color1]b shows x - ray sources and x - ray agns . the irac colors suggested by @xcite to select most broad - line agns are enclosed by _ solid lines_. however , here we see that the stern et al . selection only picks up a fraction of the x - ray agns . by comparing both panels in figure [ fig_color_color1 ] as well as figures [ fig_z_color1]b and d , we see that the same region also selects a larger number of high - redshift galaxies . these galaxies do not show up in the relatively shallow data of ( * ? ? ? * their figure 1 ) . this is consistent with the finding in b08 and @xcite that high - redshift reddened objects occupy similar color spaces to agns . the selection of @xcite does not pick up the majority of agns and has substantial galaxy contamination . b08 also presented @xmath169 vs. @xmath170 color - color diagrams and used them to analyze selections of @xmath172 galaxies and the @xcite selection of agns . as with the @xmath133 vs. @xmath163 color - color diagram , b08 concluded that the irac selection of @xmath172 galaxies can be very complete ( @xmath176 ) , but it has a large contamination ( @xmath177 ) from low - redshift galaxies . b08 also found that the @xcite selection of agns is incomplete ( though better than the @xcite selection ) and has severe galaxy contamination . we present the updated @xmath169 vs. @xmath170 color - color diagrams in figure [ fig_color_color2 ] , and we confirm all the conclusions made by b08 . to sum up , using irac colors to select high - redshift galaxies can be either incomplete or suffer from contaminations from agns or low - redshift galaxies . the selection can be fine tuned to balance the completeness and contamination , but this requires a good spectroscopic sample and/or extremely well measured irac colors . we also found that galaxies and agns occupy a very similar region of color space . the irac selections of agns based on the @xcite and @xcite methods always include large numbers of galaxy contaminations and have various degrees of incompleteness . the above discussion only includes irac colors . in this work , we presented ultradeep @xmath0 photometry , which can be a powerful tool for studying galactic colors . we show the @xmath178 vs. @xmath133 color - color diagrams for 4.5 @xmath5 m and x - ray sources in figure [ fig_color_color3 ] . here we only adopt the two shortest irac bands for their higher s / n and for their continuous availability in the _ spitzer _ warm phase mission . again , we see that the x - ray agns occupy a color space that is nearly identical to that of galaxies . such a diagram is extremely ineffective for separating galaxies and agns . on the other hand , this color - color diagram is much more structured than the previous irac - only diagrams , suggesting that these colors can provide an excellent redshift diagnostic . figure [ fig_color_color3 - 2 ] shows the same diagram as figure [ fig_color_color3 ] , but only for the spectroscopic subsample drawn from b08 . there is a very subtle difference in the distributions of galaxies in these two diagrams ( 4.5 @xmath5 m selected vs. spectroscopically identified ) , because the spectroscopic identifications suffer from the redshift desert and are somewhat biased to galaxies with strong emission lines that can have unusual @xmath178 colors ( see [ sec_compare_sed ] ) . however , three distinct redshift groups can be seen in both diagrams : @xmath179 , @xmath71.2 , and @xmath175 . in particular , we find that the following selection criteria are especially effective in selecting intermediate redshifts of @xmath71.2 : @xmath180 this selection is shown as the _ solid lines _ in figure [ fig_color_color3 - 2 ] . in figure [ fig_color_color3 - 2 ] a total of 2834 spectroscopic identifications are included , among which 1079 are at @xmath71.2 . the above @xmath0 and irac selection picks up 870 galaxies at @xmath71.2 , corresponding to a selection completeness of 81% . on the other hand , the selection also picks up 86 galaxies at @xmath179 , 26 galaxies at @xmath175 , and 3 stars , corresponding to a contamination of 11.6% . it is possible to fine tune the selection criteria to achieve higher completeness at the expense of increasing contamination , or vice versa . for example , after we moved the selecting region in the diagram towards the lower - right by 0.1 magnitude along both the @xmath29 and @xmath30 axes , the selection completeness increased to 89% , but the contamination also increased to 17% . @xmath178 and @xmath133 colors can still be measured in the _ spitzer _ warm phase mission and can be used in different ways . the above @xmath71.2 selection picks up galaxies at redshifts just lower than that in the popular @xmath181 selection ( @xmath1372.5 , @xcite ) . galaxies at @xmath175 and @xmath179 can also be selected from the same diagram . in addition , in a subsequent paper , we will use the same diagram to select @xmath182 massive and dusty galaxies ( wang et al . , in preparation ) . generally speaking , the addition of the @xmath0-band data increases the redshift resolution at @xmath183 . at higher n this will greatly help irac - based photometric redshift studies ( e.g. , * ? ? ? * ) , which can be more powerful than simple color selections . on the other hand , in all the examples discussed in this paper , irac colors become much less efficient in separating galaxies at @xmath182 , even after the inclusion of @xmath0 . for these high - redshift galaxies , optical colors or photometric redshifts that combine optical , nir , and irac bands ( e.g. , * ? ? ? * ) may be much better . however , on optically faint objects such as high - redshift submillimeter galaxies ( e.g. , * ? ? ? * ; * ? ? ? * ) , colors from deep nir and irac imaging are valuable in studying their redshifts . we carried out ultradeep @xmath0 band imaging in the hdfn and its flanking field with wircam on the cfht . the mosaic image has a size of @xmath184 , fully covering the goods - n field . with nearly 50 hr of integration , we reached a 90% completeness limit of 1.38 @xmath5jy ( @xmath185 ) in the 0.25 deg@xmath2 region , and 0.94 @xmath5jy ( @xmath186 ) in the 0.06 deg@xmath2 goods - n irac region . the faintest detected objects have @xmath0 fluxes of @xmath187 @xmath5jy ( @xmath188 ) . the photometry is uniform across the entire field , and the variation is well within 2% . in the goods - n region the astrometry is accurate to @xmath65 for compact objects with high s / n . the @xmath0 images and the associated catalogs are available through the internet . we used the @xmath0 image and catalog as priors to measure mir magnitudes in the _ spitzer _ irac 3.6 to 8.0 @xmath5 m bands with our realclean method . we verified the realclean results using data taken in two different epochs of observations , as well as by comparing our colors with the colors of stars and with the colors derived from model seds of galaxies at various redshifts . the realclean method appears to provide reasonably good photometry even on galaxies that have close neighbors within @xmath108 . this also provides a relatively clean way to identify faint irac objects that are not detected in the @xmath0 image . we make the realclean irac catalogs for both @xmath0 detected and undetected objects available to the community . additional to the realclean method , we also measured galactic colors by convolving the @xmath0 image with the irac psfs and convolving the irac images with the @xmath0 psf . the derived colors are consistent with the realclean results . we found our @xmath0 sample to be extremely powerful in studying the multiwavelength properties of high - redshift galaxies . the @xmath0 image detected most ( if not all ) of the x - ray , 24 @xmath5 m , and radio sources published in this field . on the other hand , only 40% of the _ hst _ acs sources in the goods - n are detected at @xmath0 at the current depth . we studied various @xmath0 and irac color - magnitude and color - color diagrams . because of the rest - frame 1.6 @xmath5 m bump in the seds of galaxies , galaxies show significant color evolution in the @xmath0 and irac bands from @xmath189 to @xmath183 . we found that using irac colors to select galaxies at certain redshifts is extremely sensitive to the color measurements and that balancing selection completeness and contamination will require careful tuning of the selection criteria with a good spectroscopic sample . we discussed an effective selection of @xmath71.2 galaxies using the @xmath0 band and the two shorter irac bands that will remain available in the _ spitzer _ warm phase mission . we also confirmed our previous studies ( b08 ) that the irac agn selections in @xcite and @xcite are either incomplete at the low - luminosity end or suffer severely from galaxy contamination . we thank the referee for comments that greatly improve the manuscript . we are grateful to the cfht staff for help on obtaining the data , c .- h . yan for providing useful parameters about wircam , and b .- c . hsieh and l. lin for useful discussions . this work started when w.h.w . was a jansky fellow at the national radio astronomy observatory . we gratefully acknowledge support from nrao ( w.h.w . ) , the national science council of taiwan grant 98 - 2112-m-001 - 003-my2 ( w.h.w . ) , nsf grants ast 0709356 ( l.l.c . ) and ast 0708793 ( a.j.b . ) , the university of wisconsin research committee with funds granted by the wisconsin alumni research foundation ( a.j.b . ) , and the david and lucile packard foundation ( a.j.b . ) . alexander , d. m. et al . 2003 , , 126 , 539 anderson , j. , & king , i. r. 2003 , , 115 , 113 barger , a. j. , cowie , l. l. , & wang , w .- h . 2008 , , 689 . 687 ( b08 ) bertin , e. , & arnouts , s. 1996 , , 117 , 393 biggs , a. d. & ivison , r. j. 2006 , , 371 , 963 brinchmann , j. , & ellis , r. 2000 , , 536 , l77 calzetti , d. , armus , l. , bohlin , r. c. , kinney , a. l. , koornneef , j. , & storchi - bergmann , t. 2000 , , 533 , 682 capak , p. , et al . 2004 , , 127 , 180 capak , p. , et al . 2007 , , 172 , 99 cohen , m. , wheaton , wm . a. , & megeath , s. t. 2003 , , 126 , 1090 cornwell , t. , braun , r. , & briggs , d. s. 1999 , in asp conf . 180 , synthesis imaging in radio astronomy ii , e.d . g. b. taylor , c. l. carilli , & r. a. perley , 151 cowie , l. l. , gardner , j. p. , hu , e. m. , songaila , a. , hodapp , k .- w . , & wainscoat , r. j. 1994 , , 434 , 114 cowie , l. l. , & barger , a. j. 2008 , , 686 , 72 cutri , r. m. , et al . 2003 , explanatory supplement to the 2mass all sky data release ( washington : nasa ) , http://www.ipac.caltech.edu/2mass/releases/allsky/doc/explsup.html daddi , e. , cimatti , a. , renzini , a. , fontana , a. , mignoli , m. , pozzetti , l. , tozzi , p. , & zamorani , g. 2004 , , 617 , 746 devlin , m. j. , et al . 2009 , , 458 , 737 elsner , f. , feulner , g. , & hopp , u. 2008 , , 477 , 503 fazio , g. g. , et al . 2004 , , 154 , 10 fontana , a. , et al . 2006 , , 459 , 745 giavalisco , m. et al . 2004 , , 600 , l93 grazian , a. , et al . 2006a , , 449 , 951 hatziminaoglou , e. , et al . 2005 , , 129 , 1198 hgbom , j. a. 1974 , , 15 , 417 kajisawa , m. , et al . 2006 , , 58 , 951 kajisawa , m. , et al . 2009 , , 702 , 1393 kauffmann , g. , & charlot , s. 1998 , , 297 , l23 keenan , r. c. , trouille , l. , barger , a. j. , cowie , l. l. , & wang , w .- h . 2010 , , 186 , 94 kinney , a. l. , calzetti , d. , bohlin , r. c. , mcquade , k. , storchi - bergmann , t. , & schmitt , h. r. 1996 , , 467 , 38 lacy , m. , et al . 2004 , , 154 , 166 laidler , v. g. , et al . 2007 , , 119 , 1325 mancini , c. , matute , i. , cimatti , a. , daddi , e. , dickinson , m. , rodighiero , g. , bolzonella , m. , & pozzetti , l. 2009 , , 500 , 705 marsden , g. , et al . 2009 , , in press ( arxiv:0904.1205 ) perera , t. a. , et al . 2008 , , 391 , 1227 prez - gonzlez , p. g. , et al . 2008 , , 686 , 72 pope , a. , et al . 2006 , , 370 , 1185 richards , e. a. 2000 , , 533 , 611 reddy , n. a. , steidel , c. c. , erb , d. k. , shapley , a. e. , & pettini , m. 2006 , , 653 , 1004 reach , w. t. , et al . 2005 , , 117 , 978 rieke , g. , et al . 2004 , , 154 , 25 sajina , a. , lacy , m. , & scott , d. 2005 , , 621 , 256 sawicki , m. 2002 , , 124 , 3050 silva , l. , granato , g. l. , bressan , a. , & danese , l. 1998 , , 509 , 103 simpson , c. , & eisenhardt , p. 1999 , , 111 , 691 skrutskiee , m. f. , et al . 2006 , , 131 , 1163 stern , d. , et al . 2005 , , 631 , 163 wang , w .- h . , cowie , l. l. , van sader , j. , barger , a. j. , & williams , j. p. 2007 , , 670 , l89 wang , w .- h . , barger , a. j. , & cowie , l. l. , 2009 , , 690 , 319 younger , j. d. , et al . 2009 , , 704 , 803 yun , m. s. , et al . 2008 , , 389 , 333	we present an ultradeep @xmath0-band image that covers @xmath1 deg@xmath2 centered on the great observatories origins deep survey - north ( goods - n ) . the image reaches a 5 @xmath3 depth of @xmath4 in the goods - n region , which is as deep as the goods - n _ spitzer _ infrared array camera ( irac ) 3.6 @xmath5 m image . we present a new method of constructing irac catalogs that uses the higher spatial resolution @xmath0 image and catalog as priors and iteratively subtracts fluxes from the irac images to estimate the irac fluxes . our iterative method is different from the @xmath6 approach adopted by other groups . we verified our results using data taken in two different epochs of observations , as well as by comparing our colors with the colors of stars and with the colors derived from model spectral energy distributions ( seds ) of galaxies at various redshifts . we make available to the community our wircam @xmath0-band image and catalog ( 94951 objects in 0.25 deg@xmath2 ) , the interactive data language ( idl ) pipeline used for reducing the wircam images , and our irac 3.6 to 8.0 @xmath5 m catalog ( 16950 objects in 0.06 deg@xmath2 at 3.6 @xmath5 m ) . with this improved @xmath0 and irac catalog and a large spectroscopic sample from our previous work , we study the color - magnitude and color - color diagrams of galaxies . we compare the effectiveness of using @xmath0 and irac colors to select active galactic nuclei ( agns ) and galaxies at various redshifts . we also study a color selection of @xmath71.2 galaxies using the @xmath0 , 3.6 @xmath5 m , and 4.5 @xmath5 m bands .
over the years knot theory has worked with knots and links in the three - dimensional sphere @xmath2 . however , together with improving knowledge about 3-manifolds , great attention has been paid to knots and links in manifolds different from @xmath2 : seifert fiber spaces @xcite ( in particular , lens spaces @xcite ) , thickened surfaces @xcite , arbitrary 3-manifolds @xcite . in this paper we focus on knots and links in lens spaces @xmath0 with @xmath1 . these manifolds are the simplest ones ( beside @xmath2 ) where we can study knots and links since every link in @xmath3 can be presented by integer @xmath4-surgery over the unknot ( in cases of other manifolds the surgery link can be much more difficult ) . moreover , there are interesting articles explaining applications of knots in lens spaces to other fields of science : @xcite exploits them to describe topological string theories and @xcite uses them to describe the resolution of a biological dna recombination problem . a lot of link invariants can be generalized from links in @xmath2 to links in @xmath0 . kauffman bracket skein module @xcite , knot floer homology @xcite , homfly - pt polynomial @xcite all have analogues in lens spaces . despite the fact that some invariants can be generalized to links in lens spaces , sometimes it is very difficult to use them . for example , the fundamental quandle of a link which has a very simple topological description for links in @xmath2 can be easily generalized to links in @xmath0 . however , an explicit procedure to write down a presentation of a fundamental quandle for links in @xmath0 by generators and relations directly from a ( band , disc , grid ) diagram is known only in the case @xmath5 ( see @xcite ) , so it is almost impossible to use this invariant . another disadvantage of the ordinary fundamental quandle for links in @xmath0 is that the fundamental quandle of the link @xmath6 is isomorphic to the fundamental quandle of its lift @xmath7 , where @xmath8 denotes the universal covering map . so we can not distinguish links with equivalent liftings . such invariants are called _ inessential _ @xcite . in this paper we define the notion of virtual quandle for links in @xmath3 . virtual quandles were firstly introduced in @xcite as generalizations of quandles for virtual links . even if in the present paper we do not work with virtual links , we use the name `` virtual quandle '' since our invariant has the same algebraic structure . the paper is organized as follows . in section [ band ] we review the definition of band diagrams for links in lens spaces , while in section [ quandle ] we recall those of quandle and virtual quandle , proving some results that will be useful in the following . in section [ s33 ] we construct a virtual quandle for links in @xmath3 and prove that it is an invariant ( see theorem [ tt2 ] ) . finally , in section [ exex ] we investigate some properties of this invariant , proving that it is essential ( see proposition [ esss ] ) . in this section we recall the notion of band diagrams for links in lens spaces ( which were firstly introduced in @xcite ) and recall how reidemeister moves for such band diagrams look like . we start by giving the definition of dehn surgery in @xmath2 . let @xmath9 be a knot in @xmath2 , denote by @xmath10 a closed tubular neighborhood of @xmath9 and let @xmath11 be a simple closed curve on @xmath12 . _ the dehn surgery on @xmath9 along @xmath11 _ is the manifold @xmath13 obtained by gluing @xmath14 with a solid torus @xmath15 along their boundaries via a homeomorphism which identifies @xmath11 with the boundary of a meridian disc . the curve @xmath11 is called the _ slope _ of the surgery and the homeomorphism type of the resulting manifold depends only on the homology class of @xmath11 in @xmath12 , up to orientation change . we fix a base @xmath16 for @xmath17 , such that @xmath18 is a meridian of @xmath9 and @xmath19 has algebraic intersection 1 with @xmath18 . if the homology class of @xmath11 is @xmath20 , with @xmath21 , we call the surgery _ rational _ and we say that it has _ framing index _ @xmath22 . every link @xmath23 in a manifold @xmath13 obtained via a @xmath22-surgery over @xmath9 , can be presented by a diagram of a link @xmath24 in @xmath2 , where the knot @xmath9 is equipped with a number @xmath25 , the surgery along @xmath9 gives @xmath13 and @xmath23 is the image of @xmath26 under the surgery operation . this presentation of @xmath23 is called _ a mixed link diagram _ and the knot @xmath9 is called the _ surgery knot_. in order to simplify notation we use the same symbol @xmath23 to denote both @xmath26 and @xmath23 . the _ lens space _ @xmath0 is the 3-manifold obtained by a rational surgery on the unknot @xmath27 in @xmath2 with framing index @xmath28 . so , a link @xmath23 in @xmath0 can be described by a mixed link diagram of the link @xmath29 ( see , for example , figure [ mixxx ] ) . in order to construct a band diagram for a link @xmath23 in @xmath0 we present @xmath2 as a one - point compactification of @xmath30 and fix coordinates @xmath31 in @xmath30 such that the surgery knot ( which is the unknot in @xmath2 ) is described as the @xmath32 axis in @xmath30 . a regular orthogonal projection of a mixed link diagram onto the plane @xmath33 is called _ a punctured disc diagram _ of @xmath23 ( see figure [ punban ] on the left ) . in a punctured disc diagram the surgery knot is depicted as a puncture . a _ band diagram _ for a link @xmath23 in @xmath0 can be obtained from a punctured disk diagram by ( i ) cutting a disc containing the punctured disk diagram and centered in the pucture along a half - line starting from the puncture and avoiding the crossings of @xmath23 and ( ii ) deforming the resulting annulus into a rectangle ( see figure [ punban ] on the right ) . clearly , an orientation of the link induces an orientation on the resulting band diagram . [ punban ] the intersection points of the link with the left ( resp . right ) side of the band diagram are called _ left ( resp . right ) boundary points_. left and right boundary points together are called _ boundary points _ of a band diagram . two band diagrams are equivalent under the three classical reidemeister moves @xmath34 , @xmath35 , @xmath36 inside the rectangle and a global move called _ the @xmath37-move _ ( see @xcite ) . in the case of lens spaces @xmath3 the @xmath37-move is depicted in figure [ sl ] . in this section we review some facts about quandles and virtual quandles from the algebraic point of view . a _ quandle _ is a set @xmath38 endowed with a binary algebrabraic operation @xmath39 satisfying the following axioms : 1 . for all @xmath40 @xmath41 ; 2 . for all @xmath40 the map @xmath42 is a bijection on @xmath38 ; 3 . for all @xmath43 @xmath44 . here for simplicity we use symbols @xmath45 and @xmath46 to denote @xmath47 and @xmath48 respectively . the notion of a quandle was first introduced by joyce @xcite and , independently , matveev @xcite as an invariant for links in the @xmath49-sphere . after that quandles were also studied from an algebraic point of view ( see , for example , @xcite ) . for a group @xmath50 and a number @xmath51 denote by @xmath52 the quandle having the same underlying set as @xmath50 and with operation @xmath53 . this quandle is called the _ @xmath54-th conjugacy quandle . _ a bijection @xmath55 is called an _ automorphism _ of a quandle @xmath38 if for all @xmath56 the equality @xmath57 holds . in particular , from the second and the third axioms follows that the map @xmath58 is an automorphism of @xmath38 for each @xmath40 . the group generated by @xmath58 for all @xmath40 is called _ the group of inner automorphisms of @xmath38 _ and is denoted by @xmath59 . we denote by @xmath60 the image of the element @xmath61 under the map @xmath62 : this notation should not suggest that @xmath63 is an element of @xmath38 , however it is practically convenient since @xmath64 for all @xmath65 . from the third axiom of quandles for all @xmath65 we have @xmath44 , and replacing @xmath66 by @xmath67 , we obtain the equality @xmath68 if an element of a quandle @xmath38 has the form @xmath69 for @xmath40 and @xmath70 , then we say that the element @xmath66 is on the _ primary level _ and the element @xmath71 is on the _ operator level_. equality ( [ nomore ] ) says that we do not need more than these two levels and so every element of a quandle has the form @xmath69 . moreover , if @xmath72 for all @xmath40 then we use the notation @xmath73 and call this expression _ the operator relation _ ( see @xcite ) . a _ virtual quandle _ is a set @xmath74 endowed with a binary algebraic operation @xmath39 and a unary algebraic operation @xmath75 satisfying the following axioms : 1 . @xmath74 with the operation @xmath39 is a quandle ; 2 . @xmath76 is a bijection of @xmath74 such that @xmath77 . let @xmath50 be a group , @xmath51 and @xmath78 be a fixed element of @xmath50 . if we add to the quandle @xmath52 the operation @xmath79 , then we obtain a virtual quandle . a bijection @xmath80 is an _ automorphism _ of the virtual quandle @xmath74 if for all @xmath81 we have @xmath57 and @xmath82 . the selfmap @xmath76 is clearly an automorphism of @xmath74 , moreover @xmath76 belongs to the center of the group @xmath83 . on the contrary the bijection @xmath58 is not necessarily an automorphism of a virtual quandle . indeed if @xmath58 is an automorphism , then for all @xmath84 we have @xmath85 and @xmath86 . so , @xmath58 is an automorphism of @xmath74 if and only if @xmath87 , for all @xmath88 . if @xmath74 is a virtual quandle , then we denote by @xmath89 the quandle obtained from @xmath74 just forgetting about the operation @xmath76 . every automorphism of @xmath74 induces an automorphism of @xmath89 and every map @xmath58 induces an automorphism of @xmath89 . as for groups , it is possible _ to present ( virtual ) quandle by generators and relations _ as follows . given a set of letters @xmath90 denote by @xmath91 the inductively defined set of words obtained from @xmath92 by all possible subsequent application of the operations @xmath39 , @xmath93 , @xmath75 , @xmath94 . let @xmath95 be some set of words from @xmath91 and @xmath96 be a set of formal equalities . denote by @xmath97 an equivalence relation on @xmath91 which consists of the following relations : 1 . for all @xmath98 @xmath99 ; 2 . for all @xmath100 @xmath101 ; 3 . for all @xmath100 @xmath102 ; @xmath103 for all @xmath104 ; 5 . for all @xmath105 @xmath106 , @xmath107 , + @xmath108 , @xmath109 ; 6 . @xmath110 if the equality @xmath111 belongs to @xmath112 . the quotient @xmath113 with well defined operations @xmath39 and @xmath75 is obviously a virtual quandle . we say that this virtual quandle is _ presented by the set of generators @xmath92 and the set of relations @xmath112 _ and denote it by @xmath114 . every relation in the virtual quandle @xmath74 has the form @xmath115 , where @xmath116 are elements of @xmath74 and @xmath117 belong to the free group @xmath118 generated by @xmath74 . let @xmath76 be an automorphism of a quandle @xmath38 . in general we can not write the equalities @xmath120 and @xmath121 since the expressions @xmath122 and @xmath123 are not defined . however it is easy to show that the following equalities hold @xmath124 moreover , an automorphism @xmath76 of a quandle @xmath38 induces an automorphism @xmath125 of @xmath59 which acts on the generators of @xmath59 by the rule @xmath126 . as a consequence , on the operator level we can write expressions of type @xmath123 and @xmath122 meaning that @xmath127 and @xmath128 . moreover if @xmath66 belongs to the free group @xmath129 generated by the elements of @xmath38 then on the operator level we can write expressions of type @xmath130 . an automorphism @xmath76 of a quandle @xmath38 is said to be an _ @xmath119-splitting automorphism _ of @xmath38 if for every element @xmath131 the equality @xmath132 holds . in other words , @xmath76 is an @xmath119-splitting automorphism of @xmath38 if @xmath133 for all @xmath134 . let @xmath50 be a group , @xmath135 be a subgroup of @xmath50 generated by @xmath119-th powers of elements from @xmath50 and @xmath136 be a fixed element of the quotient . then @xmath79 is an @xmath119-splitting automorphism of the quandle @xmath137 . we end the section proving some useful properties of @xmath119-splitting automorphisms . [ lem1 ] let @xmath138 be an @xmath119-splitting automorphism of a quandle @xmath38 . the following properties hold : 1 . @xmath139 for all @xmath131 ; 2 . @xmath140 for all @xmath131 ; 3 . for every @xmath141 the automorphism @xmath142 is an @xmath119-splitting automorphism of @xmath38 . moreover if @xmath143 , then @xmath144 . \1 ) from the @xmath119-splitting relation we have @xmath145 therefore @xmath139 . \2 ) follows from 1 ) and the @xmath119-splitting relation . \3 ) let @xmath61 be an arbitrary element of @xmath129 , then the following sequence of equalities holds @xmath146 in the last equality we used item 2 ) of the lemma . so @xmath78 is an @xmath119-splitting automorphism of @xmath38 . for every element @xmath40 we have @xmath147 , therefore if @xmath143 , then @xmath144 . note that equality 1 ) of lemma [ lem1 ] does not imply the equality @xmath148 . it just says that @xmath149 for all @xmath134 . in this section we define the notion of the virtual quandle @xmath150 for a link @xmath9 in a lens space @xmath3 and prove that it is invariant under all moves @xmath34 , @xmath35 , @xmath36 , @xmath37 . an _ arc _ of a band diagram of a link @xmath9 in @xmath3 is a connected part of the diagram bounded by two overpasses , or two boundary points , or one overpass and one boundary point . referring to figure [ ban ] , we label ( i ) the arcs which are bounded by ( at least ) one left boundary point by the letters @xmath151 , ( ii ) the arcs which are bounded by ( at least ) one right boundary point by the letters @xmath152 and ( iii ) all the remaining arcs by the letters @xmath153 . note that some of @xmath154 may be identified . choose an orientation of the link and for each right boundary point @xmath155 define the number @xmath156 by the following rule : if the arrow @xmath155 runs from the left to the right , then @xmath157 , otherwise @xmath158 . we define the _ virtual quandle _ @xmath150 of a link @xmath159 as the virtual quandle having the presentation @xmath160 , where : * _ the set of generators @xmath92 _ is @xmath161 ; * _ the set of inner relations @xmath162 _ consists of relations which can be written from the part @xmath163 of the band diagram contained in the interior of the rectangle . here we have ( i ) relations of identifications between some of @xmath164 , @xmath165 , @xmath166 ( if it is necessary ) and ( ii ) as in the classical case , for each crossing , the relation @xmath167 , where @xmath66 , @xmath61 , @xmath168 are the labels of the arcs involved in the crossing as depicted in figure [ crosrule ] ; * _ the set of boundary relations @xmath169 _ consists of the @xmath119 relations @xmath170 for @xmath171 and one operator relation @xmath172 ; * _ the set of splitting relations @xmath173 _ consists of relations which state that @xmath76 is a @xmath4-splitting automorphism of @xmath174 and has order @xmath4 : @xmath175 for all @xmath176 and @xmath177 for all @xmath178 . [ xxx ] let @xmath9 be the knot in @xmath3 , whose band diagram is depicted in figure [ exx4 ] . we have four generators @xmath179 , @xmath180 , @xmath181 , @xmath182 . since both @xmath181 , @xmath182 run from the left to the right , @xmath183 . inner relations are @xmath184 , @xmath185 . boundary relations are @xmath186 , @xmath187 , @xmath188 . splitting relations are @xmath175 , @xmath177 for all @xmath66 . therefore @xmath150 has the following presentation @xmath189 using the relations @xmath184 , @xmath185 we can delete the elements @xmath190 from the set of generators . then @xmath150 can be rewritten as @xmath191 using the relation @xmath192 we can delete @xmath179 from the set of generators . then if we replace the letter @xmath180 just by @xmath66 we have @xmath193 from the relations @xmath194 and @xmath195 we obtain the relation @xmath196 . if @xmath4 is odd , then from the relations @xmath196 and @xmath177 we get @xmath197 and the virtual quandle has the following presentation @xmath198 if @xmath4 is even , then the relation @xmath177 is a consequence of the relation @xmath196 , and the relation @xmath175 can be obtained from the relation @xmath195 . then the virtual quandle has the following presentation . let @xmath27 be the unknot in @xmath3 . a band diagram for @xmath27 is depicted in figure [ exx45 ] . here we have only one generator @xmath66 and only splitting relations , so @xmath200 [ xx ] let @xmath23 be the @xmath119-component link in @xmath3 , whose band diagram is depicted in figure [ exx6 ] . we have @xmath119 generators @xmath164 and no inner relations ( @xmath201 ) . boundary relations are @xmath202 , @xmath203 , @xmath204 , @xmath205 and splitting relations are @xmath175 and @xmath177 for all @xmath66 , so the virtual quandle is @xmath206 from the relations @xmath207 , it follows that @xmath197 for all @xmath66 . then the virtual quandle has the following presentation @xmath208 [ tt2 ] the virtual quandle @xmath150 is an invariant for links in @xmath3 . in order to prove that the virtual quandle is an invariant for links in @xmath3 we need to prove that it is invariant under all the reidemeister moves @xmath34 , @xmath35 , @xmath36 and the @xmath37-move . the proof of the invariance under the moves @xmath34 , @xmath35 , @xmath36 is the same as in the case of ordinary quandle for links in @xmath2 ( see , for example , the proof of ( * ? ? ? * theorem 3.4 ) ) . so we need to consider only the case of the @xmath37-move . let @xmath9 be a link in lens space @xmath3 whose band diagram is depicted in figure [ ban ] . the virtual quandle @xmath150 has generators @xmath209 and the set of relations @xmath210 . denote by @xmath211 the link having the band diagram which is obtained from the band diagram of @xmath9 by the @xmath37-move ( see figure [ exx69 ] ) . fix an orientation of @xmath9 and take the induced orientation on @xmath211 : note that the arcs of @xmath211 labelled by @xmath212 , @xmath203 , @xmath213 are oriented as the arc of @xmath9 labelled by @xmath181 . the virtual quandle @xmath214 has generators @xmath179 , @xmath203 , @xmath215 , @xmath181 , @xmath203 , @xmath216 , @xmath217 , @xmath203 , @xmath218 , @xmath219 , @xmath203 , @xmath220 , @xmath212 , @xmath203 , @xmath221 , @xmath222 , @xmath203 , @xmath223 . the set of inner relations @xmath224 consists of the set @xmath162 and the following relations @xmath225 the set of boundary relations @xmath226 consists of the following @xmath227 relations @xmath228 and one operator relation @xmath229 the set of splitting relations @xmath230 is the same as the set of splitting relations @xmath173 and consists of the relations @xmath175 , @xmath177 for all @xmath66 . from the set of relations ( [ eq6 ] ) and ( [ eq5 ] ) we see that every element @xmath231 can be expressed via @xmath181 and @xmath232 . therefore we can delete the elements @xmath233 from the set of generators and replace the set of relations ( [ eq6 ] ) and ( [ eq5 ] ) by the relation @xmath234 using boundary relation ( [ eq8 ] ) and the fact that @xmath235 we conclude from ( [ eq80 ] ) that @xmath236 . from inner relations ( [ eq3 ] ) and ( [ eq4 ] ) we see that all the generators @xmath237 can be expressed via @xmath238 , @xmath239 . therefore we can delete all these generators from the generating set changing them by their expressions via @xmath238 , @xmath239 in the relations where they take part . moreover , from the set of relations ( [ eq7 ] ) we can express @xmath238 via @xmath179 ( @xmath240 ) and delete all the elements @xmath238 from the set of generators . after these manipulations we see that @xmath214 is generated by the elements @xmath164 , @xmath239 , @xmath166 which are the same as for @xmath150 . the set of inner relations @xmath224 coincides with the set @xmath162 . the set @xmath226 of boundary relations consists of the following relations @xmath241 and one operator relation @xmath242 using the splitting relation and the equality @xmath243 we can rewrite the boundary relations ( [ eq10 ] ) in the following way @xmath244 by lemma [ lem1 ] we have @xmath245 and using the fact that @xmath246 we can rewrite equality ( [ eq11 ] ) in the form @xmath247 , which is equivalent to the relation @xmath248 . now the only difference between @xmath150 and @xmath214 is in the boundary relations ( [ eq12 ] ) . however if we change @xmath76 by @xmath249 , then we obtain completely the same relations in both quandles since by lemma [ lem1 ] the splitting relations for @xmath76 and for @xmath78 are the same . in this section we investigate some properties of the virtual quandle and we prove that it is an essential invariant . at first , we recall some results on the homology of links in lens spaces ( see ( * ? ? ? * lemma 4 ) and ( * ? ? ? * proposition 2 ) ) . if @xmath9 is an oriented link in @xmath3 with components @xmath250 represented via a band diagram with @xmath119 right boundary points , then the homology class @xmath251 $ ] of @xmath252 in @xmath253 has the following form @xmath254=\sum_{i_j=1_j}^{n_j}\varepsilon_{i_j}~{\rm mod}~p,\ ] ] where @xmath255 are the labels of the right boundary points which correspond to @xmath252 and @xmath256 according to the rule described in the previous section . moreover , @xmath257 where @xmath258,\ldots , [ k_m])$ ] . [ prprpr ] let @xmath159 be a link with components @xmath250 . let @xmath259 be the virtual quandle obtained from @xmath150 by adding relations @xmath260 , for all @xmath261 . then we have @xmath262,p)}(t_j)=t_j , \forall x~~x\equiv1\rangle\ ] ] where @xmath251 $ ] denotes the homology class of @xmath252 in @xmath253 . the virtual quandle @xmath259 is obviously a homomorphic image of @xmath150 and we can easily represent @xmath263 by generators and relations transforming relations from @xmath264 using relations @xmath260 for all @xmath66 . the boundary relation @xmath172 and the splitting relation @xmath175 follow directly from the relation @xmath260 for all @xmath66 . therefore in @xmath263 from the set @xmath169 of boundary relations we have only @xmath186 , @xmath203 , @xmath265 and from the set @xmath173 of splitting relations we have only the relation @xmath177 for all @xmath66 . every inner relation which follows from crossings of the diagram has the form @xmath167 ( see figure [ crosrule ] ) , therefore since @xmath266 in @xmath263 this relation will give us @xmath267 . it means that inner relations just identify all arcs belonging to the same component of a link and lying between two boundary points . if we denote by @xmath268 the generator corresponding to the @xmath269-th component of a link , we obtain the following presentation for @xmath259 . @xmath270 the relations @xmath271 and @xmath272 together are equivalent to the relation @xmath273 which by equality ( [ homj ] ) is equivalent to the relation @xmath274,p)}(t_j)=t_j$ ] , therefore @xmath259 has the following presentation @xmath262,p)}(t_j)=t_j , \forall x~~x\equiv1\rangle\ ] ] and the statement is proved . let @xmath8 be the universal covering . an invariant of links in @xmath0 is called _ essential _ if it is able to distinguish at least a couple of links having equivalent liftings under @xmath275 . the fundamental quandle for links in lens spaces is an inessential invariant since it is invariant under cyclic coverings ( see @xcite ) . on the contrary , for the virtual quandle invariant we have the following result . in order to prove that the virtual quandle is an essential invariant we need to find two links @xmath211 and @xmath276 in @xmath3 which have equivalent liftings and different virtual quandles . let us consider two links in @xmath277 depicted in figure [ ccc2 ] . this links have equal liftings ( see @xcite ) . by example [ xxx ] and example [ xx ] we have @xmath278 let @xmath279 and @xmath280 be virtual quandles obtained from @xmath214 and @xmath281 , respectively , adding the relations @xmath282 for all @xmath168 . then the virtual quandles @xmath283 are not isomorphic : they both have two elements , but in @xmath279 the automorphism @xmath76 permutes the elements , while in @xmath280 the automorphism @xmath76 fix both of them . then also @xmath214 and @xmath281 are obviously not isomorphic . 99 k. baker , j. grigsby , m. hedden , grid diagrams for lens spaces and combinatorial knot floer homology , int . res . not . imrn 10 , 2008 , art . i d rnm024 , 39 pp . d. buck , m. mauricio , connect sum of lens spaces surgeries : application to hin recombination , math . cambridge philos . soc . , v. 150 , 2011 , 505 - 525 . j. carter , a survey of quandle ideas , introductory lectures on knot theory , 22 - 53 , ser . knots everything , 46 , world sci . hackensack , nj , 2012 . a. cattabriga , e. manfredi , m. mulazzani , on knots and links in lens spaces , topology appl . , v. 160 , 2013 , 430 - 442 . a. cattabriga , e. manfredi , l. rigolli , equivalence of two diagram representations of links in lens spaces and essential invariants , acta math . hungar . , v. 146 , n. 1 , 2015 , 168 - 201 . c. cornwell , a polynomial invariant for links in lens spaces , j. knot theory ramifications , v. 21 , 2012 , 1250060 , 31 pp . drobotukhina , an analogue of the jones polynomial for links in @xmath284 and a generalization of the kauffman - murasugi theorem , ( russian ) algebra i analiz , v. 2 , n. 3 , 1990 , 171191 ; translation in leningrad math . j. , v. 2 , n. 3 , 1991 , 613-630 r. fenn , c. rourke , racks and links in codimension two , j. knot theory ramifications , v. 1 , n. 4 , 1992 , 343 - 406 . j. hoste , j. przytycki , the @xmath285-skein module of lens spaces ; a generalization of the jones polynomial , j. knot theory ramifications , v. 2 , 1993 , 321 - 333 . d. ilyutko , v. manturov , virtual knots : the state of the art , world scientific publishing co , singapore , 2013 . d. joyce , a classifying invariant of knots , the knot quandle , j. pure appl . algebra , v. 23 , 1982 , 37 - 65 . l. kauffman , virtual knot theory , eur . j. comb . , v. 20 , n. 7 , 1999 , 663 - 690 . s. lambropoulou , c. rourke , markov s theorem in 3-manifolds , topology appl . , v. 78 , 1997 , 95 - 122 . e. manfredi , lift in the 3-sphere of knots and links in lens spaces , j. knot theory ramifications , v. 23 , n. 5 , 2014 , 1450022 . v. manturov , invariants of virtual links , ( russian ) dokl . nauk , v. 384 , n. 1 , 2002 , 11 - 13 , dokl . math , v. 65 , n. 3 , 329 - 331 . v. manturov , on invariants of virtual links , acta appl . , v. 72 , n. 3 , 2002 , 295 - 309 .	we construct a virtual quandle for links in lens spaces @xmath0 , with @xmath1 . this invariant has two valuable advantages over an ordinary fundamental quandle for links in lens spaces : the virtual quandle is an essential invariant and the presentation of the virtual quandle can be easily written from the band diagram of a link . + _ mathematics subject classification 2010 : _ primary 57m27 , 08a99 ; secondary 57m10 . + _ keywords : _ links in lens spaces , link invariants , virtual quandles .
the general motivation is the question where are the fundamental particles corresponding to the other representations of the permutation group ? " so far , fermions go in totally anti - symmetric representations and bosons go in totally symmetric representations . but there are mixed representations . particles associated with such representations obey parastatistics @xcite . the order of the paraparticles is denoted by @xmath0 " such that @xmath1 " corresponds to the ordinary bosons and fermions . in general , @xmath0 " is the maximum number of parafermions ( parabosons ) that can occur in a totaly symmetric ( anti - symmetric ) state . from r - parity , all supersymmetric particles are assigned a new conserved quantum number @xmath2 so @xmath3 respectively for particles and superparticles . consequently , superparticles are pair - produced and the lightest superpartner ( lsp ) must be stable . from locality , the selection rules @xcite for the production of paraparticles are similar : transitions with only one external @xmath4 particle are forbidden , so the lightest such particle is stable . the number of parafermions is conserved modulo 2 . for a @xmath0-even family , the number of external lines is even , so an even number of @xmath4 must be produced . for a @xmath0-odd family , the number of external lines can not be an odd number less than @xmath0 . for instance , if @xmath5 , then there can not be 3 paraparticles produced . however , in the case of @xmath6 there is the important exception that while a single paraparticle can not be produced , the production of 3 such particles is allowed . in the case of susy , once the masses are known , the cross - sections are fixed . for instance , sleptons and electroweak gauginos interact with strength @xmath7 similarly , squarks and gluinos interact with strength @xmath8 . in leading order in the coupling , the corresponding cross - section in the pair - production of two analogous order @xmath0 paraparticles is a factor of @xmath0 larger . in comparison with susy , such a cross - section test can be used to exclude the production of paraparticles . however , if a cross - section were found to be a factor of @xmath0 greater than expected , this would be the same prediction as in the case of the production of @xmath0-fold degenerate fermions / bosons . in particular , there is the issue `` how to distinguish @xmath9 parabosons from two - fold degenerate ordinary bosons ? '' in the case when susy is not involved in the pair - production processes , the magnitude of the cross - sections might not be known apriori , but this same question arises . the multiplicity signatures for @xmath9 parabosons and ordinary bosons are indeed different . to show this , we use a parameter - free statistical model @xcite to study multiplicity signatures for coherent production of charged - pairs of parabosons of order @xmath9 in comparison with those arising in the case of ordinary bosons , @xmath1 . this model gives 3d plots of the pair probability @xmath10@xmath11 the probability of @xmath12 paraboson charged - pairs + @xmath13 positive parabosons " versus @xmath14 and @xmath15 . as shown in the figure from ref . [ 3 ] , the @xmath1 curve is found to lie on the relatively narrow 2d @xmath9 surface . such signatures distinguish between @xmath9 parabosons and two - fold degenerate ordinary bosons , @xmath1 . circa 1970 , for ordinary bosons horn and silver @xcite constructed the analogous parameter - free statistical model , using conserved - charge coherent states , and showed that it well described inelastic @xmath16 pair production from fixed targets with laboratory kinetic - energies up to 27 gev . the model agreed with the universal trends of an experimental regularity reported by wang . horn and silver argued for a statistical treatment of the gross features because ( i ) momentum conservation should be a weak constraint since the emitted pions occupy a small part of the available phase space , ( ii ) total isospin conservation on the distribution of charged pions should also be weak since neutral pions are summed over , and so ( iii ) charge conservation remains as the important constraint . a charged - paraboson pair in order @xmath9 consists of one a quantum of charge ` + 1 ' and one b quantum of charge ` -1 ' . the hermitian charge operator is defined by @xmath17 where @xmath18 are the parabose number operators . @xmath19 does not commute with @xmath20 or @xmath21 , and the paraboson pair operators @xmath22 . nevertheless , since @xmath23=0,\quad \lbrack q , ba]=0,\quad \lbrack ab , ba]=0 $ ] , the @xmath9 conserved - charge coherent state can be defined as a simultaneous eigenstate of @xmath24 and @xmath25 : @xmath26 unlike in the @xmath1 case @xcite where only one parameter arises , here two complex numbers @xmath27 and @xmath28 arise because @xmath29 and @xmath25 are fundamentally distinct operators . consequently , in the following multiplicity considerations , two non - negative parameters occur which are the moduli of these two complex numbers , @xmath30 and @xmath31 . these parameters can be interpreted as the intensity strengths of the @xmath29 " and @xmath25 " sources . the explicit expressions for the conserved - charge coherent states @xmath32 are given in ref . [ 5 ] . for @xmath13 fixed , the percentage of events with @xmath12 such pairs , @xmath33 , is the square of the moduli of the expansion coefficients of @xmath32 in terms of the two - mode parabose number fock states . the figure shows the pair - probability @xmath34 for the production of two @xmath35 and one @xmath36 . the formulas for the @xmath1 curve are for the source - intensity - strength @xmath37 non - negative @xmath38 and for the @xmath9 ribbon are for the intensity strengths @xmath39 , @xmath40 non - negative @xmath41 where the @xmath42 s are modified bessel functions . in the figure , the solid line is the @xmath1 curve . this @xmath1 curve is also the @xmath43 line on the @xmath44 ribbon . near the peak , there is a fold " at the bottom edge of the ribbon . as shown , the @xmath9 ribbon consists of open - circles for the non - folded @xmath45 region , and of solid - circles for the folded @xmath46 region . the upper edge of the ribbon is the @xmath47 set of points . slightly to the right of the peak , one can see from the solid - circles that each line of dots travelling leftward down the page , bends under ( or over " , whichever as the viewer prefers ) the fold to reach the @xmath43 line on the ribbon . for the cases of @xmath48 and @xmath49 , the @xmath1 curve always lies on the @xmath9 two dimensional surface . for @xmath13 odd , as shown in the figure , the ribbon is @xmath50 asymmetric and the ribbon is only partly folded over . for @xmath13 even , there is a complete fold @xmath51 of the ribbon because then the @xmath52 and @xmath53 surfaces are identical . the line of dots travelling leftward down the page " in the peak region of the figure are a set of @xmath54 values from a unit - negative - slope diagonal in the @xmath54 domain . in independent - particle - emission models , the total energy in the emitted particles is monotonically related to the intensity strength of the source . this suggests that with the sum @xmath55 fixed , there will be a significant @xmath56 energy dependence in coherent paraboson pair - production . @xmath57 are respectively the total emitted @xmath58 and @xmath59 quanta s energies . there is overall @xmath60 symmetry ( @xmath61 charge symmetry ): the probability for @xmath62 @xmath35 s and @xmath63 @xmath36 s " equals probability for @xmath64 @xmath35 s and @xmath65 @xmath36 s " , because @xmath66 and @xmath67 for @xmath68integer . nevertheless , for q odd , physical observables such as @xmath33 are not symmetric under the @xmath69 exchange of the @xmath29 " and @xmath25 " intensity strengths . ( 5 - 7 ) . this work was partially supported by u.s . dept . of energy contract de - fg 02 - 86er40291 . green , phys.rev . * 90 * , 270 ( 1953 ) ; d.v . volkov , sov . jetp , * 11 * , 375 ( 1960 ) ; o.w . greenberg and a.m.l . messiah , phys.rev . * 136 * , b248 ( 1964 ) ; _ ibid . _ , * 138 * , b1155 ( 1965 ) ; y. ohnuki and s. kamefuchi , _ quantum field theory and parastatistics _ ( 1982 ) ( univ . press of tokyo ) ; and o.w . greenberg and a.k . mishra , `` path integrals for parastatistics '' , math - ph/0406011 . h. feshbach , phys.rev.lett . * 3 * , 317 ( 1963 ) . n. frascino and c.a . nelson , hep - ph/0308089 . color versions of some of the figures and talk - transparencies are available at vietnam 2004 " and small x and diffraction 2003 " , fnal , websites . d. horn and r. silver , phys.rev . * d2 * , 2082 ( 1970 ) ; c.p . wang , phys.rev . * 180 * , 1463 ( 1969 ) ; h.a . kastrup , nuc . phys . * b1 * , 309 ( 1967 ) . s. jing and c.a . nelson , j. phys . a : * 32 * , 4131 ( 1999 ) . figure 1 : the pair - probability @xmath34 for the production of two @xmath35 parabosons and one @xmath36 paraboson versus the mean number of charged - pairs @xmath14 and @xmath15 .	due to selection rules , new particles are sometimes discovered / predicted to be produced in pairs . in the current search for susy particles this will occur if r - parity is conserved . in local relativistic field theory , there can be identical particles which are neither bosons nor fermions which are associated with higher - dimensional representations of the permutation group . such particles will generally be pair - produced and so empirical tests are required to exclude them . a parameter - free statistical model is used to study the unusual multiplicity signatures in coherent paraboson production versus the case of ordinary bosons . = -5 mm * tests for the statistics of pair - produced new particles * + charles a. nelson + _ department of physics , state university of new york at binghamton + binghamton , n.y . 13902 _ +
in a recent study of the quantum energies of interfaces in field theory , we discovered a set of consistency conditions on scattering data that take the form of finite energy sum rules in potential scattering , @xcite @xmath0 + \sum_j ( -\kappa_{\ell j}^{2})^{n } = 0\ , , \qquad\quad m \ge n \ , . \label{main}\ ] ] here @xmath1 denotes the scattering phase shift in the channel with angular momentum @xmath2 , and @xmath3 is the @xmath4 born approximation . the sum on @xmath5 ranges over the bound states with angular momentum @xmath2 and @xmath6 is the binding energy . note that for @xmath7 , eq . ( [ main ] ) is simply levinson s theorem . in fact , these identities are the natural generalizations of levinson s theorem . @xmath8 is the minimal number of born subtractions necessary to render the integral in eq . ( [ main ] ) finite . since we may generally subtract further born approximations , eq . ( [ main ] ) also implies oversubtraction " rules such as @xmath9 in ref . @xcite these sum rules appeared as consistency conditions in quantum field theory leading to finite expressions for the casimir energies of interfaces . here we derive them within scattering theory , and consider applications and consequences in ordinary quantum mechanics . our proof will employ jost function techniques commonly used to prove levinson s theorem @xcite . for sufficiently singular potentials , however , our sum rules fail , even though levinson s theorem continues to hold . in section ii we derive the sum rules for the antisymmetric channel in one dimension , where the analysis is simplest . this derivation applies to the @xmath10-wave in three dimensions as well . the extension to higher partial waves in three dimensions is straightforward and is presented in the appendix . in section iii we discuss generalizations . the generalization to fermion scattering ( via the dirac equation ) is also straightforward and is left to the reader . we also mention the generalization to multichannel problems with internal symmetries . the symmetric channel in one dimension requires special consideration ( as it does for levinson s theorem @xcite ) and is treated in detail in section iv . in section v we describe some singular potentials for which the sum rules do not hold . finally , in section vi we study the semiclassical limit , where the sum rules take a particularly compact form and have a simple physical interpretation . these sum rules could have been derived many years ago in the heyday of potential scattering theory . however , we have been unable to find them in the literature . they bear some resemblance to results based on the gelfand - diki equation , obtained in the russian literature @xcite , although the physical foundations and the resulting sum rules themselves are quite different . in our conclusion we mention this earlier work and compare it with our own . also , for the special case of a separable potential a related sum rule was obtained in ref . we consider the scattering of a spinless particle in a symmetric potential @xmath11 in one dimension , described by the schrdinger equation , @xmath12 this is a two channel problem . the antisymmetric channel is specified by the boundary condition @xmath13 . the symmetric channel corresponds to @xmath14 . here we consider the antisymmetric channel . let @xmath15 denote the scattering phase shift in this channel , defined by the asymptotic form of the wave function @xmath16 at large @xmath17 , @xmath18 our goal is to derive the sum rules @xmath19 = - \sum_j ( -\kappa_{-,j}^{2})^{n } \ , , \qquad\quad m \ge n \label{main1}\ ] ] where the sum on @xmath5 ranges over the antisymmetric bound states of @xmath20 with binding energies @xmath21 . for the remainder of this section we suppress the subscript labeling the antisymmetric channel . for real @xmath22 , the phase shift @xmath23 is given in terms of the s - matrix @xmath24 , which in turn is related to the jost function @xmath25 by @xmath26 \,.\ ] ] the born approximation is an expansion of the phase shift @xmath27 ( not the jost function @xmath28 itself ) in powers of the interaction @xmath20 . the jost function is obtained from the _ jost solution _ @xmath29 to eq . ( [ schr1 ] ) , which is asymptotic to an outgoing wave at infinity , @xmath30 and @xmath31 . as is well known , the integral equation for @xmath29 has a unique solution in the upper half @xmath22-plane , where it is holomorphic and continuous as @xmath32 , provided that the potential @xmath20 is locally integrable and from the so - called faddeev class " @xmath33 in addition , @xmath25 has zeros at the bound states , @xmath34 , on the positive imaginary axis.@xcite to quantitatively estimate the behavior of the born approximation at large momenta @xmath35 , we furthermore have to assume that the interaction @xmath20 is bounded and sufficiently smooth to allow for integration by parts . unless stated otherwise , we will restrict our analysis to non - singular potentials @xmath20 with these properties . to proceed , we take @xmath36 and introduce an auxiliary function , @xmath37 , with the following properties : 1 . @xmath37 is analytic and has no zeros in the upper half @xmath22-plane including @xmath38 . @xmath39 falls like @xmath40 as @xmath41 in the upper half plane . after completing the derivation of our sum rules we will construct @xmath37 and relate it to the born approximation to @xmath23 . for real @xmath22 we introduce @xmath42 \ , , \label{deltam}\ ] ] and consider @xmath43 since the integrand is manifestly even in @xmath22 , we can extend the integration range to @xmath44 . applying the substitution @xmath45 we obtain , @xmath46 where the contour @xmath47 is the real axis plus the semicircle of infinite radius in the upper half plane . the semicircle gives no contribution to the integral because of property ( b ) . the contour integral can now be performed using cauchy s theorem by recognizing that @xmath48 has poles of unit residue at each bound state . by property ( a ) , @xmath49 is has no poles inside @xmath47 . the result is the sum rule , eq . ( [ main1 ] ) . in this section we construct an auxiliary function with the two properties required in the previous subsection . it is convenient to parameterize the jost solution @xmath29 in terms of an exponent @xmath50 , @xmath51 substituting into the schrdinger equation we find that the complex function @xmath50 satisfies @xmath52 subject to the boundary condition @xmath53 where @xmath54 . combining eqs . ( [ smatrix ] ) and ( [ beta ] ) with the boundary condition @xmath55 , it is easy to see that @xmath56 eqs . ( [ betaeq ] ) and ( [ betabc ] ) can be converted into a non - linear integro - differential equation , @xmath57 where @xmath58 note that by differentiation @xmath59 obeys a similar equation , @xmath60 denote the term in @xmath50 that is @xmath61 order in the potential by @xmath62 and define @xmath63 . an equation for @xmath62 can be obtained from eq . ( [ betaint ] ) by iteration @xmath64 here @xmath65 is the term in the expansion of @xmath66 which is of @xmath61 order in the potential . for @xmath67 , @xmath65 involves only @xmath68 with @xmath69 . thus we are led to equations for @xmath70 , the first few of which are @xmath71^{2},\cr \beta^{(3)}(k ) & = & \frac{1}{2k}\int_{0}^{\infty}dy\,\left ( 1-e^{2iky}\right)2\beta'^{(1)}(k , y)\beta'^{(2)}(k , y)\ , , \label{bornseries}\end{aligned}\ ] ] and so forth . similarly for the @xmath72 , which appear as sources in eqs . ( [ bornseries ] ) , @xmath73^{2},\cr \beta'^{(3)}(k , x ) & = & i\int_{x}^{\infty}dy\ , e^{2ik(y - x)}2\beta'^{(1)}(k , y)\beta'^{(2)}(k , y)\ , . \label{bornprimeseries}\end{aligned}\ ] ] the exponential factors in eq . ( [ bornbeta ] ) guarantee that @xmath70 is analytic in the upper half @xmath22-plane provided that @xmath65 is , and likewise for @xmath72 . starting with @xmath74 we derive the required analytic properties of @xmath75 and @xmath76 inductively . to obtain the large @xmath35 behavior of @xmath75 and @xmath76 from their respective integral representations , we integrate by parts once and estimate the remainder by sequentially applying the riemann - lebesgue lemma .. in particular , the riemann - lebesgue lemma requires the existence of @xmath77 . ] the result is that * for the class , eq . ( [ faddeev ] ) of potentials , @xmath70 is holomorphic in the upper half plane including at @xmath38 ; * @xmath78 as @xmath41 in the upper half plane @xmath79 . to complete the derivation , we define @xmath80 \label{aux}\ ] ] the required properties of @xmath37 follow directly from those of @xmath70 the convergence of the born series @xmath81 for sufficiently large @xmath35 in the upper half plane @xcite ( see also the appendix ) . the quantity that enters the sum rule is @xmath82 , given by eq . ( [ deltam ] ) . from eq . ( [ betaeq ] ) , it follows that @xmath83 for real @xmath22 . as a result , we have @xmath84 so that @xmath82 is the sum of the first @xmath85 terms in the born expansion of @xmath23 . this completes the derivation of the sum rules in the antisymmetric channel in one dimension . the antisymmetric channel in one dimension is actually generic . the sum rules can easily be extended to scattering from a central potential in any number of space dimensions @xmath86 . the computation proceeds for each partial wave in analogy to the antisymmetric case in @xmath87 . of course , the appropriate generalized hankel functions must replace the simple exponentials that appear in one dimension . we summarize the derivation for @xmath88 in the appendix . the sum rules also extend to the case of fermion scattering in a straightforward way . for a scalar potential , the dirac equation decomposes into partial waves labeled by total spin @xmath5 and parity @xmath89 , and the sum rules again hold in each partial wave individually . when there are internal symmetries , so that there are several channels @xmath90 in each partial wave , we expect that the sum rules will continue to hold with the phase shifts replaced by the sum of the eigenphases , which is given by the trace of the logarithm of s : @xmath91 similarly , the sum over bound states will include all bound states in the channel . for example , consider an isodoublet of fermions scattering in a background generated by an isodoublet scalar higgs field in three dimensions . if the higgs background is of the `` hedgehog '' form @xmath92 , then the fermion spectrum will decompose into channels labeled by parity @xmath89 and grand spin @xmath93 . in each channel , @xmath94 is a 2-by-2 matrix , which can not be simultaneously diagonalized for all @xmath22 . ( each degree of freedom also appears with the usual @xmath95 degeneracy . ) if we introduce a chiral su(2 ) gauge field that maintains grand spin conservation , then states with different parity but the same @xmath93 will mix , leaving a 4-by-4 @xmath94-matrix labeled only by @xmath93 . the symmetric channel in one dimension introduces additional subtleties , which are treated in the following section . the result is that the sum rules may be modified by an anomalous piece if too many subtractions are attempted . specifically , the sum rules in the symmetric channel read @xmath96 = - \sum_j ( -\kappa_{+,j}^{2})^{n } + i_{n , m}^{\rm anom}\ , , \label{mainsym}\ ] ] for @xmath97 . the anomalous term vanishes if @xmath98 . as a result , the `` minimally subtracted '' form of the sum rules , where @xmath8 , hold without modification except for the case @xmath99 , which is levinson s theorem . in that case @xmath100 and we recover the extra term that appears in levinson s theorem in the symmetric channel @xcite . we have checked these results numerically in a variety of simple , generic potentials . in one dimension , we have also checked them for the exactly solvable reflectionless scalar potentials of the form @xmath101 , with @xmath2 integer @xcite , and the corresponding potentials in the fermionic case @xcite . as in the case of levinson s theorem , the symmetric channel requires special attention . the regular solution to the schrdinger equation obeying the boundary conditions , @xmath102 and @xmath103 , can be written in terms of the jost solution @xmath29 , @xmath104,\ ] ] where @xmath105 . comparing to the @xmath94-matrix parameterization as @xmath106 , @xmath107 we see that @xmath108 \,.\ ] ] the derivation proceeds exactly in analogy to the antisymmetric channel except , as we shall see , it is not possible to find an auxiliary function that is regular at @xmath38 . instead we introduce an auxiliary function @xmath109 with the following properties : 1 . @xmath109 is analytic and has no zeros in the upper half @xmath22-plane excluding @xmath38 . @xmath110 falls like @xmath40 as @xmath41 in the upper half plane . 3 . at @xmath38 , @xmath111 has a pole with residue @xmath112 . 4 . @xmath113 for @xmath98 . then , using the properties of @xmath93 and @xmath114 we find , @xmath115 \cr & = & -\frac{1}{2\pi i}\int_{-\infty}^{\infty}dk \ k^{2n}\frac{d}{dk } \left(\ln g(k)-\ln g_{m}(k)\right ) \cr & = & -\frac{1}{2\pi i}\oint_{\cal c}dk \ k^{2n}\frac{d}{dk } \left(\ln g(k)-\ln g_{m}(k)\right)\cr & = & \left\{\begin{array}{ll } - \sum_j ( -\kappa_{j}^{2})^{n } & \qquad2n > m\ge n\\ - \sum_j ( -\kappa_{j}^{2})^{n } + i_{n , m}^{\rm anom } & \qquad m\ge 2n \end{array}\right . \label{contoursymm}\end{aligned}\ ] ] where @xmath116 \,.\ ] ] the factor of @xmath117 difference between the residue of the pole at @xmath38 and the anomalous term in the sum rule arises because the integration contour @xmath47 passes through @xmath38 and therefore captures only half the residue . given the restrictions on @xmath85 and @xmath118 in eq . ( [ contoursymm ] ) , it is clearly possible to derive a non - anomalous sum rule in the symmetric channel by making the minimal subtraction , @xmath8 . the only exception is levinson s theorem , @xmath7 , which we discuss in detail below . otherwise anomalies arise if one attempts to `` oversubtract '' for a given @xmath118 . we consider specific examples after constructing the auxiliary function @xmath109 . as in the antisymmetric channel , the auxiliary function is obtained from the expansion of the jost function in powers of the potential . the difference is that the relevant jost function is @xmath119 defined by @xmath105 . using the exponential parameterization of eq . ( [ beta ] ) , we find @xmath120 where , @xmath121 . the prefactor @xmath122 gives the difference from the antisymmetric channel . comparing with eq . ( [ aux ] ) we are led to the ansatz @xmath123_{m}+\ln f_{m}(k ) \label{gansatz}\ ] ] where the notation @xmath124_{m}$ ] is an instruction to make the formal expansion of @xmath125 in powers of the potential and keep all terms up to @xmath126 order . for example , @xmath127_{0}&=&\ln k\cr \left[\ln(k+\beta'(k))\right]_{1}&= & \ln k+\beta'^{(1)}(k)/k\cr \left[\ln(k+\beta'(k))\right]_{2}&= & \ln k+\beta'^{(1)}(k)/k + [ \beta'^{(1)}(k)]^{2}/k^{2}+\beta'^{(2)}(k)/k \,.\end{aligned}\ ] ] this process is necessary to reproduce the asymptotic behavior of @xmath119 at large @xmath35 as required by condition ( b ) above . the cost is the introduction of poles up to @xmath126 order at @xmath38 . it is straightforward to verify that @xmath109 defined in eq . ( [ gansatz ] ) satisfies requirements ( a ) and ( b ) above . the argument is essentially the same as for the antisymmetric channel . it is clear from the definition of @xmath109 that its contribution to the integral along the real axis , proportional to @xmath128 is just the sum of the first @xmath85 terms in the born approximation to the phase shift in the symmetric channel . it remains to characterize the singularity in @xmath109 at @xmath38 . the term in the contour integral in eq . ( [ contoursymm ] ) which is potentially singular at @xmath38 is proportional to @xmath129_{m}&= & k^{2n}\left[\frac{1+\frac{d\beta'(k)}{dk}}{k+\beta'(k)}\right]_{m } \cr & = & k^{2n-1}\left[\left(1+\frac{d\beta'(k)}{dk}\right)\sum_{p=0}^{\infty}\left ( \frac{-\beta'(k)}{k}\right)^{p}\right]_{m } \label{expand}\end{aligned}\ ] ] an anomalous contribution to the sum rule will result if the @xmath130 singularities in the expansion of @xmath131 to @xmath126 order in the potential overcome the prefactor of @xmath132 . since the functions @xmath133 are all analytic in the vicinity of @xmath38 , the most singular term in eq . ( [ expand ] ) comes from the term @xmath134 , which is singular if @xmath135 . if @xmath136 there is a simple pole at @xmath38 from this term . if @xmath137 there are poles of higher order as well . it is straightforward ( but increasingly tedious ) to pull out the residue of the simple pole , which determines the anomalous contribution to the sum rule . once having identified the residue , the expression for the anomalous contribution to @xmath138 is @xmath139_{m } \label{residue}\ ] ] we illustrate this result with some important special cases : * @xmath7 : levinson s theorem + for @xmath7 we need the coefficient of @xmath130 in the term zeroth order in the potential in @xmath140 , which is unity . thus @xmath141 , and we obtain levinson s theorem in the symmetric channel:@xcite @xmath142 * @xmath143 : minimal subtraction + for @xmath144 , the minimum born subtraction we can make in order to render @xmath138 convergent is to take @xmath8 . the most singular term in the expansion of eq . ( [ expand ] ) through @xmath126 order is proportional to @xmath145 . thus the integrand goes like @xmath146 near @xmath38 . this has no pole when @xmath143 . so @xmath147 for @xmath148 , and the minimally subtracted form of the sum rules is not altered in the symmetric channel . * @xmath98 : oversubtraction without an anomaly : + there is no singularity at @xmath38 as long as @xmath98 . therefore it is possible to subtract further born approximations from the phase shift without introducing anomalies into the sum rules . for @xmath149 the first born approximation must be subtracted for convergence and no further subtraction is possible without anomaly . for @xmath150 the first and second born approximations must be subtracted and the third may be subtracted without anomaly , and so forth . * @xmath136 : computation of the anomaly . + for fixed @xmath118 , as further subtractions are attempted , one finally reaches @xmath136 , where an anomaly appears . the anomaly comes entirely from the term proportional to @xmath151 in the expansion of the integrand . referring back to the definition of @xmath133 , we find @xmath152 so @xmath153^{2n}\ ] ] + in particular , the first non - trivial case is the @xmath149 sum rule from where both the first and second terms in the born approximation have been subtracted : @xmath154 = \sum_j \kappa_j^{2 } -\frac{1}{2}\left[\int_{0}^{\infty}dyv(y)\right]^{2}\ ] ] this result was first discovered in conjunction with the work of @xcite by direct evaluation of the feynman graph corresponding to the second born approximation . here we see that it follows from a careful analysis of the analytic properties of the born approximation near @xmath38 and has essentially the same origin as the extra factor of @xmath155 that appears in levinson s theorem for the symmetric channel . among the cases that we have checked numerically is the square well in one dimension . even though its sharp edges seem to invalidate our use of complex analysis in the proof above , the sum rules still hold . as is the case with levinson s theorem , the difference between a sharp edge and a smooth , very steep edge can be made arbitrarily small . if we take the limit where the width of the well goes to zero with the area held fixed , we obtain the delta - function potential @xmath156 where we have written the dirac delta function as @xmath157 to distinguish it from a phase shift . the phase shift in this potential vanishes in the antisymmetric channel because @xmath20 is localized at the origin , where the antisymmetric wavefunction vanishes . the symmetric channel phase shift and its first born approximation are easily calculated : @xmath158 the symmetric channel has a bound state at @xmath159 . like the square well , this potential obeys the one - dimensional version of levinson s theorem @xcite relating the phase shifts to the number of bound states in each channel @xmath160 however , sum rule with @xmath161 , eq . ( [ mainsym ] ) , fails . one expects @xmath162 , but obtains instead @xmath163 . examining a sequence of square well potentials approaching the delta function reveals that for any square well , the sum rule is satisfied , but the support of the integral moves out to larger and larger @xmath22 as the potential gets narrower and deeper . the @xmath164-function limit and the @xmath22-integration do not commute . it is instructive to examine more closely what goes wrong in the @xmath164-function case . a straightforward calculation shows that the proper jost function for the symmetric channel in one dimension is @xmath165 note that it has a zero at @xmath166 as expected and is analytic in the upper half @xmath22-plane . according to the symmetric channel analysis , @xmath167 the derivation of levinson s theorem using @xmath168 proceeds without difficulty . to derive the sum rule for @xmath169 it is necessary to consider @xmath170 . this quantity vanishes like @xmath171 for large @xmath35 , _ not _ @xmath172 as expected on the basis of property ( b ) listed in the previous section . as a result , the integral around the semicircle at infinity does not vanish . specifically , @xmath173 = \frac{\lambda^{2}}{8}\ ] ] where @xmath174 is the semicircle at infinity in the upper half @xmath22-plane . we combine this result with the integral along the real axis , @xmath175-\lambda^{2}/8\ ] ] now @xmath169 can be evaluated by contour integration , yielding the same anomalous result obtained by direct integration of @xmath176 along the real axis : @xmath177 . it remains to explain why @xmath178 falls only like @xmath171 . consider @xmath179 since @xmath180 is proportional to @xmath181 , @xmath182 and @xmath183 determine the large @xmath22 behavior of @xmath184 . from their definitions , ( [ betaint ] ) and ( [ betaprime ] ) , @xmath185 for @xmath186 . in particular , as @xmath187 , @xmath188 . in contrast , if @xmath20 is any bounded function of @xmath17 , including a square well , @xmath189 . this is the ultimate source of the breakdown of the sum rule in the case of the @xmath164 function , for which @xmath190 is ill - defined . our sum rules become especially simple in the wkb approximation . they provide formulas for the sum of powers of the binding energies as integrals over powers of the potential . in this way , the wkb approximation yields some insight into the physical origin of the sum rules . we have checked the accuracy of these results in some simple potentials . we work in one dimension with a potential @xmath20 that is everywhere negative . we therefore define @xmath191 . we assume that @xmath192^{n}$ ] exists for all @xmath193 . the reflection coefficient is exponentially small in the wkb approximation , so the even and odd parity phase shifts are equal and are given by @xmath194 the @xmath61 born approximation to @xmath23 is merely the term of order @xmath195 in the expansion of the integrand . the wkb approximation should be valid when @xmath196 , where @xmath197 is the local de broglie wavelength . for a deep , smooth potential this criterion is satisfied for all @xmath17 . the first correction to the wkb approximation gives only a modulation of the magnitude of the wavefunction and does not change its phase . so we expect eq . ( [ wkbvalid ] ) to be quite a good approximation . to evaluate the sum rule with minimal subtraction ( @xmath8 ) we must calculate @xmath198 a straightforward calculation yields the wkb estimate @xmath199^{n+\frac{1}{2 } } \equiv i_{n , n}^{\rm wkb}\ , . \label{wkbresult}\ ] ] note that in the wkb approximation , the @xmath61 born approximation is proportional to @xmath200 so oversubtraction of the sum rules is not allowed in this case . in fig . [ wkbplot ] we show the relative error that arises due to the wkb approximation for various sum rules . for sufficiently strong potentials this error is indeed small . alternatively , we can use zeta function regularization to define a regularized integral @xmath201 and evaluate for @xmath202 , where it converges to a beta function . we then analytically continue to @xmath203 , and obtain the same result . curiously , this result implies that in the wkb approximation , the contribution of the born terms vanish in zeta function regularization . note that @xmath204 gives the wkb approximation to levinson s theorem : @xmath205 and @xmath206 gives a particularly simple formula for the sum of the binding energy of all bound states in the wkb approximation : @xmath207^{\frac{3}{2}},\ ] ] not surprisingly , eq . ( [ wkbresult ] ) has a simple semiclassical interpretation . we replace the sum over bound states by an integral over the density of states , @xmath208 , @xmath209 where the integral extends only over bound states ( @xmath210 ) . if we approximate @xmath211 semiclassically by @xmath212 , where @xmath213 , then @xmath214 direct evaluation of this simple integral yields eq . ( [ wkbresult ] ) . our sum rules are related to the results of buslaev and faddeev based on the gelfand - diki equation @xcite . they studied solutions to the schrdinger equation on the half - line subject to the boundary condition @xmath215 . in our language , this system is equivalent to the antisymmetric channel in one dimension , though without the restriction that the potential be smooth at the origin ( so they can have @xmath216 ) . they obtained a sequence of sum rules , beginning with @xmath217 their identities have a similar structure to our sum rules ( after integrating by parts ) . there are some significant differences , however . instead of subtracting the born approximation as we have done , they instead have subtracted the leading local asymptotic expansion of the phase shift in powers of @xmath130 . their expressions are simpler just integrals of the potential over space divided by powers of @xmath22 but more singular at the origin . as a result , it is not possible to form the oversubtracted versions of their identities . this difference also accounts for the need for extra terms proportional to the potential and its derivatives at the origin . in field theory applications , the born subtractions arise naturally in the process of renormalization in a definite scheme.@xcite we know of no similar application of the buslaev - faddeev results . we thank e. abers , g. dunne , and j. goldstone for helpful discussions , suggestions , and references . m. q. and h. w. are supported in part by deutsche forschungsgemeinschaft under contracts qu 137/1 - 1 and we 1254/3 - 1 . r. l. j. is supported in part by the u.s . department of energy ( d.o.e . ) under cooperative research agreement # df - fc02 - 94er40818 . n. g. is supported by the u.s . department of energy ( d.o.e . ) under cooperative research agreement # de - fg03 - 91er40662 . in this section , we extend the proof of the sum rules to spherically symmetric potentials in three dimensions . our analysis follows the case of the antisymmetric channel in one dimension closely . we consider the scattering of a spinless particle in a central potential @xmath218 in three dimensions described by the radial schrdinger equation , @xmath219\,\psi_\ell = k^2 \psi_\ell \ , . \label{schr}\ ] ] let @xmath1 denote the scattering phase shift in the channel with angular momentum @xmath2 , and @xmath3 the @xmath61 born approximation . our goal is to derive the sum rules @xmath0 = - \sum_j ( -\kappa_{\ell j}^{2})^{n } \ , , \qquad\quad m \ge n \label{mainell}\ ] ] where the sum on @xmath5 ranges over the bound states with angular momentum @xmath2 and @xmath6 is the binding energy . for real @xmath22 , the phase shift @xmath1 is given in terms of the s - matrix @xmath220 , which in turn is related to the jost function @xmath221 by@xmath222 \,.\ ] ] the born approximation is an expansion of the phase shift @xmath223 in powers of the interaction @xmath218 . the jost function is obtained from the jost solution @xmath224 , which is the solution to eq . ( [ schr ] ) asymptotic to an outgoing wave at infinity @xmath225 uniqueness of @xmath224 and analyticity in the upper @xmath22-plane can be verified for locally integrable potentials @xmath218 with the property @xcite @xmath226 it is convenient to introduce the free jost solution @xmath227 where @xmath228 is the spherical hankel function . @xmath229 is determined by the @xmath230 limit of @xmath231 , @xmath232 as in previous cases , @xmath229 is analytic in the upper half @xmath22-plane with zeros at the bound state momenta , @xmath34 . for real @xmath22 we define @xmath237 \label{deltalm}\ ] ] and the remainder of the derivation follows exactly as in the antisymmetric channel in one dimension : the sum rule eq . ( [ mainell ] ) follows from cauchy s theorem . as in one dimension , it is convenient to parameterize @xmath238 in terms of an exponential , @xmath239 substituting into the radial schrdinger equation , we find that the complex function @xmath240 satisfies @xmath241 subject to the boundary condition @xmath242 where @xmath243 . here we have introduced @xmath244 , @xmath245\ , , \label{defetaell}\ ] ] which is a simple rational function of @xmath246 . it is convenient to convert eq . ( [ betaeqell ] ) into a non - linear integro - differential equation , @xmath247 where @xmath248 \ , . \label{betaintell}\ ] ] by definition the value of @xmath249 at @xmath250 is the logarithm of the jost function , @xmath251 furthermore , the analytic properties of the jost solution imply @xmath252 , yielding the phase shift @xmath253 \label{deltaeqell}\ ] ] for real @xmath22 . the born series for @xmath254 is constructed by iterating eq . ( [ betaeqell ] ) and keeping track of powers of @xmath218 in the source @xmath255 on the r.h.s . of ( [ betaint1 ] ) . with these definitions , we are prepared to construct @xmath233 . we begin by proving two important properties of the @xmath256 using induction . first , we need to show that the @xmath257 approximation @xmath258 of eqs . ( [ betaeqell ] ) and ( [ betabcell ] ) is a holomorphic function of @xmath22 in the upper half plane @xmath259 . to see this , we go back to the initial value problem for @xmath260 and rewrite eq . ( [ betaeqell ] ) as @xmath261 with the boundary condition @xmath262 for all @xmath22 . @xmath263 is the @xmath257 term in the iteration of @xmath264 , including all combinations of @xmath265 and @xmath266 that give a total order of @xmath267 : @xmath268 and @xmath269 we proceed by induction and therefore assume that the right hand side of eq . ( [ betaelldiffeq ] ) is holomorphic in the upper half @xmath22-plane . this is certainly the case for @xmath270 . since the boundary condition is independent of @xmath22 , poincar s theorem ensures that the solution @xmath271 of eq . ( [ betaelldiffeq ] ) is in fact a holomorphic function of @xmath22 in every domain where the coefficients are . it thus remains to show that @xmath272 as defined in eq . ( [ defetaell ] ) is holomorphic in @xmath246 for @xmath273 . this will be the case if the free jost solution @xmath274 is non - vanishing in the upper half plane . for @xmath275 it is trivial , since @xmath276 . for @xmath277 , it suffices to note that any zero of @xmath278 in the upper half plane would correspond to a bound state of the free schrdinger equation , which is forbidden by the repulsive centrifugal barrier . second , we have to establish the convergence of the born series . the iteration of the integral equation for @xmath224 yields an expansion in the interaction @xmath279 that is uniformly and absolutely convergent in the upper half plane @xmath280 . from the bound @xcite @xmath281 and the limit @xmath282 in the upper half plane @xmath283 , it is clear that for any given @xmath284 , we can always find a radius @xmath285 such that @xmath286 for @xmath287 . for such large @xmath35 , the argument of the logarithm in @xmath288 is entirely contained in the circle around unity of radius @xmath155 , where the logarithm is holomorphic . the absolute convergence of @xmath279 thus implies the convergence of the born series for @xmath240 outside a semi - circle of sufficiently large radius @xmath287 . thus the ( born ) series @xmath289 converges absolutely and uniformly at sufficiently large @xmath35 in the upper half plane , finally , we need to show that the difference @xmath290 vanishes at least as @xmath291 for large @xmath35 in the upper half @xmath22-plane . we proceed inductively from eq . ( [ betaintell ] ) using integration by parts . what we will actually show is that the approximation @xmath292 decays as @xmath293 . from the convergence of the born series it is then clear that the remainder @xmath294 vanishes at least as the leading term @xmath295 . to estimate the large-@xmath22 behavior of @xmath296 , we rewrite the integral equation , eq . ( [ betaintell ] ) in the form @xmath297 where @xmath298 \gamma_\ell(k , r_{1})\,.\ ] ] integrating by parts once and estimating the remainder by the riemann - lebesgue lemma , the leading asymptotic behavior of the kernel @xmath299 is easily found to be @xmath300\ , . \label{gap}\ ] ] from this estimate and the limit @xmath301 in the upper half plane , we infer @xmath302 \ , . \label{betaintagain}\ ] ] the starting iteration for the source is @xmath303 , which is independent of @xmath22 ( and @xmath2 ) , whence @xmath304\,.\ ] ] as a check , the large @xmath22 behavior of the first born approximation in the @xmath10-channel is correctly predicted as @xmath305 for the next iteration , we note @xmath306 ^ 2 \sim k^{-2}$ ] ( with a real constant factor ) , so that @xmath307 the general iteration step follows from the assumption that @xmath308 for @xmath309 . to find the behavior of @xmath296 , we need to consider the @xmath61 order term @xmath310 that is given in eq . ( [ gammaexp ] ) . by assumption , each @xmath311 decays as @xmath312 at large @xmath22 . hence all the terms in eq . ( [ gammaexp ] ) are of order @xmath313 . thus the source @xmath314 vanishes as @xmath315 . from eq . ( [ betaintagain ] ) , we easily complete the induction , @xmath316 , as claimed . it should be noted that the @xmath317 estimates in eq . ( [ gap ] ) ( [ gap2 ] ) require the potential @xmath218 to be bounded and sufficiently smooth to allow for integration by parts . moreover , the riemann - lebesgue lemma imposes certain restrictions on @xmath265 and its derivatives , as discussed in the main text . these restrictions are certainly satisfied for smooth bounded potentials from the faddeev class ( [ faddeev ] ) , but more general cases such as a step function may also be handled . having established these two properties of the @xmath318 , it is clear that the auxiliary function satisfying ( a ) and ( b ) in the previous subsection takes the same form as in the antisymmetric channel in one dimension , @xmath319 . \label{auxell}\ ] ] the required properties of @xmath37 follow directly from those of @xmath320 and the convergence of the born series for sufficiently large @xmath35 . the quantity that enters the sum rule is @xmath321 , given by eq . ( [ deltalm ] ) . clearly @xmath321 is the sum of the first @xmath85 terms in the born expansion of @xmath254 . this completes the derivation of the sum rules in @xmath322 partial wave in three dimensions . n. graham , r. l. jaffe , m. quandt and h. weigel , hep - th/0103010 . r. newton , _ scattering theory of waves and particles _ ( mcgraw - hill , new york , 1966 ) ; m. goldberger and k. watson , _ collision theory _ ( john wiley & sons , new york , 1964 ) ; k. gottfried _ quantum mechanics _ ( w. a. benjamin , new york , 1966 ) .	we study scattering theory identities previously obtained as consistency conditions in the context of one - loop quantum field theory calculations . we prove the identities using jost function techniques and study applications .
it is well known that a locally lipschitz function can be estimated point - wise by the riesz potential of its gradient in bounded john domains , ( * ? ? ? * theorem ) , ( * ? ? ? * theorem 10 ) , and hence , especially , in lipschitz domains and in convex domains , ( * ? ? ? * lemma 7.16 ) . by modifying the riesz potential , point - wise estimates can be generalized for functions which are defined in more irregular domains than john domains , ( * ? ? ? * theorem 3.4 ) , ( * ? ? ? * theorem 4.4 ) . more precisely , for every function @xmath4 whose weak distributional partial derivatives are in @xmath5 , the pointwise estimate @xmath6 . here , @xmath7 is a domain in the @xmath1-dimensional euclidean space and the regularity of the boundary is controlled by the function @xmath8 . hedberg s method ( * ? ? ? * lemma , theorem 1 ) can be extended so that this point - wise estimate leads to the sobolev - type inequality where an orlicz - space is the target space . hedberg s method has been used by a. cianchi and b. stroffolini for the classical riesz potential when functions are orlicz functions , ( * ? ? ? * theorem 1 , corollary 1 ) , and by the authors for the modified riesz potential with a special orlicz function , ( * ? ? ? * theorem 1.1 ) and ( * ? ? ? * theorem 1.1 ) , and with a general orlicz function in ( * ? ? ? * corollary 3.4 , corollary 5.4 ) . for other papers on orlicz embeddings of cianchi we refer to @xcite , @xcite . in the present paper we show that the optimal orlicz function for the modified riesz potential in can be found as a function of @xmath8 which depends on the geometry of the domain @xmath7 . our main theorem is the following theorem where we give the formula to the orlicz function . [ thm : defn_h_intro ] let @xmath9 . let the continuous , strictly increasing function @xmath10 be such that @xmath11 and @xmath12 satisfies the @xmath13-condition and the inequality @xmath14 whenever @xmath15 . if @xmath16 then there exists an @xmath17-function @xmath18 that satisfies the @xmath13-condition , and @xmath19 with this function we obtain the following point - wise estimate . [ thm_main ] let @xmath7 be a domain in @xmath20 , @xmath21 . let @xmath3 . if @xmath18 is the function from theorem [ thm : defn_h_intro ] and @xmath22 , then there exists a constant @xmath23 such that the point - wise estimate @xmath24 holds for every @xmath25 . here , @xmath26 is the hardy - littelewood maximal operator of @xmath27 and the constant @xmath23 depends on @xmath1 , @xmath28 , and the @xmath13-constant of @xmath18 only . by this point - wise estimate we obtain embedding results for bounded and unbounded non - smooth domains . examples of these domains are lipschitz domains and convex domains , but also domains with suitable outward cusps are allowed . we define a class of domains which are controlled by the function @xmath8 from . we call these domains in definition [ john ] as @xmath12-cigar john domains , since our definition is a modification of @xcite where j. visl has defined unbounded john domains with @xmath29 . hence , examples of @xmath12-john domains are the classical bounded and unbounded john domains , but also so called @xmath30-john domains when @xmath31 . we have the following corollary which recovers some of the known results of the poincar inequality . [ corollary_bdd ] if there exists @xmath32 such that @xmath33 is increasing for @xmath34 and if @xmath35 is a bounded or an unbounded @xmath12-cigar john domain with a constant @xmath36 in @xmath20 , @xmath37 and if @xmath9 , then with the function @xmath18 in theorem [ thm : defn_h_intro ] there exists a constant @xmath23 such that the inequality @xmath38 holds for every @xmath39 with @xmath40 . here the constant @xmath23 depends on @xmath1 , @xmath28 , @xmath13-constants of @xmath18 and @xmath12 , and john constant @xmath36 only . we point out that if @xmath35 is a bounded @xmath30-john domain , then @xmath41 , @xmath42 , and this corollary yields that the @xmath43-poincar inequality holds . if @xmath44 , the result is optimal . thus the corollary recovers some of the known results of ( * ? ? ? * theorem 10 ) , ( * ? ? ? * corollaries 5 and 6 ) , and ( * ? ? ? * theorem 2.3 ) , but our proof is completely different from the previous proofs . especially , in section 6 we construct an example of an unbounded domain which shows that the lebesque space can not be the target space in this corresponding embedding if @xmath45 . the outline of the paper is as following : we define the domains we consider in section 2 and we call them @xmath12-cigar john domains . we find the suitable orlicz function in section 3 , prove embedding theorems in section 4 , recover some poincar inequalities in section 5 , and in section 6 we construct an example of an unbounded @xmath12-cigar domain . throughout the paper we let the function @xmath10 satisfy the following conditions 1 . @xmath12 is continuous , 2 . @xmath12 is strictly increasing , 3 . @xmath11 , 4 . there exists a constant @xmath46 such that @xmath47 whenever @xmath15 , 5 . @xmath12 satisfies the @xmath13-condition i.e. there exists a constant @xmath48 such that @xmath49 for every @xmath34 . we write @xmath50 now , if @xmath12 satisfies the conditions ( 1)(5 ) , then @xmath8 does , too , and the constant in ( 4 ) is the same for the functions @xmath12 and @xmath8 , that is @xmath51 . the definition of a bounded john domain goes back to f. john ( * ? ? ? * definition , p. 402 ) who defined an inner radius and an outer radius domain , and later this domain was renamed as a john domain in @xcite . we extend the definition of john domains following j. visl @xcite in the classical case . let @xmath52 in @xmath20 , @xmath21 , be a closed rectifiable curve with endpoints @xmath53 and @xmath54 . the subcurve between @xmath55 is denoted by @xmath56 $ ] . for @xmath57 we write @xmath58 \big ) , \ell\big(e[x , b ] \big ) \bigg\},\ ] ] where @xmath59\big)$ ] is the length of the subcurve @xmath60 $ ] . [ john ] a bounded or an unbounded domain @xmath35 in @xmath20 is a @xmath61-cigar john domain if there exists a constant @xmath62 such that each pair of points @xmath63 can be joined by a closed rectifiable curve @xmath52 in @xmath35 such that @xmath64 where @xmath65 is an open ball centered at @xmath66 with a radius @xmath67 and the function @xmath8 is defined as in . the set @xmath68 is called a cigar with core @xmath52 joining @xmath53 and @xmath54 . we point out that if @xmath35 is a @xmath61-cigar john domain with @xmath69 , @xmath70 , then it is a @xmath61-cigar john domain with @xmath71 for every @xmath72 . for the case @xmath73 for all @xmath42 , in definition [ john ] , we refer to @xcite and ( * ? ? ? * and 2.13 ) . if @xmath35 is a bounded domain then the following definition from ( * ? ? * definition 4.1 ) for a @xmath8-john domain gives an equivalent definition to a bounded @xmath61-cigar john domain . [ bounded - john ] a bounded domain @xmath35 in @xmath74 is a @xmath8-john domain if there exist a constants @xmath75 and a point @xmath76 such that each point @xmath77 can be joined to @xmath78 by a rectifiable curve @xmath79 \to d$ ] , parametrized by its arc length , such that @xmath80 , @xmath81 , @xmath82 and @xmath83.\ ] ] the point @xmath78 is called a john center of @xmath35 and @xmath84 is called a john curve of @xmath66 . if the function @xmath8 is defined as in with the function @xmath12 , then a bounded domain is a @xmath8-john domain if and only if it is a @xmath12-john domain . if @xmath85 , then our definition for bounded @xmath8-john domains coincides with the definition of the classical john domains . if @xmath86 , @xmath87 , then our definition for bounded @xmath8-john domains coincides with the definition of @xmath30-john domains . [ thm : john - john ] let @xmath35 be a bounded domain . if @xmath35 is a @xmath8-john domain then @xmath35 is a @xmath61-cigar john domain . on the other hand , if @xmath35 is a @xmath61-cigar john domain with a constant @xmath36 , then @xmath35 is a @xmath8-john domain with constants @xmath88 @xmath89 note that when @xmath90 , then @xmath91 with the same speed as @xmath92 . assume first that @xmath35 is a @xmath8-john domain with a john center @xmath78 . let @xmath63 and let the john curves @xmath93 and @xmath94 connect them to @xmath78 , respectively . we may assume that @xmath95 , since inside the ball the points can be connect by two straight lines going via the center of the ball @xmath96 . let @xmath97 . then , @xmath98 } b \left ( \gamma_1(t ) , \frac{\psi(t)}{\alpha/\dist(x_0 , \partial d ) } \right ) \cup \bigcup_{t \in ( 0 , \ell(\gamma_2 ) ] } b \left ( \gamma_2(t ) , \frac{\psi(t ) } { \alpha/\dist(x_0 , \partial d ) } \right)\\ \end{split}\ ] ] and thus @xmath35 is a @xmath61-cigar john domain . assume then that @xmath35 is a @xmath61-cigar john domain . let us carefully choose a suitable john center so that the center is not too close to the boundary of @xmath35 . let @xmath99 such that @xmath100 . let @xmath52 be a core of a john cigar that connects@xmath66 and @xmath101 . then the length of @xmath52 is at least @xmath102 . let @xmath78 be the center of @xmath52 . then @xmath103 so we choose @xmath104 , and hence @xmath105 . from now on this @xmath106 and the point @xmath78 are fixed in this proof . for every @xmath107 there exists a curve @xmath52 such that @xmath108 . let @xmath109 be the length of @xmath52 , then @xmath110 or by the definition @xmath111 i.e. @xmath112 . to @xmath78 ( the solid line ) , the core @xmath52 ( the dotted line ) and a new carrot given by the constant @xmath113 ( the dashed line).,width=11 ] note that the length of @xmath52 inside the ball @xmath114 is at least @xmath106 and thus for the points in @xmath115 the distance to the boundary is at least @xmath116 . let us choose that @xmath117 since @xmath118 and @xmath8 is increasing , we have @xmath119 . let @xmath120 be the first point from @xmath53 that satisfies @xmath121 . let us replace @xmath122 $ ] by the radius of the ball @xmath114 , if necessary . let us denote this new arc by @xmath52 . let @xmath84 be an arc @xmath52 parametrized by its curve length , such that @xmath123 , @xmath124 . since @xmath125 we obtain that @xmath126.\ ] ] this yields that @xmath127 and thus @xmath128 this yields that we may choose @xmath129 . thus , @xmath35 is a @xmath8-john domain with these @xmath130 and @xmath131 . we note that by the condition ( 4 ) of @xmath12 @xmath132 we recall a covering lemma from ( * ? ? ? * lemma ) which is valid for a bounded @xmath61-john domain . for the previous versions in classical case we refer to ( * theorem 9.3 ) and in a special case to ( * ? ? ? * lemma 3.5 ) . * lemma).[lem : covering ] let @xmath12 satisfies the conditions ( 1)(5 ) . let @xmath133 be defined as in . let @xmath35 in @xmath74 be a bounded @xmath8-john domain with john constants @xmath130 and @xmath131 . let @xmath76 the john center . then for every @xmath134 there exists a sequence of balls @xmath135 such that @xmath136 is in @xmath137 for each @xmath138 and for some constants @xmath139 , @xmath140 , and @xmath141 * @xmath142 ; * @xmath143 , and @xmath144 as @xmath145 ; * no point of the domain @xmath35 belongs to more than @xmath17 balls @xmath146 ; and * @xmath147 . \(1 ) the constant @xmath148 in the previous lemma can be taken to be @xmath149 + ( 2 ) if @xmath35 is a @xmath12-cigar john domain and the john center has been chosen as in theorem [ thm : john - john ] , then @xmath150 as @xmath90 . we recall the following definitions . let @xmath7 be an open set of @xmath20 . we denote the lebegue space by @xmath151 , @xmath152 . by @xmath153 , @xmath152 , we denote those locally integrable functions whose first weak distributional derivatives belongs to @xmath154 i.e. @xmath155 . by @xmath156 , @xmath152 , we denote those functions from @xmath154 whose first weak distributional derivatives belongs to @xmath154 i.e. @xmath157 . theorem [ thm : john - john ] and lemma [ lem : covering ] give the following point - wise estimate which we recall from ( * ? ? ? * theorem ) . [ thm : riesz ] let @xmath12 satisfy the conditions ( 1)(5 ) . let @xmath133 be as defined in . let @xmath35 in @xmath74 be a bounded @xmath12-cigar john domain with a john constant @xmath36 . then there exists a finite constant @xmath23 and @xmath76 such that for every @xmath158 and for almost every @xmath77 the inequality @xmath159 holds . here we recall the definitions of @xmath17-functions and orlicz spaces . a function @xmath161 is an @xmath17-function if 1 . @xmath18 is continuous , 2 . @xmath18 is convex , 3 . @xmath162 and @xmath163 . continuity and @xmath162 yield that @xmath164 . let @xmath165 by convexity @xmath166 and thus @xmath167 for @xmath165 . this implies that @xmath18 is a strictly increasing function . by the notation @xmath168 we mean that there exists a constant @xmath169 such that @xmath170 for all @xmath66 . the notation @xmath171 means that @xmath172 . two @xmath17-functions @xmath18 and @xmath148 are equivalent , which is written as @xmath173 , if there exists @xmath174 such that @xmath175 for all @xmath34 . equivalent @xmath17-functions give the same space with comparable norms . we point out that @xmath173 if and only if for the inverse functions @xmath176 . we assume that @xmath18 satisfies the @xmath13-condition , that is , there exists a constant @xmath177 such that @xmath178 if an @xmath17-function satisfies the @xmath13-condition then the relations @xmath179 and @xmath180 are equivalent . the constant @xmath177 is called the @xmath13-constant of @xmath18 . let @xmath7 in @xmath20 be an open set . the orlicz class is a set of all measurable functions @xmath4 defined on @xmath7 such that @xmath181 we study the orlicz space @xmath182 which means the space of all measurable functions @xmath4 defined on @xmath7 such that @xmath183 for some @xmath184 . whenever the function @xmath18 satisfies the @xmath13-condition , then the space @xmath182 is a vector space and it is equivalent to the corresponding orlicz class . we study these orlicz spaces and call their functions orlicz functions . the orlicz space @xmath182 equipped with the luxemburg norm @xmath185 is a banach space . we recall the following theorem from ( * ? ? ? * theorem ) . [ thm : pointwise - maximal ] let @xmath12 satisfy the conditions ( 1)(5 ) . let @xmath133 be defined as in . let @xmath9 be given . suppose that there exists a continuous function @xmath186 such that @xmath187 let @xmath188 be a continuous function and let @xmath189 be an n - function satisfying the @xmath190-condition . suppose that there exists a finite constant @xmath191 such that the inequality @xmath192 holds for all @xmath34 . let @xmath7 in @xmath20 be an open set . if @xmath22 , then there exists a constant @xmath23 such that the inequality @xmath193 holds for every @xmath25 . here the constant @xmath23 depends on @xmath1 , @xmath28 , @xmath194 , @xmath191 , and the @xmath13-constants of @xmath12 and @xmath18 only . our goal is to find a formula which would give all suitable functions @xmath18 . examples of some of these functions were given in ( * ? ? ? * section 6 ) . here we do the preparations to find @xmath18 . assume that there exists @xmath195 such that @xmath33 is increasing for @xmath34 . this yields that @xmath196 is increasing , too . under this condition inequality holds : since @xmath197 we have @xmath198 let us define the functions @xmath199 and @xmath200 such that @xmath201 then , @xmath202 if we choose @xmath203 and assume that the inverse function of @xmath204 exists , that is @xmath205 exists , then @xmath206 and thus @xmath207 unfortunately , there is a problem with this function @xmath208 to be a suitable function @xmath18 ; namely , the function @xmath208 is not necessary convex . for example , if @xmath209 , @xmath210 , and @xmath211 , then the function @xmath208 is not convex , see figure [ fig : not_convex ] . the angle at the point @xmath212 comes from the angle of @xmath8 at the point @xmath213 . our main theorem , theorem [ thm : defn_h_intro ] in introduction , corrects this point : we show that there exists an @xmath17-function @xmath18 that is equivalent with @xmath208 . is not necessary convex.,width=226 ] let us write that @xmath214 for @xmath34 and @xmath215 . let us first show that @xmath204 is strictly increasing . assume then that @xmath216 . the inequality @xmath217 is equivalent to the inequality @xmath218 recall that if @xmath12 satisfies the condition @xmath219 , then @xmath8 does , too , and the constant is the same for both functions . thus by the condition ( 4 ) and the inequality @xmath220 we obtain @xmath221 thus the function @xmath204 is strictly increasing . this yields that the function @xmath208 exists and is strictly increasing . let us show that @xmath222 . since @xmath223 we obtain @xmath224 let us show that @xmath225 . since @xmath226 is decreasing , by the condition ( 4 ) , and by @xmath220 we obtain @xmath227 we have shown that @xmath228 is bijective . let us then study the condition @xmath229 since @xmath204 is a strictly increasing bijection , inequality is equivalent to @xmath230 since @xmath33 is increasing , then @xmath231 is decreasing and @xmath232 is decreasing , too . we note that @xmath233 , since @xmath234 . we obtain @xmath235 and thus inequality holds . let us then show that @xmath236 for all @xmath237 with @xmath238 . the inequality @xmath236 is equivalent to @xmath239 by the condition @xmath219 of @xmath12 and the inequality @xmath220 , we obtain @xmath240 the inequality @xmath236 yields that @xmath208 satisfies the @xmath13-condition . let us write @xmath241 . then @xmath242 . since @xmath208 is increasing , we have @xmath243 p. hst has shown in ( * ? ? ? * proposition 5.1 ) that if @xmath244 satisfies the @xmath13-condition and @xmath245 is increasing , then @xmath27 is equivalent to a convex function . since @xmath208 satisfies inequality and the @xmath13-condition , we obtain that @xmath208 is equivalent to a convex function @xmath18 . using @xmath246 and the bijectivity , we obtain @xmath247 and thus also @xmath162 . this gives that @xmath18 is right continuous at the origin . thus by convexity the function @xmath18 is continuous on @xmath248 . since @xmath231 is decreasing and @xmath234 , we obtain @xmath249 since the functions @xmath208 and @xmath18 are equivalent , this yields that @xmath250 thus we have shown that the function @xmath18 satisfies the conditions ( n1 ) ( n3 ) . [ rem : h - pienilla - arvoilla ] later it is crucial to us that @xmath251 for @xmath252 . namely , then for every @xmath12 the function @xmath18 satisfies @xmath253 whenever @xmath252 . functions @xmath254 , @xmath255 and @xmath256 , satisfy the assumptions of theorem [ thm : defn_h_intro ] . now , the proof for our second main theorem , theorem [ thm_main ] in introduction , follows easily : theorem [ thm : pointwise - maximal ] and theorem [ thm : defn_h_intro ] . as a corollary we obtain from theorem [ thm_main ] and theorem [ thm : riesz ] : [ cor : funktiosta - gradientiin ] let @xmath9 . let the function @xmath18 be as in theorem [ thm : defn_h_intro ] . if @xmath35 is a bounded @xmath12-cigar john domain with a constant @xmath36 , then there exit a constant @xmath23 and a point @xmath76 such that the point - wise estimate @xmath257 holds for all @xmath258 with @xmath259 and for almost every @xmath77 . here the constant @xmath23 depends on @xmath1 , @xmath28 , @xmath191 , @xmath260 , @xmath261 , @xmath36 , @xmath262 and @xmath263 only . corollary [ cor : funktiosta - gradientiin ] is essential in the proofs of the following theorem [ thm : sobolev - poincare - p>1 ] and theorem [ thm : sobolev - poincare - p=1 ] . [ thm : sobolev - poincare - p>1 ] assume that @xmath12 satisfies the conditions @xmath264 @xmath265 , @xmath266 in the condition @xmath219 , and there exists @xmath195 such that @xmath33 is increasing for @xmath34 . let @xmath8 be defined as in . let @xmath267 , @xmath37 , be a bounded @xmath12-cigar john domain with a constant @xmath36 . let @xmath268 . then there exists an @xmath17-function @xmath18 , that satisfies @xmath13-condition and @xmath269 and there exists a constant @xmath270 such that the inequality @xmath271 holds for every @xmath272 . here the constant @xmath23 depends on @xmath1 , @xmath28 , @xmath260 , @xmath261 , @xmath36 and @xmath273 only . assume that @xmath259 . corollary [ cor : funktiosta - gradientiin ] yields that @xmath274 where the constant @xmath23 depends on @xmath1 , @xmath28 , @xmath260 , @xmath261 , @xmath36 , and @xmath273 only . by integrating over @xmath35 and using the fact that the maximal operator is bounded whenever @xmath275 , we obtain that @xmath276 this yields that the inequality @xmath277 holds for every @xmath258 with @xmath259 . by applying this inequality to the function @xmath278 we obtain that @xmath279 we may assume w.l.o.g . that @xmath280 . let denote @xmath281 . by the triangle inequality @xmath282 here , @xmath283 where @xmath284 is the conjugate function of @xmath18 and @xmath23 is the constant in hlder s inequality . next we show that @xmath285 . since the function @xmath18 is continuous and strictly increasing , there exists a unique @xmath286 such that @xmath287 i.e. @xmath288 . by solving @xmath289 we obtain @xmath290 similarly , we obtain @xmath291 since @xmath292 for all @xmath293 , see for example ( * ? ? ? * lemma 2.6 , p. 56 ) , we obtain that @xmath294 hence , we have shown that @xmath295 for every @xmath272 . [ exm : h - funktio ] let us choose that @xmath41 , @xmath296 . we have calculated in remark [ rem : h - pienilla - arvoilla ] that for every @xmath12 the function @xmath18 satisfies @xmath253 whenever @xmath252 . if @xmath297 , then @xmath298 and thus we have that @xmath299 for @xmath300 . [ thm : sobolev - poincare - p=1 ] assume that the function @xmath12 satisfies the conditions @xmath264 @xmath265 , @xmath266 in the condition @xmath219 , and there exists @xmath195 such that @xmath33 is increasing for @xmath34 . let @xmath8 be defined as in ( [ equ : psi ] ) let @xmath267 , @xmath37 , be a bounded @xmath12-cigar john domain with a constant @xmath36 . then there exists an @xmath17-function @xmath18 , that satisfies @xmath13-condition and @xmath301 such that the inequality @xmath302 holds for some constant @xmath23 and for every @xmath303 . here the constant @xmath23 depends on @xmath1 , @xmath260 , @xmath261 and @xmath36 only . let us consider functions @xmath304 such that @xmath305 . the center ball @xmath306 is written as @xmath307 . in the proof of theorem [ thm : john - john ] we had chosen @xmath78 so that @xmath308 . we show that there exists a constant @xmath270 such that the inequality @xmath309 holds whenever @xmath310 . this yields the claim as in the proof of theorem [ thm : sobolev - poincare - p>1 ] . since @xmath18 is increasing , we first estimate @xmath311 let us define @xmath312 for all @xmath77 . if @xmath313 , then @xmath314 . we obtain @xmath315 by the triangle inequality we have @xmath316 by the @xmath317-poincar inequality in a ball @xmath307 , ( * ? ? ? * section 7.8 ) , there exists a constant @xmath318 such that @xmath319 we continue to estimate the right hand side of inequality @xmath320 where @xmath321 . assume first that @xmath322 is so large that @xmath323 . when @xmath324 , then @xmath325 by and thus @xmath326 thus for @xmath324 we obtain that @xmath327 . this yields that @xmath328 this constant does not blow up when @xmath90 : @xmath329 assume then that @xmath92 is small . this yields that for every @xmath330 the sum @xmath331 is finite and depends on @xmath332 . we obtain @xmath333 then , we will find an upper bound for the sum @xmath334 since @xmath335 , corollary [ cor : funktiosta - gradientiin ] yields that @xmath336 we choose for every @xmath337 a ball @xmath338 , centered at @xmath66 and with radius @xmath339 depending on @xmath66 , such that @xmath340 with the understanding that @xmath341 is zero outside @xmath35 . by the besicovitch covering theorem ( or the @xmath342-covering theorem ) we obtain a subcovering @xmath343 so that we may estimate by the @xmath13-condition of @xmath18 @xmath344 let @xmath345 . since @xmath341 is zero almost everywhere in @xmath346 and @xmath347 for almost every @xmath348 , we obtain @xmath349 estimates , , and imply inequality . corollary [ corollary_bdd ] in introduction follows from theorem [ thm : sobolev - poincare - p>1 ] and theorem [ thm : sobolev - poincare - p=1 ] . in theorem [ thm : sobolev - poincare - p=1 ] the @xmath17-function @xmath18 is the best possible in a sense that it can not be replaced by any @xmath17-function @xmath148 that satisfies the @xmath13-condition and @xmath350 in ( * ? ? ? * theorem 7.2 ) we have shown that the corresponding embedding in theorem [ thm : sobolev - poincare - p=1 ] does not hold if @xmath351 this is valid for this function @xmath148 . by the definitions of @xmath352 and @xmath8 we obtain that @xmath353 and thus there does not exists a constant @xmath354 such that @xmath355 for every @xmath303 . [ thm : sobolev - poincare - unbounded ] assume that the function @xmath12 satisfies the conditions @xmath264 @xmath265 , @xmath266 in the condition @xmath219 , and there exists @xmath195 such that @xmath33 is increasing for @xmath34 . let the function @xmath8 be defined as in . let @xmath35 in @xmath20 , @xmath37 , be an unbounded domain that satisfies the following conditions : * @xmath356 , where @xmath357 ; * @xmath358 for each @xmath359 ; * each @xmath360 is a bounded @xmath12-cigar john domain with a constant @xmath36 . let @xmath9 . then , there exists an @xmath17-function @xmath18 , that satisfies @xmath13-condition and @xmath361 and there exits a constant @xmath23 such that the inequality @xmath362 holds for every @xmath363 . here the constant @xmath23 depends on @xmath1 , @xmath28 , @xmath260 , @xmath261 and @xmath36 only . the proof follows the proof of ( * ? ? ? * theorem 4.1 ) . by theorems [ thm : sobolev - poincare - p>1 ] and [ thm : sobolev - poincare - p=1 ] there exits a constant @xmath23 such that the inequality @xmath364 holds for each @xmath360 and all @xmath303 . the constant @xmath23 does not blow up when the diameter of @xmath360 tends to infinity . in the case @xmath275 this is clear . in the case @xmath44 , we refer to the discussion after in the proof of theorem [ thm : sobolev - poincare - p=1 ] . the constant depends on @xmath365 but this does not cause a problem . when @xmath366 inequality yields that there exists a constant @xmath270 such that the inequality @xmath367 holds ; here the constant @xmath23 is independent of @xmath359 . let us write @xmath368 the triangle inequality yields that @xmath369 since @xmath360 satisfies inequality , we have @xmath370 and thus the second term is finite . again , by inequality we obtain that @xmath371 thus the real number sequence @xmath372 is bounded and hence there exists a convergent subsequence @xmath373 and @xmath374 such that @xmath375 . since @xmath18 is continuous , @xmath376 fatou s lemma and the modular form of yield that @xmath377 for every @xmath39 with @xmath378 . this yields that there exists a constant @xmath23 such that the inequality @xmath379 holds for every @xmath258 with @xmath259 . the claim follows by applying this inequality to the function @xmath278 . let the function @xmath12 be defined as in theorem [ thm : sobolev - poincare - unbounded ] . the following unbounded domains satisfy the assumptions of theorem [ thm : sobolev - poincare - unbounded ] . \(a ) @xmath380 . \(b ) @xmath381 . as a special case we recover results for poincar domains . we recall that a bounded domain @xmath35 is called a @xmath382-poincar domain , where @xmath383 , if there is a constant @xmath270 such that the inequality @xmath384 holds for all @xmath385 . inequality is the @xmath382-poincar inequality . we note that for a bounded domain @xmath35 inequality holds if and only the inequality @xmath386 holds , the constants @xmath23 and @xmath387 depend on each other and @xmath388 only . let us recall results for bounded @xmath12-john domains in the case @xmath41 , for a fixed @xmath87 . a bounded @xmath389-john domain is usually called @xmath30-john domain . a bounded @xmath30-john domain is a @xmath390-poincar domain whenever @xmath391 , ( * ? ? ? * theorem 10 ) . so , a bounded @xmath30-john domain is a @xmath390-poincar domain for all @xmath70 if @xmath392 . a bounded @xmath30-john domain is a @xmath393-poincar domain if @xmath394 , where @xmath395 $ ] is the minkowski dimension of the boundary of the domain , ( * ? ? ? * theorem 1.3 ) . a bounded @xmath30-john domain is a @xmath396-poincar domain for every @xmath397 if @xmath398 $ ] ; ( * ? ? ? * corollaries 5 and 6 ) , ( * ? ? ? * theorem 2.3 ) . the exponent @xmath399 is the best possible in the class of bounded @xmath30-john domains , we refer to @xcite . our theorems [ thm : sobolev - poincare - p>1 ] and [ thm : sobolev - poincare - p=1 ] with @xmath400 give that a bounded @xmath30-john domain is a @xmath43-poincar domain if @xmath401 and @xmath402 . thus our result is optimal in the case @xmath44 . on the other hand , our method does not cover the case @xmath403 . note that our proof totally differs from the previous proofs in ( * ? ? ? * theorem 10 ) , ( * ? ? ? * corollaries 5 and 6 ) , and ( * ? ? ? * theorem 2.3 ) . in this section we give an example which shows that for certain unbounded @xmath12-cigar john domains the target space can not be a lebesgue space . the idea is that at near the infinity the target space should be @xmath404 but local structure of the domain may not allow so good integrability . we assume a priori that the function @xmath12 has the properties ( 1)(5 ) . later on we give extra conditions to the function @xmath12 . we construct a mushrooms - type domain . let @xmath405 be a decreasing sequence of positive real numbers converging to zero . let @xmath406 , @xmath407 , be a closed cube in @xmath20 with side length @xmath408 . let @xmath409 , @xmath407 , be a closed rectangle in @xmath20 which has side length @xmath410 for one side and @xmath411 for the remaining @xmath412 sides . let @xmath413 be the first quarter of the space i.e. all coordinates of the points in @xmath413 are positive . we attach @xmath406 and @xmath409 together creating mushrooms which we then attach , as pairwise disjoint sets , to the side @xmath414 of @xmath413 so that the distance from the mushroom to the origin is at least @xmath415 and at most @xmath416 , see figure [ fig : domain ] . we assume that a priori the function @xmath12 has the properties ( 1)(5 ) , but we have to assume here also that @xmath417 . we need copies of the mushrooms . by an isometric mapping we transform these mushrooms onto the side @xmath418 of @xmath413 and denote them by @xmath419 and @xmath420 . so again the distance from the mushroom to the origin is at least @xmath415 and at most @xmath416 . we define @xmath421 see figure [ fig : domain ] . we omit a short calculation which shows that @xmath7 is a @xmath12-cigar john domain . let us define a sequence of piecewise linear continuous functions @xmath422 by setting @xmath423 where the function @xmath208 will be given in . then the integral average of @xmath424 over @xmath7 is zero for each @xmath425 . the gradient of @xmath426 differs from zero in @xmath427 only and @xmath428 note that @xmath429 we require that @xmath430 hence , we define @xmath431 let @xmath18 be an @xmath17-function . then , @xmath432 hence , we have @xmath433 thus , there does not exist a positive constant @xmath23 such that the inequality @xmath434 could hold for all @xmath4 from the appropriate space if @xmath435 assume that @xmath436 . if @xmath437 , then we obtain that the inequality does not hold if @xmath438 assume then that we have a sequence @xmath439 of positive numbers going to infinity . for each @xmath440 we may choose points @xmath441 and @xmath442 such that the balls @xmath443 and @xmath444 are subsets of the first quadrant and @xmath445 . we define a sequence of piecewise linear continuous functions @xmath446 by setting @xmath447 now we have @xmath448 for some constant @xmath23 . on the other hand , for any @xmath449 @xmath450 thus , there does not exist a positive constant @xmath387 such that the inequality @xmath451 could hold for all @xmath4 from the appropriate space if @xmath452 if @xmath436 and if there were an embedding with the lebesgue space @xmath454 as a target space , then by we would have @xmath455 and by we would have @xmath456 . thus the target space can not be a lebesgue space . the target space can be @xmath454 only if @xmath457 . and in this case @xmath458 . note that the limit @xmath459 exists since @xmath12 is increasing and @xmath460 . if @xmath461 , then there exists @xmath462 such that @xmath463 . let @xmath12 satisfy ( 1)(5 ) , and assume that @xmath45 . let @xmath7 be the unbounded @xmath12-cigar john domain constructed in . let @xmath464 . then there do not exist numbers @xmath465 and @xmath466 such that the inequality @xmath467 could hold for all @xmath468 . let @xmath12 satisfy ( 1)(5 ) , and assume that @xmath469 . let @xmath7 be the unbounded @xmath12-cigar john domain constructed in . assume that there exist numbers @xmath465 and @xmath466 such that the inequality @xmath467 holds for all @xmath468 . then @xmath458 and there exists @xmath470 such that @xmath471 for all @xmath472 $ ] . p. hajasz , _ sobolev inequalities , truncation method , and john domains _ , papers on analysis : a volume dedicated to olli martio on the occasion of his 60th birthday . edited by j. heinonen , t. kilpelinen , and p. koskela , report univ . jyvskyl , * 83 * , university of jyvskyl , jyvskyl , 2001 , pp . 109126 . g. reshetnyak , _ integral representations of differentiable functions in domains with nonsmooth boundary _ ( russian ) , _ sibirsk . mat . 21(1980 ) , 108116 ; translation in _ sib . math . j. _ 21 * ( 1981 ) , 833839 .	@xmath0-functions which are defined in non - smooth domains in the @xmath1-dimensional euclidean space can be estimated point - wise by the modified riesz potential of their gradients . these point - wise estimates imply embeddings into orlicz spaces from the space @xmath2 , @xmath3 , where the functions are defined in bounded or unbounded domains with minimum requirement of the smoothness of the boundary . the results are sharp for @xmath0-functions .
the cooler synchrotron cosy - jlich provides proton and deuteron beams phase - space cooled and polarized if desired with momenta up to 3.7 gev / c . it is thus well suited to produce the light scalars @xmath0(980 ) since masses up to 1.1 ( 1.5 , 1.03 ) gev / c@xmath4 can be produced in @xmath5 ( @xmath6 , @xmath7 ) collisions . such hadronic interactions offer a few particular advantages : * due to large production cross sections rare processes , like radiative decays into vector mesons @xmath8 @xcite or the isospin - violating @xmath9-@xmath10 mixing @xcite , are expected to occur at reasonable count rates . * the isospin of the initial state and of the produced scalar meson can be selected . @xmath11 or @xmath12 reactions must lead to @xmath13 ( @xmath14 ) production , a @xmath15 , @xmath16 ( using a deuterium target ) or a @xmath17 reaction can produce both the @xmath9 and the @xmath10 , whereas the @xmath18 process is a filter for the @xmath10 ( @xmath19 ) resonance , because the initial deuterons and the @xmath20 particle in the final state both have isospin @xmath19 . * for the @xmath21 decays the maximum accessible excess energy @xmath22 is rather small . thus with a forward magnetic spectrometer like anke ( see below ) , large acceptances and an unprecedented mass resolution @xmath23 can be reached . this is important to unravel effects induced by the opening of the @xmath24 and @xmath25 thresholds , which are separated by only 8 mev / c@xmath4 . at the same time the following drawbacks should be mentioned : * also the cross sections for background processes like multi - pion production are large . * the final states contain at least two baryons . therefore , the scalar meson signal can be distorted by final - state interactions ( fsi ) between these baryons and/or between one or more baryons and the mesons from @xmath0 decays . this effect has , _ e.g. _ been seen for the @xmath26 reaction @xcite and also in @xmath27 data @xcite . on the other hand such experiments can be exploited for the investigation of the low energy @xmath28 and @xmath29 interactions , see _ e.g. _ the analyses in refs . @xcite . @xmath0 production has been or will be studied at cosy in @xmath5 , @xmath30 , @xmath6 and @xmath7 interactions for the strong decays into @xmath1 and @xmath2/@xmath3 as well as radiative decays @xmath31 into vector mesons . while near - threshold decay channels with at least one charged kaon can well be investigated with the magnetic spectrometer anke ( `` apparatus for the detection of nucleonic and kaon ejectiles '' ) , the @xmath2 , @xmath3 and @xmath31 final states will be measured with wasa ( `` wide angle shower apparatus '' ) which is available for measurements at cosy since 2007 . the magnetic anke spectrometer @xcite consists of three dipoles and detection systems for identification of charged particles emitted under forward angles . for our measurements an h@xmath32/d@xmath32 cluster - jet target , which can provide areal densities of up to @xmath33 @xmath34s@xmath35 , has been used . together with @xmath36 particles in the cosy ring , this corresponds to luminosities up to a few times @xmath37 @xmath34s@xmath35 . @xmath38-mesons are detected in a positive side detection system @xcite , using time - of - flight ( tof ) measurement between 23 scintillation start counters , which are placed near a side exit window of the spectrometer magnet , and the range telescopes system or a wall of scintillation counters . the momentum reconstruction algorithm uses the track information provided by two multiwire proportional chambers ( mwpcs ) . this information as well as the kaon energy losses in the scintillators are used in order to suppress background . high momentum particles ( @xmath39 , @xmath40 , @xmath41 or he ) produced in coincidence with the kaons are detected by a forward detection system which consists of three mwpcs ( used for momentum reconstruction ) and two layers of scintillation counters ( particle i d ) . as a selection criteria , the energy loss of the particles and time difference between the hits in the side and forward systems are used . @xmath42-mesons are observed in a negative side detection system containing layers of scintillation counters and two mwpcs , which also provide the possibility to use the time difference between negative and positive detection systems , and @xmath43 techniques and to reconstruct @xmath42 momenta @xcite . the 4@xmath44 detector facility wasa @xcite has been designed for studies of production and decays of light mesons . wasa makes use of a hydrogen and deuterium pellet target . the pellet concept is crucial to achieve a close to 4@xmath44 detection acceptance in this internal target storage ring experiment . the target system provides small spheres of frozen hydrogen or deuterium and allows for high luminosities of up to @xmath45 @xmath34s@xmath35 . the wasa detectors comprise a forward part for measurements of charged target - recoil particles and scattered projectiles ( @xmath39 , @xmath40 , @xmath41 or he ) , and a central part for measurements of the scalar meson decay products . the forward part consists of eleven planes of plastic scintillators and of proportional counter drift tubes . the central part comprises an electromagnetic calorimeter with @xmath46 csi(na ) crystals surrounding a superconducting solenoid . inside the solenoid a cylindrical chamber with drift tubes and a plastic scintillator barrel are placed . close - to - threshold data on kaon pair production in nucleon - nucleon scattering , like in the reactions @xmath48 or @xmath49 , allow to study the @xmath1 and @xmath50 subsystems in the final state . the strength of the @xmath1 interaction is of relevance for a possible @xmath1 molecule interpretation of the scalar resonances @xmath9(980 ) and @xmath10(980 ) . similarly , a better understanding of the @xmath50 system is one prerequisite to infer the nature of the @xmath51(1405 ) . the @xmath48 reaction has been measured with anke at two proton kinetic energies @xmath52 gev , and 2.83 gev , corresponding to excess energies of @xmath53 mev and 105 mev with respect to the @xmath54 threshold @xcite . figure [ fig : k+k0 ] shows the invariant @xmath55 mass distribution ( normalized to the phase - space volume ) for the higher beam energy . the data seem to indicate a two - peak structure which clearly deviates from the expected flatt - like behaviour for the @xmath13 resonance ( indicated by the lines ) . while the enhancement at higher @xmath1 masses probably is a reflection of the @xmath56 fsi @xcite , there is no obvious interpretation of the lower one . interestingly , the same peak structure ( although , due to the lower mass resulution , only supported by the lowest - mass data point ) is also observed for data from @xmath57 annihilations . a possible explanation of the unexpected shape is that interference effects by chance lead to the same structure in both data sets . invariant @xmath55 mass distributions normalized to the phase - space volume . the crystal - barrel data ( open squares ) are from the @xmath58 reaction @xcite ; the anke data ( full circles ) for @xmath59 at @xmath60 mev @xcite . the mass resolution of the anke data is @xmath61 mev / c@xmath4 ( fwhm ) . the lines denote the flatt fits to crystal - barrel data from refs . @xcite ( solid ) and @xcite ( dashed).,scaledwidth=80.0% ] the measurements of the @xmath62 reaction were performed at three energies of @xmath52 gev , 2.70 gev and 2.83 gev , _ i.e. _ at @xmath22 values of 51 mev , 67 mev and 108 mev with respect to the @xmath63 threshold . figure [ fig : k+k- ] presents the invariant @xmath24 mass resolution for the lowest beam energy . the lines show the expected shape of the distribution from a monte - carlo simulation that takes into account the anke acceptance @xcite . three data points at the lowest @xmath1 masses lie significantly above the simulation . this behaviour is also visible for the other two beam energies and in disto results on the same reaction at @xmath64 mev @xcite as well as in our data on the @xmath65 reaction @xcite . this effect is demonstrated in the right part of fig . [ fig : k+k- ] where the anke differential cross section is normalized to the simulated spectra . at all three beam energies a sigificant enhancement between the @xmath24 and @xmath66 thresholds is observed . left : differential cross section for the @xmath67 reaction with respect to the @xmath24 invariant mass at @xmath52 gev @xcite . the solid ( black ) line shows the expected shape of the distribution from a monte - carlo simulation that takes into account the @xmath68 , non-@xmath69 contributions as well as the @xmath70 and @xmath5 fsis ( best fit ) @xcite . for illustration , the dashed ( blue ) and dotted ( red ) lines show the shape of the non-@xmath69 mass distribution with the @xmath24 being in an s(p ) wave . right : data for all measured energies 2.65 gev , 2.70 gev , and 2.83 gev normalized to the best - fit distributions . the mass resolution is @xmath71 mev / c@xmath4 ( fwhm).,title="fig:",scaledwidth=35.0% ] left : differential cross section for the @xmath67 reaction with respect to the @xmath24 invariant mass at @xmath52 gev @xcite . the solid ( black ) line shows the expected shape of the distribution from a monte - carlo simulation that takes into account the @xmath68 , non-@xmath69 contributions as well as the @xmath70 and @xmath5 fsis ( best fit ) @xcite . for illustration , the dashed ( blue ) and dotted ( red ) lines show the shape of the non-@xmath69 mass distribution with the @xmath24 being in an s(p ) wave . right : data for all measured energies 2.65 gev , 2.70 gev , and 2.83 gev normalized to the best - fit distributions . the mass resolution is @xmath71 mev / c@xmath4 ( fwhm).,title="fig:",scaledwidth=35.0% ] the mass scale of the low - mass variation in fig . [ fig : k+k- ] is not that of the widths of the scalar resonances , which are much larger . it is tempting to suggest that this structure might be due to the opening of the @xmath25 channel at a mass of 0.995 gev / c@xmath4 , which induces some cusp structure that also changes the energy dependence of the total cross section near threshold . this would require a very strong @xmath72 channel coupling , which might be driven by the @xmath0 resonances . thus , although the @xmath73 reaction may not be ideal for investigating the properties of scalar states , their indirect effects might still be crucial . an evident test of isospin symmetry is to scrutinize the @xmath74 and @xmath13 mass distributions . so far , the best knowledge of the @xmath74 shape comes from a flatt fit to the @xmath75 mass distribution from @xmath76 annihilations , measured with crystal barrel at lear @xcite . the fit yields the coupling constants with small statistical uncertainty , however , the @xmath75 channel could , in principle , be distorted by isospin - violating @xmath74-@xmath10 mixing effects , _ e.g. _ distortions of a basically symmetric breit - wigner type ( @xmath77 ) @xmath9 mass distribution by admixtures of the ( @xmath78 ) @xmath10 resonance . this effect might be sizable if the @xmath79 cross section is significantly larger than for the @xmath74 @xcite it is thus desirable to obtain high statistics data for the @xmath80 @xmath13 state , where mixing effects with the @xmath10 are strictly forbidden . the best data so far on the @xmath81 decay from an experiment at bnl on pion - induced reactions @xcite are displayed in fig . [ fig : a0-vs .- a+ ] together with the above mentioned crystal - barrel data . the measured @xmath13 shape can as well be fitted by a flatt distribution as by a breit - wigner ( the latter corresponds to the case of zero coupling of the @xmath13 to kaons ) . invariant @xmath75 mass distribution from @xmath82 crystal - barrel data ( green squares ) @xcite in comparison with that for the positive charge state @xmath83 from pion - induced reactions ( blue circles ) @xcite . the line denotes a flatt fit from ref . @xcite to the crystal - barrel data.,scaledwidth=70.0% ] it has been predicted that @xmath9-@xmath10 mixing can lead to a comparatively large isospin violation in the reactions @xmath84 and @xmath85 close to the corresponding production thresholds @xcite . one may either test whether the cross - section ratios for these reactions follow isospin relations , or whether the @xmath74 and @xmath13 mass distributions exhibit different shapes . both require data with high statistical accuracy for the strong scalar decay channels and backgound , _ e.g. _ from non - resonant @xmath2 production must be well understood . a first test beam time took place in november 2007 at wasa and the data are currently being analyzed . the @xmath86 and @xmath87 data that have been taken at wasa in november 2007 will also be used to search for events from radiative decays into vector mesons @xmath88 . according to our estimates @xcite a few thousand of such events could be measured in 10 weeks of beam time , however , scheduling of such a long beam time at cosy has to await the result of the above mentioned test measurement . another proposed measurement for wasa aims at data for the isospin - violating @xmath89 reaction @xcite . that cross section is expected to be dominated by a primary reaction of the type @xmath90 , followed by an @xmath91 conversion , and an isospin - conserving @xmath92 decay . in order to determine the @xmath91 mixing strength , the @xmath93 production cross section must be experimentally determined . for that purpose , the @xmath94 reaction has been measured at anke . a preliminary analysis yields a total cross section in the range 50100 pb . if that cross section is dominated by the @xmath95 decay , then isospin violation in the @xmath96 reaction should be measurable at wasa within a few weeks of beam time . this work has been supported by : deutscher akademischer austausdienst , deutsche forschungsgemeinschaft , china scholarship council , cosy - ffe program , helmholtz - gemeinschaft , russian academy of science , russian fund for basic research , european community . contributions by the anke and wasa collaborations are greatfully acknowledged . m. bscher _ et al . _ , `` first measurement of radiative decays of scalar mesons @xmath97 with wasa '' , cosy proposal # 174 , available via www.fz-juelich.de/ikp/publications/list_of_all_cosy-proposals.shtml . m. bscher _ et al . _ , `` investigation of the neutral scalar mesons @xmath9/@xmath10(980 ) in the reaction @xmath98 with anke '' , cosy proposal # 140 , available via www.fz-juelich.de/ikp/publications/list_of_all_cosy-proposals.shtml .	the light scalar mesons @xmath0(980 ) are being investigated at cosy - jlich by detecting the strong decays into @xmath1 and @xmath2/@xmath3 as well as radiative decays into vector mesons . selected results are discussed with emphasis on recent measurements at the anke and wasa spectrometers . address = institut fr kernphysik , forschungszentrum jlich , 52425 jlich , germany
let @xmath9 and @xmath10 be a nonnegative integer or infinity , and denote by @xmath11 the indicator function on a set @xmath12 . in this note , we consider _ bilevel causal signals _ @xmath13 with unknown transition values ( positions ) @xmath14 , satisfying @xmath15 and also a uniform generalized sampling process @xmath16 with sampling kernel @xmath6 being causal and uniform sampling taken every @xmath17 seconds . for the bilevel causal signal @xmath0 in ( [ bilevel.def ] ) , define its _ maximal local rate of innovation _ @xmath1 by reciprocal of the maximal positive number @xmath18 such that there is at most one transition position @xmath14 , in any time period @xmath19 , that is , @xmath20 the concept of signals with finite rate of innovation was introduced by vetterli , marziliano and blu @xcite . examples of signals with finite rate of innovation include streams of diracs , piecewise polynomials , band - limited signals , and signals in a finitely - generated shift - invariant space @xcite@xcite . in the past ten years , the paradigm for reconstructing signals with finite rate of innovation from their samples has been developed , see for instance @xcite , @xcite and @xcite@xcite and references therein . precise identification of transition positions is important to reach meaningful conclusions in many applications . vetterli , marziliano and blu show in @xcite that a bilevel signal @xmath0 defined in ( [ bilevel.def ] ) can be reconstructed from its samples ( [ sampling.def ] ) when the sampling kernel @xmath6 is the box spline @xmath21 ( or the hat spline @xmath22 ) and the sample rate @xmath8 is at ( or above ) the maximal local rate of innovation @xmath1 of the signal @xmath0 . in this note , we show that bilevel causal signals @xmath0 defined in ( [ bilevel.def ] ) are uniquely determined from their samples @xmath5 , in ( [ sampling.def ] ) if the sampling kernel @xmath6 is causal and positive on @xmath23 , and the sample rate @xmath24 is at ( or above ) the maximal local rate of innovation @xmath1 , see theorem [ uniqueness.tm ] . our numerical simulations indicate that the bilevel signal recovery procedure from noisy samples @xmath25 is stable when there are limited numbers of transition positions for the bilevel signal @xmath0 . in this section , we provide a necessary condition on the sampling kernel @xmath6 such that bilevel signals @xmath0 in ( [ bilevel.def ] ) are uniquely determined from their samples @xmath26 in ( [ sampling.def ] ) . also in this section , we propose an algorithm for the bilevel signal recovery . the main theorem of this note is as follows : [ uniqueness.tm ] let @xmath9 and set @xmath27 . if @xmath6 is a causal sampling kernel with @xmath28 on @xmath7 , then any bilevel causal signal @xmath0 in ( [ bilevel.def ] ) with maximal local rate of innovation @xmath1 being less than or equal to the sampling rate @xmath24 can be recovered from its samples @xmath5 . let @xmath29 then @xmath30 and @xmath31 is a strictly increasing function on @xmath32 as @xmath6 is strictly positive on @xmath7 . denote its inverse function on @xmath33 $ ] by @xmath34\longmapsto [ 0 , t]$ ] . let @xmath0 be a bilevel causal signal in ( [ bilevel.def ] ) with transition positions @xmath14 , satisfying ( [ innovationassump ] ) . then its first sample @xmath35 is given by @xmath36 where the first two equalities hold by the causality of the signal @xmath0 and the sampling kernel @xmath6 , and the fourth equality follows from ( [ bilevel.def ] ) and the observation that @xmath37 by ( [ innovationassump ] ) , ( [ innovativerate.def ] ) and the assumption that @xmath38 . recall that @xmath31 is strictly increasing on @xmath39 . then there exists a transition position in the time range @xmath32 if and only if @xmath40 . moreover , if it exists , it is given by @xmath41 thus for a bilevel causal signal , we may determine from its first sample @xmath42 the ( non-)existence of its transition position in the time period @xmath39 and further its transition position in that time period if there is one . inductively , we assume that all transition positions of the bilevel signal @xmath0 in the time range @xmath43 have been determined from its samples @xmath44 . we examine four cases to determine its transition position in the time period @xmath45 from the sample @xmath46 . * case 1 * : there is no transition position in @xmath43 . in this case , following the above argument to determine transition positions in the time range @xmath39 , we have that there exists a transition position in @xmath45 if and only if @xmath47 . if there is , the transition position is the first transition position @xmath48 of the bilevel causal signal @xmath0 , and @xmath49 * case 2 * : the last transition position in @xmath43 is @xmath50 for some @xmath51 . in this case , @xmath52 and @xmath53 for all @xmath54 . thus @xmath55 hence there exists a transition position @xmath56 in the time range @xmath45 if and only if @xmath57 is positive . moreover if @xmath58 , the transition position @xmath56 in the time range @xmath45 is determined by @xmath59 * case 3 * : the last transition position in @xmath43 is @xmath56 for some @xmath60 . in this case , the @xmath61-th sample @xmath46 is given by @xmath62 thus there exists a transition value @xmath63 if and only if @xmath64 is positive . also we see that if @xmath58 , then the transition value @xmath65 can be obtained by @xmath66 * case 4 * : the last transition position in @xmath43 is @xmath67 . in this case , all transition positions of the bilevel signal @xmath0 have been recovered already . hence the bilevel signal @xmath0 is fully recovered . this completes our inductive proof . from the above argument of theorem [ uniqueness.tm ] , we can use the following algorithm to recover a bilevel causal signal @xmath0 in ( [ bilevel.def ] ) from its samples @xmath68 , where @xmath69 : * bilevel signal recovery algorithm * : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ * if all samples @xmath70 , are zero , then set @xmath71 and stop ; else find the first nonzero sample , say @xmath72 , the first transition position of the bilevel signal @xmath0 is located at @xmath73 , and set @xmath74 . * do step 2a if the last transition position in the time range @xmath43 is @xmath50 for some @xmath51 ; do step 2b elseif the last transition position in the time range @xmath43 is @xmath56 for some @xmath60 ; and do step 4 else . * * step 2a : define @xmath56 as in ( [ ti.case2 ] ) if @xmath75 in ( [ yn+1.case2 ] ) is positive , else do step 3 . * * step 2b : define @xmath65 as in ( [ ti.case3 ] ) if @xmath75 in ( [ yn+1.case3 ] ) is positive , else do step 3 . * set @xmath76 . do step 2 if @xmath77 , and step 4 if @xmath78 . * stop as all transition positions @xmath14 , of the bilevel signal @xmath0 are recovered . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ we finish this section with a remark that the requirement @xmath38 in theorem [ uniqueness.tm ] can be relaxed to the following : there is at most one transition position @xmath14 , in each sampling range @xmath79 . in this section , we consider the maximal sampling error @xmath80 of two bilevel signals @xmath0 and @xmath81 when maximal error of their transition positions are small . we then present some numerical simulations on recovery of a bilevel signal @xmath0 in ( [ bilevel.def ] ) from its noisy samples @xmath82 in ( [ sampling.def ] ) , where @xmath83 , are bounded noises of low levels . first we notice that sampling procedure from bilevel signals @xmath0 to their samples @xmath26 are stable in bounded norm . [ stability.tm ] let @xmath9 , @xmath6 be a bounded filter supported in @xmath84 , @xmath85 be a bilevel causal signal with maximal local innovation rate @xmath86 , and @xmath87 be a perturbation of the bilevel signal @xmath0 with perturbed transition positions @xmath88 satisfying @xmath89 then the sample errors between @xmath90 and @xmath91 , are dominated by @xmath92 , i.e. , @xmath93 where @xmath94 is the @xmath95 norm of the sampling kernel @xmath6 . by the assumption on maximal local innovation rate @xmath1 of the bilevel signal @xmath0 and the maximal transition position perturbation @xmath96 between bilevel signals @xmath0 and @xmath81 , we have that @xmath97 this together with the support assumption for the sampling kernel @xmath6 gives that @xmath98 therefore @xmath99 where the first inequality holds as @xmath100 if @xmath101 and @xmath102 have nonempty intersection , the second inequality is true as @xmath103 for all @xmath104 , and the last inequality follows from the assumptions that @xmath105 and @xmath38 . this proves the sampling error estimate ( [ error.prop.eq1 ] ) between the bilevel causal signals @xmath0 and @xmath81 . now we consider the corresponding nonlinear inverse problem how to recover a bilevel signal @xmath0 from its noisy samples @xmath82 in ( [ sampling.def ] ) , where @xmath83 , are bounded noises . let us start by looking at two examples . [ example1 ] take @xmath106 as the original bilevel signal and @xmath107 as the sampling kernel . for sufficiently small @xmath108 , define @xmath109 . then for every @xmath110 , the @xmath111-th transition positions of bilevel signals @xmath112 and @xmath113 are @xmath111 and @xmath114 respectively ( hence their difference is @xmath115 that could be arbitrary large for sufficiently large @xmath111 ) , but on the other hand , maximal sampling errors for those two bilevel signals @xmath112 and @xmath113 are bounded by @xmath116 , @xmath117 this leads to instability of the recovery procedure from samples @xmath118 to the bilevel signal @xmath112 in the presence of bounded noises . [ example2 ] take @xmath112 and @xmath119 in example [ example1 ] as the original bilevel signal and the sampling kernel respectively . define @xmath120 for sufficiently small @xmath108 . then for every @xmath110 the difference between @xmath111-th transition positions of bilevel signals @xmath112 and @xmath121 is always @xmath116 , and there is no difference between their @xmath122-th samples except for @xmath123 . this suggests that the recovery procedure from samples @xmath118 to the bilevel signal @xmath112 is not locally - behaved and the reconstruction error on transition positions could disseminate . from the above two examples , we see that the nonlinear recovery procedure from samples @xmath124 to bilevel signals @xmath0 is _ unstable _ in the presence of bounded noises and that it is _ globally - behaved _ in general . in this note , we present some initial numerical simulations with small numbers of transition positions , sampling rate over maximal local rate of innovation and very low levels of noise . detailed noise performance analysis and stable recovery in the presence of other types of noises will be discussed in the coming paper . take a sampling kernel @xmath125 , and a bilevel signal @xmath126 containing 10 transition positions , see figure [ bilevelsignal.fig ] . [ cols="^,^ " , ] in this note , we show that bilevel causal signals @xmath0 could be reconstructed from their samples @xmath5 , if the sampling kernel @xmath6 is causal and positive on @xmath23 and if the sample rate is at ( or above ) the maximal local rate of innovation of the bilevel signal @xmath0 . we also propose a stable bilevel signal recovery algorithm in the presence of bounded noise if the number of transition positions of bilevel signals is not large . we remark that the bilevel signal recovery algorithm proposed in this note is applicable when uniform sampling @xmath127 every @xmath17 second replaced by nonuniform sampling @xmath128 with sampling density @xmath129 , and bilevel causal signal @xmath130 with maximal local rate of innovation @xmath131 replaced by box causal signals @xmath132 with maximal local rate of innovation @xmath133 , where for every @xmath134 , @xmath135 is height of the box located on the time period @xmath136 . the authors would like to thank professor ram mohapatra for his help in the preparation of this note . this work is supported in part by the national science foundation ( dms-1109063 ) .	bilevel signal @xmath0 with maximal local rate of innovation @xmath1 is a continuous - time signal that takes only two values @xmath2 and @xmath3 and that there is at most one transition position in any time period of @xmath4 . in this note , we introduce a recovery method for bilevel causal signals @xmath0 with maximal local rate of innovation @xmath1 from their uniform samples @xmath5 , where the sampling kernel @xmath6 is causal and positive on @xmath7 , and the sampling rate @xmath8 is at ( or above ) the maximal local rate of innovation @xmath1 . we also discuss stability of the bilevel signal recovery procedure in the presence of bounded noises .
in @xcite , rasmussen defined a homomorphism on the smooth concordance group of knots @xmath3 @xmath4 which he showed had the property that @xmath5 where we write @xmath6 for the smooth @xmath7-ball genus ( or _ slice genus _ ) of @xmath1 . the starting point for this paper is the following theorem of rasmussen s @xcite : [ raspos ] for positive knots @xmath1 ( that is , knots which admit a diagram with no negative crossings ) @xmath8 the point being that in the case of positive knots @xmath1 , the computation of @xmath2 is a triviality and agrees with twice the genus of an obvious candidate for a minimal - genus slicing surface ( namely the one obtained by pushing the seifert surface given by seifert s algorithm into the @xmath7-ball ) . the invariant @xmath2 is equivalent to all the information contained in @xmath9 , where @xmath10 is the perturbed version of standard khovanov homology first defined and studied by lee @xcite . there is a spectral sequence with @xmath11 page being the standard khovanov homology of a knot @xmath1 and @xmath12 page being the bigraded group @xmath13 and many efforts to compute @xmath14 for knots other than for positive knots have made use of the existence of spectral sequences ( for some nice examples see @xcite ) . however , since it is known that @xmath15 for @xmath16 , to define @xmath2 only requires knowledge of the partial chain complex @xmath17 where @xmath0 is a diagram of @xmath1 . in fact , since explicit representatives for a basis of @xmath9 are known at the chain level , one only needs to know the map @xmath18 for a positive diagram @xmath0 , @xmath19 . this is what made theorem [ raspos ] a trivial corollary once the properties of @xmath14 were established . by studying this map we obtain a diagram - dependent upperbound @xmath20 for @xmath2 . we also give an error estimate @xmath21 for this upperbound . the resulting lowerbound @xmath22 for @xmath2 improves upon previously known rudolph - bennequin - type inequalities . we give a list of particular cases where @xmath23 vanishes and so @xmath20 necessarily agrees with @xmath2 . the following results are stated for knots , since the rasmussen invariant is most familiar in this setting . some results however admit a generalization to links ( via the definition of @xmath14 for links as found for example in @xcite ) . we discuss this in section [ linksection ] . our results concern an easily - computable number @xmath24 which is defined from an oriented knot diagram @xmath0 . postponing an explicit description of how to compute @xmath20 until definition [ maindefn ] , we begin by giving some results . [ upperbound ] @xmath25 of course , we must remember that @xmath26 depends only on the isotopy class of the knot represented by @xmath0 , whereas the same is not true of @xmath20 . hence in order for the bound of theorem [ upperbound ] to be a good bound , we should expect to be forced to give some restrictions on diagrams @xmath0 : [ posandneg ] the bound of theorem [ upperbound ] is tight for positive diagrams @xmath0 and for negative diagrams @xmath0 . [ braid ] let @xmath27 for @xmath28 . then if @xmath29 is any word in the @xmath30 letters @xmath31 and @xmath32 is a knot diagram which is the closure of the @xmath33-stranded braid represented by @xmath29 , then we have @xmath34 we note that knots admitting such a braid presentation are known to be fibered @xcite , so in particular not every knot admits such a presentation . [ alternating ] let @xmath0 be an alternating diagram of a knot . then we have @xmath35 propositions [ posandneg ] , [ braid ] , and [ alternating ] are each consequences of theorem [ trees ] for which we need a few definitions . given a diagram @xmath0 we write @xmath36 for the oriented resolution . we form a decorated graph @xmath37 , known as the seifert graph of @xmath0 , as follows : we start with a node for each component of @xmath36 . each crossing in @xmath0 , when smoothed , lies on two distinct components of @xmath36 ; for each positive ( respectively negative ) crossing of @xmath0 we connect the corresponding nodes by an edge decorated with @xmath38 ( respectively @xmath39 ) . note that @xmath37 by itself is not enough to recover the full khovanov chain complex of the diagram @xmath0 , but if we added extra data of an ordering of the edges at each node , we would be able to recover the full complex . from @xmath37 we now form two other graphs : we form a subgraph @xmath40 ( respectively @xmath41 ) from @xmath37 by removing all edges of @xmath37 decorated with a @xmath38 ( respectively @xmath39 ) . [ maindefn ] we define the number @xmath42 where @xmath43 is the writhe of @xmath0 . [ errordefn ] we define the number @xmath44 then we have [ trees ] if @xmath45 then @xmath46 . in fact we can say more : @xmath47 theorem [ trees ] enables us to improve on previously known easily - computable combinatorial lower bounds for the slice genus . we have : [ benny ] @xmath48 which is stronger than the rudolph - bennequin inequalities as proved in @xcite , @xcite , and @xcite ( for a nice discussion see @xcite ) . of propositions [ posandneg ] , [ braid ] , and [ alternating ] . this is just a matter of checking that the condition @xmath45 of theorem [ trees ] holds in each case . this is only a non - trivial check for the case of @xmath0 being alternating . suppose @xmath0 is an alternating diagram . the complement of the oriented resolution @xmath36 is a number of regions of the plane . if @xmath0 is not the trivial diagram , there is a unique way to associate to each region either a @xmath38 or a @xmath39 such that only positive ( respectively negative ) crossings of @xmath0 occur in regions associated with a @xmath38 ( respectively @xmath39 ) and such that adjacent regions have different associated signs . see figure [ alt ] for an example . then each region with associated sign @xmath38 ( respectively @xmath39 ) corresponds to exactly one component of @xmath41 ( respectively @xmath49 ) . since there is one more region than there are circles of @xmath36 ( or equivalently nodes of @xmath37 ) we must have @xmath45 . . we indicate which crossings are positive and which negative . on the right of the figure is the oriented resolution @xmath36 on which we indicate how to uniquely associate @xmath38 or @xmath39 to each component of the complement of @xmath36.,title="fig:",width=355,height=480 ] we note that proposition [ alternating ] gives a combinatorial formula for the rasmussen invariant of an alternating diagram . it is known @xcite that the rasmussen invariant of an alternating knot agrees with the signature of the knot , and there is also known @xcite a combinatorial formula for the signature of an alternating diagram . proposition [ alternating ] gives an equivalence between these two results . there is a nice topological interpretation of @xmath50 which is useful in computing it by hand : form a graph @xmath51 which has a node for each component of @xmath49 and a node for each component of @xmath41 . each circle in @xmath36 is a member of exactly one component of @xmath49 and exactly one component of @xmath41 ; for each circle in @xmath36 let @xmath51 have an edge connecting the corresponding pair of nodes . then @xmath52 , the first betti number of @xmath51 . this follows from the connectedness of @xmath51 so that we have @xmath53 just prior to posting on the arxiv , we heard from tomomi kawamura @xcite that she has independently obtained several of the results in this paper , using entirely different methods . we thank tetsuya abe and cornelia van cott for their comments on an earlier draft of this paper . we assume familiarity with the definition of the khovanov chain complex defined from a knot diagram @xmath0 , and with rasmussen s paper @xcite . we write @xmath54 for lee s perturbed chain complex with complex coefficients ( where the tqft is induced from the frobenius algebra @xmath55/(x^2 - 1)$ ] ) , with the @xmath56 representing the quantum filtration : @xmath57 and the superscript @xmath58 denoting the homological grading : @xmath59 similarly we write @xmath60 for the homology of the chain complex @xmath61 . there is a distinguished subspace of @xmath62 which i shall write as @xmath63 ; @xmath36 being the oriented resolution of @xmath0 and @xmath64 being a shift in the quantum filtration by the writhe of @xmath0 . here one can think either of @xmath65 as being lee s tqft functor or of @xmath66 as being the perturbed khovanov homology of the ( @xmath67-crossing ) diagram @xmath36 . by lee @xcite we know that given a knot diagram @xmath0 with orientation @xmath68 , there exist @xmath69 such that @xmath70 furthermore , the homology @xmath60 is 2-dimensional and supported in homological grading @xmath71 with @xmath72 , [ { \mathfrak{s}}_{{\overline}{o } } ] > $ ] . there is an explicit description of these generators at the chain level : [ generators ] the orientation @xmath68 on @xmath0 induces an orientation on @xmath36 . for each circle @xmath73 in @xmath36 we give a invariant which is the mod 2 count of the number of circles in @xmath36 separating @xmath73 from infinity , to which we add @xmath67 ( respectively @xmath74 ) if @xmath73 has the counter - clockwise ( respectively clockwise ) orientation . we label @xmath73 with @xmath75 ( respectively @xmath76 ) if the invariant is @xmath67 ( respectively @xmath74 ) @xmath77 . this determines an element @xmath78 , @xmath79 being given in the same way using the opposite orientation @xmath80 on @xmath0 . we know that , in rasmussen s notation , @xmath81 and @xmath82 is the filtration grading of the highest filtered part of @xmath83 to contain @xmath84 $ ] ( or equivalently @xmath85 $ ] - this interchangeability is taken as understood from now on ) . this is the same as the filtration grading of the highest filtered part of @xmath86 containing @xmath84 $ ] . it follows that [ ronen ] let @xmath87 be the projection onto the vector space summand . then @xmath88 where @xmath89 is the filtration grading in @xmath90 of the highest filtered part containing @xmath91 $ ] . @xmath92 + ( of theorem [ upperbound ] ) given a knot diagram @xmath0 with orientation @xmath68 , we write @xmath93 , @xmath94 for the number of positive , negative crossings of @xmath0 respectively so that the writhe @xmath95 . form the diagram @xmath96 by taking the oriented resolution at each of the positive crossings . note that diagram @xmath96 is also oriented with writhe @xmath97 . suppose there are @xmath98 components @xmath99 of @xmath96 ( where we mean components as a subset of the plane , so that the standard 2-crossing diagram of the hopf link would be considered as a single component , for example ) and suppose that @xmath100 has @xmath101 crossings for @xmath102 . we observe that , up to quantum filtration shift by @xmath103 , the map @xmath104 can be identified with the map @xmath105 this latter map is in fact @xmath106 where @xmath107 is the @xmath108th differential in the chain complex @xmath109 and @xmath110 is the identity map . inductively on @xmath111 we observe a canonical identification @xmath112 now @xmath113 , where @xmath114 is either the element @xmath115 or @xmath116 where we use @xmath117 to stand for the induced orientation on the oriented resolution of @xmath100 . this is because the mod 2 invariant associated to each circle @xmath118 via definition [ generators ] differs by @xmath67 or @xmath74 from the invariant associated to @xmath119 via definition [ generators ] , and it is the same difference for all circles of @xmath120 . suppose the number of components of @xmath120 is @xmath121 . we observe that @xmath122 is the highest filtered part of @xmath123 to be non - zero and is @xmath74-dimensional . therefore by lemma 3.5 @xcite , @xmath124 $ ] has filtered degree less than or equal to @xmath125 in @xmath126 . we compute for @xmath89 in lemma [ ronen ] : @xmath127 hence we have @xmath128 ( of theorem [ trees ] ) given an oriented knot diagram @xmath0 , let @xmath129 be the mirror image of @xmath0 . it is then easy to check that @xmath130 so we have @xmath131 given an @xmath111-component link @xmath132 , let @xmath133 be the genus of a connected minimal - genus smooth surface in the @xmath7-ball which has @xmath134 as boundary . we extend the definition of the slice genus @xmath135 to links by defining 99999 a. beliakova and s. wehrli , categorification of the colored jones polynomial and the rasmussen invariant of links , to appear in canada j. math . t. kawamura , _ an estimate of the rasmussen invariant for links _ , preprint t. kawamura , _ the rasmussen invariants and the sharper slice - bennequin inequality on knots _ , topology 46 ( 2007 ) 29 - 38 e. s. lee _ an endomorphism of the khovanov invariant _ , adv . ( 2005 ) 554 - 586 o. plamenevskaya , _ transverse knots and khovanov homology _ , math . ( 2006 ) 571 - 586 j. rasmussen , _ khovanov homology and the slice genus _ , to appear in inventiones mathematicae a. shumakovitch , _ rasmussen invariant , slice - bennequin inequality , and sliceness of knots _ , j. knot theory ramifications ( 2007 ) , 1403 - 1412 j. stallings , _ constructions of fibered knots and links _ , algebraic and geometric topology , proc . pure math . xxxii.2 , amer . soc . , providence , ri , 1978 , pp . 55 - 60 a. stoimenow , _ some examples related to knot sliceness _ , j. pure applied algebra 210(1 ) ( 2007 ) , 161 - 175 p. traczyk , _ a combinatorial formula for the signature of alternating diagrams _ , fund . math . 184 ( 2004 ) 311 - 316	given a diagram @xmath0 of a knot @xmath1 , we give easily computable bounds for rasmussen s concordance invariant @xmath2 . the bounds are not independent of the diagram @xmath0 chosen , but we show that for diagrams satisfying a given condition the bounds are tight . as a corollary we improve on previously known bennequin - type bounds on the slice genus .
gutzwiller s work has now become a milestone in the understanding of the properties of a quantum system whose classical counterpart depicts chaotic dynamics @xcite . starting from feynman s path formulation of quantum mechanics , he has been able to complete the early studies of van vleck @xcite , deriving expressions for the semiclassical propagator , and from this , for the quantum level density : the well - known gutzwiller trace formula . the later is an asymptotic series in @xmath0 and can be separated into two parts ; the leading order corresponds to the thomas - fermi ( or extended thomas - fermi when including @xmath0 corrections ) average density of states @xcite ; the other part corresponds to the oscillations around the preceding term and involves contributions from all periodic orbits of the system . this formula has been widely used to obtain approximate values for the quantum energy eigenvalues of classically chaotic systems : the hydrogen atom in magnetic field @xcite , the helium atom @xcite , anisotropic kepler problem @xcite , resonant tunnel diode @xcite , billiards @xcite , etc . since then , the gutzwiller trace formula has also been generalized to take into account contributions of other kinds : diffractive effects @xcite , continuous families of periodic orbits@xcite , ghost orbits , etc . at the same time , because the trace formula as derived by gutzwiller only contained the leading term of the asymptotic expansion of the quantum level density , the systematic expansion of the semiclassical propagator in powers of @xmath0 has been the purpose of several studies @xcite . however , these corrections to the trace formula have only been tested for billiards , for which both classical and quantum properties are easier to calculate . in the present paper , we will show how , for quantum systems whose hamiltonian separates into kinetic and smooth potential energies , @xmath0 corrections can be computed with great accuracy , extending the method described in refs . @xcite , based on classical green s functions . especially , we will show that the previous derivation @xcite of the correction to gutzwiller trace formula is partially wrong . from a numerical point of view , all quantities involved in the calculation of the @xmath0 corrections for a given classical path can be obtained as solutions of sets of first order differential equations to be integrated along this path using standard time integrators like the runge - kutta method . the number of equations in these sets can be quite large and can be probably reduced with a deeper analysis of their structures , in the same way that the amplitude in the gutzwiller trace formula for a two - dimensional ( 2d ) system can be obtained by integrating only a @xmath1 matrix and not the whole monodromy matrix @xcite . however , it would give rise to more complicated expressions and probably to additional difficulties in the numerical implementation , whereas the expressions given in the paper can be put in the computer as they stand . also , the amount of cpu time and the memory needed by the codes are small enough , so that , on a first stage , the reduction of the number of equations can be skipped . the paper is divided as follows : in sec . [ seckqq0 ] , after a brief description of the derivation of @xmath0 corrections for the propagator @xmath2 , we explain how the classical green s functions and @xmath0 corrections can be efficiently computed . in sec . [ secgqq0 ] , we explain how to obtain expression of the additional terms , arising from the time to energy domain transformation , in @xmath0 corrections for the quantum green s function @xmath3 . in the case of the trace of the propagator , a detailed derivation of the @xmath0 corrections is carried out in sec . [ sectrk ] , leading to the proper formulas , along with the way they can be computed . the time to energy transformation is explained in sec . [ sec : tg ] , leading to the @xmath0 correction expression in the case of the quantum green s function . finally , section [ bidihy ] shows how to apply theoretical expressions obtained in the four preceding sections in the case of the 2d hydrogen in magnetic field and emphasizes the excellent agreement with numerical coefficients extracted from exact quantum calculation , using harmonic inversion @xcite . the starting point is the feynman path integral , whose discrete version , for a time independent hamiltonian which separates into kinetic and potential energies , @xmath5 , reads as follows @xcite : @xmath6,\end{gathered}\ ] ] where @xmath7 , @xmath8 and @xmath9 is the classical lagrangian . for small values of @xmath0 ( i.e. , the semiclassical limit ) , using the stationary phase approximation , all preceding integrals are expanded around the stationary solutions , that is the classical orbits @xmath10 going from @xmath11 to @xmath12 during time @xmath13 , each of them thus giving a contribution @xmath14 to the propagator , whose final expression reads formally as follows @xcite : @xmath15 where @xmath16 is the dominant semiclassical contribution to the propagator @xmath2 : @xmath17},\end{gathered}\ ] ] where @xmath18 is the classical action and @xmath19 is the morse index of the orbit . the @xmath20 expression is given by @xcite : @xmath21,\end{gathered}\ ] ] where the @xmath22 are higher - order derivatives of the potential @xmath23 , evaluated at @xmath10 . the classical green s function @xmath24 , associated with the classical orbit , is an @xmath25 matrix solution of the following equation @xcite : @xmath26 where @xmath27 is the jacobi - hill operator , controlling the linear stability around the classical orbit in the configuration space @xcite : @xmath28.\ ] ] furthermore , the fact that both initial and final point are fixed in the propagator @xmath2 imposes the following boundary conditions on the classical green s function @xcite : @xmath29.\ ] ] if @xmath30 is a conjugate point of @xmath11 , then the determinant @xmath31 in formula ( [ kl0(t ) ] ) is formally infinite , but this happens only for restricted values of @xmath13 , so that , in this section , we will focus on the general case , for which @xmath32 and @xmath11 are not conjugate points . apart from @xmath33 , @xmath24 obeying the homogeneous jacobi - hill equation @xmath34 , so that , introducing the notations @xmath35 one immediately obtains @xmath36 where @xmath37 is the @xmath38 monodromy matrix , depicting the linear stability around the classical orbit in the phase space . @xmath39 and @xmath40 are four @xmath25 matrices , whose values are determined from the boundary conditions at time @xmath33 : @xmath41 and at times @xmath42 and @xmath43 : @xmath44 for a hamiltonian which separates between kinetic and potential energy @xmath45 , @xmath37 has the following simple structure : @xmath46,\ ] ] which leads us to the following explicit expressions for the four matrices @xmath39 and @xmath40 : @xmath47 provided that @xmath48 is invertible . @xmath49 being the upper right @xmath25 submatrix of the matrix @xmath50 , gives the linear displacement of the final position for a change in the initial momentum ( the initial position being fixed to @xmath11 ) , i.e. , @xmath51 . thus , @xmath49 is the inverse matrix of @xmath52 which has been supposed to be invertible ( @xmath53 and @xmath11 are not conjugate points ) . finally the full expression for the classical green s function reads : @xmath54 & \\ & \mathrm{for } \quad 0\leq t\leq t ' , & \\ & & \\ & \left[j_2(t)-j_1(t)j_1^{-1}(t)j_2(t)\right ] & j_1^{\top}(t ' ) \\ & \mathrm{for } \quad t'\leq t\leq t. & \\ \end{array}\right.\ ] ] using the symplectic structure of @xmath55 , one can show that @xmath56 as expected because the operator @xmath27 and the boundary conditions are symmetric as it explicitly appears in the discrete version of the problem ( see ref . this is also emphasized in fig . [ figg00 ] , where the four matrix elements of a classical green s function @xmath24 ( for @xmath57 ) are plotted with respect to time @xmath58 . this example corresponds to a classical orbit of the 2d hydrogen atom in a magnetic field having initial and final points on the nucleus , namely , the closed orbit having code @xmath59 and whose trajectory in @xmath60 coordinates is also shown in the figure . ( see sec . [ bidihy ] for all details . ) as expected , the green s function vanishes at initial and final times ( i.e. , @xmath61 ) and for @xmath33 , the derivatives of each diagonal elements @xmath62 ( continuous line ) and @xmath63 ( long dashed line ) are discontinuous whereas , from property ( [ spg ] ) , the two off - diagonal elements are equal ( dotted and dashed lines ) . from eq . ( [ c1(t ) ] ) , there are three contributions to @xmath20 , namely : @xmath64 even if , in principle , one can compute @xmath24 for any @xmath65 values using eq . ( [ gttp ] ) , direct evaluation of the double integrals @xmath66 would be time consuming and numerically inefficient using standard integration routines , especially because , from its definition , @xmath24 is not a smooth function around the line @xmath33 . in what follows , we will show that the preceding integrals can be transformed in way such that their values can be obtained integrating a set of first order differential equations along the classical orbit , in the same way that , for example , the monodromy matrix @xmath55 can be computed . separating @xmath67 and @xmath68 contributions in @xmath66 , using symmetry property ( [ spg ] ) of @xmath24 and that the matrix @xmath69 is fully symmetric under index permutations , one gets , after straightforward algebra , @xmath70 in the preceding expressions the green s function @xmath24 is used only for @xmath65 values in the triangle @xmath71 and is formally written @xmath72 ( see eq . ( [ gttp ] ) ) , thus separating @xmath58 and @xmath73 contributions : @xmath74 this leads us to introduce two intermediate quantities , namely , @xmath75 and @xmath76 ( for @xmath77 , @xmath78 and @xmath79 running from @xmath80 to @xmath81 ) : @xmath82 in a way such that @xmath83 ( and @xmath84 ) are solutions of the following set of differential equations ( besides equations for @xmath85 and @xmath37 ) : @xmath86 with initial conditions @xmath87 . this set of equations , @xmath88 in total ( i.e. , 33 for a 2d system ) is easily integrated using any standard method ( fourth order runge - kutta in the present case ) . as mentioned in the introduction , the size of the preceding differential set is probably not minimal and could be reduced by a deeper analysis of the structure of these equations . however , it allows a fast and easy computation of the correction @xmath20 : * find a trajectory going from @xmath11 to @xmath12 in time @xmath13 ; * integrate the differential set for @xmath85 and @xmath37 along the trajectory to obtain the quantity @xmath89 ; * integrate the set of eqs . ( [ dsetkqq0 ] ) along the trajectory to get the three quantities @xmath90 , @xmath66 , entering in the @xmath20 expression . since the quantum green s function @xmath92 is related to the propagator @xmath2 , through a semisided fourier transform , this relation also holds between semiclassical contributions arising from each classical orbit , more precisely , @xmath93 again , a stationary phase approximation is used to perform the integral , which , for a given trajectory going from @xmath11 to @xmath12 , selects its total duration @xmath94 such that the classical motion is made at energy @xmath95 . this operation also gives rise to additional terms in @xmath0 corrections , to be summed with @xmath20 , and whose explicit expressions can be derived starting from eq . ( [ c1(t ) ] ) formally written as follows @xcite : @xmath96,\end{gathered}\ ] ] @xmath97 being thus the ( logarithm of ) usual semiclassical amplitude . then @xmath98 and @xmath97 are systematically expanded around @xmath94 : @xmath99 with @xmath100 . terms arising from @xmath101 expansion would contribute only to @xmath102 correction and can be discarded . performing the imaginary gaussian integrals leads to the additional @xmath0 corrections : @xmath103\\ -\frac{w_l^{(3)}c_0^{(1)}}{2\left(w_l^{(2)}\right)^2 } -\frac{w_l^{(4)}}{8\left(w_l^{(2)}\right)^2}+ \frac{5}{24}\frac{\left(w_l^{(3)}\right)^2}{\left(w_l^{(2)}\right)^3}.\end{gathered}\ ] ] the preceding formula is similar to the one in ref . @xcite , where the authors have expressed the coefficient @xmath104 in terms of derivatives of amplitude and action with respect to energy @xmath95 . the full expression of @xmath105 is then given by @xmath106 } \\ \times\bigl\{1+i\hbar\left[c_1(\mathbf{q},\mathbf{q}_0,t_0 ) + c_1^{t\rightarrow e}(\mathbf{q},\mathbf{q}_0,t_0)\right ] + \mathcal{o}(\hbar^2)\bigr\},\end{gathered}\ ] ] where @xmath107 is the reduced action and @xmath108 in sec . [ c1qq0 ] , we have shown that @xmath110 can be computed by integrating a set of differential equations along the classical orbit going from @xmath11 to @xmath12 in time @xmath94 . in this section we will show that it is also true for @xmath109 , which involves derivatives of both @xmath98 and @xmath111 with respect to @xmath13 . for all @xmath13 , we have the following functional relation ( @xmath11 and @xmath12 being fixed ) : @xmath112 where @xmath113 is the energy of the classical trajectory , @xmath114 , going from @xmath11 to @xmath12 in time @xmath13 , that is , the value of the hamiltonian @xmath115 taken at any point on the corresponding phase space trajectory @xmath116 . writing @xmath117 , the taylor expansion of @xmath118 is easily deduced from the taylor expansion of @xmath119 around the reference trajectory @xmath120 ( noted hereafter as @xmath121 ) : @xmath122 and from which one obtains the higher derivatives of the classical action @xmath123 at @xmath124 : @xmath125 where all derivatives of @xmath115 are evaluated at @xmath126 . equations for @xmath127 are deduced from hamilton s equations governing @xmath128 evolution : @xmath129 where again all derivatives of @xmath115 are evaluated at @xmath126 . thus , we are facing three differential sets of the form @xmath130 ( i.e , . nonhomogeneous linear differential equations ) , with the important property that the vector @xmath131 only depends on vectors @xmath132 with @xmath133 , so that they can be solved one after the other . solutions of these nonhomogeneous linear differential equations are expressed with the monodromy matrix @xmath134 : @xmath135 among the @xmath136 dimensional space of solutions given by preceding expressions , the relevant one is selected by transposing on initial values @xmath137 ( for @xmath138 ) the two boundary conditions @xmath139 introducing position @xmath140 and momentum @xmath141 parts for vectors @xmath142 , the taylor expansion of the preceding equations leads to the following boundary conditions : @xmath143 thus , the initial values @xmath144 are implicitly determined by the final values @xmath145 , through the integral expressions ( [ deltaxt ] ) , which for @xmath146 reads @xmath147 \left(\begin{array}{c } \mathbf{0 } \\ \mathbf{p}^{(1)}(0)\end{array}\right)\ ] ] showing thus that @xmath148 . then @xmath149 and @xmath150 are easily computed by integrating sets of differential equations obtained from eq . ( [ dxdt ] ) , allowing us to derive @xmath151 and @xmath152 values from eq . ( [ deltaxt ] ) , solving systems similar to eq . ( [ xtx0 ] ) : @xmath153 where we have introduced the notation @xmath154 for vectors @xmath155 . quantities like @xmath156 , @xmath157 , and @xmath158 can also be expressed in terms of @xmath159 and its derivatives . at this point , from the values of the three vectors @xmath160 and using eqs . ( [ dwdt ] ) at time @xmath94 , all derivatives @xmath161 of the classical action can be computed . we now explain how to compute derivatives of @xmath111 . more precisely one has to calculate the two coefficients @xmath162 and @xmath163 , which are derivatives of @xmath164 , so that , using the well - known formula @xmath165 ( @xmath166 being any ( invertible ) matrix ) , expressions of @xmath162 and @xmath163 become @xmath167 where @xmath168 means derivative of @xmath169 when changing total time @xmath13 ( and thus the classical orbit ) , which must not be confused with @xmath170 ( time derivative of @xmath171 along a given classical orbit ) . @xmath49 being the @xmath25 upper right submatrix of the monodromy matrix @xmath55 , @xmath172 is also stored at the same position in matrix @xmath173 , for which we will derive general expressions . for this purpose , we first introduce the explicit notation @xmath174 , representing the value of the monodromy matrix at time @xmath58 along the orbit going from @xmath11 to @xmath12 in time @xmath13 . writing @xmath117 , the taylor expansion of @xmath174 for a given time @xmath58 reads @xmath175 where @xmath176 is the monodromy matrix along the reference orbit ( i.e. , going from @xmath11 to @xmath12 in time @xmath94 ) . then @xmath177 and @xmath178 are the taylor coefficients of monodromy matrix @xmath179 and thus have the following expression : @xmath180 equations governing @xmath181 evolution are easily deduced from the one for @xmath174 : @xmath182,\\ \dot{m}^{(2)}_{ij}&=\sigma_{ik}\left[h^{(2)}_{kl}m^{(2)}_{lj } + 2h^{(3)}_{klm } x^{(1)}_m m^{(1)}_{lj}\right.\\ & \phantom{=\sigma_{ik}\left[\right . } \left.+h^{(3)}_{klm } x^{(2)}_m m^{(0)}_{lj } + h^{(4)}_{klmn } x^{(1)}_m x^{(1)}_n m^{(0)}_{lj}\right ] \end{aligned}\right.\ ] ] with initial conditions @xmath183 . obviously these equations are similar to those governing @xmath142 evolution , so that @xmath184 and @xmath185 values will be obtained by integrating similar differential sets . actually , it can be shown that all these sets ( for both @xmath142 and @xmath186 ) can be concatened in only one ( larger ) set of differential equations , whose integration can be done at once . finally , gathering all quantities in eq . ( [ dmdt ] ) , the two matrices @xmath168 and @xmath187 are inserted in eq . ( [ dc0dt ] ) thus giving values for @xmath162 and @xmath163 , which , along with the values for @xmath123 , allow us to compute the numerical value for @xmath109 . obviously , the number of equations in the preceding differential sets can be reduced , especially for hamiltonian separating into kinetic and potential energy , for which @xmath188 and @xmath189 coefficients are nonvanishing only when @xmath190 . however , these sets are straightforward to implement and need only a small amount of cpu time to be solved using any conventional integrator ( @xmath191 order runge - kutta in the present case ) . the diagonal elements @xmath193 of the propagator are related to classical orbits starting from @xmath11 and returning to this point after time @xmath13 , i.e. , closed orbits . summing all these diagonal elements , that is performing the integral @xmath194 , will select , through another stationary phase approximation , closed orbits for which initial and final momentum are equal : periodic orbits . @xmath0 corrections to leading order of the semiclassical contribution to @xmath192 from each periodic orbit can be derived following the same scheme , previously used for the propagator itself @xcite . adding the integral over the initial and final positions in eq . ( [ fpi2 ] ) yields @xcite @xmath195 \end{split}\ ] ] with @xmath196 . the stationary phase approximation around a given periodic orbit @xmath10 is made explicit when replacing the preceding @xmath197 integral with @xcite @xmath198 with @xmath199 . for @xmath200 ( i.e. , initial position ) , only deviations perpendicular to the periodic orbit @xmath201 have been introduced because the classical action @xmath202 is constant along the orbit ( depicted by @xmath203 ) . the contribution @xmath204 of this periodic orbit to @xmath192 then reads @xcite @xmath205 , \end{aligned}\ ] ] where @xmath206 when @xmath207 and @xmath208 when @xmath209 . @xmath210 is , in the large @xmath211 limit , the classical action of the periodic orbit . full expressions for @xmath212 , @xmath213 , and @xmath214 can be found in ref . @xcite . then , the next step would consist of performing all imaginary gaussian integrals , leaving out the integral along the orbit . however , in the preceding coordinate transformation ( [ coord ] ) , there is an hidden subtlety , affecting only @xmath0 corrections , which explains why probably it is not mentioned in usual textbooks @xcite , where authors are only looking at leading semiclassical amplitudes . actually , the problem is that the integral over @xmath203 corresponds to the length of the classical orbit , only when @xmath215 ; for a nonzero value , it will correspond to integration on a closed curve , slightly displaced from the original trajectory , whose length will thus depend on the @xmath201 value . to enlighten this , let us suppose that we have a bidimensional system , for which one periodic orbit is a circle of radius @xmath216 , traveled at constant speed @xmath217 . the coordinate transformation is then easily made using polar coordinates @xmath218 : @xmath219 the negative sign appears to preserve orientation . the volume element @xmath220 becomes @xmath221 which shows that , in this case , @xmath222 is not simply @xmath223 , the length on the periodic orbit , but is given by @xmath224 this simple example shows actually that the variable @xmath203 is not independent of @xmath201 , whereas @xmath225 is . for a general system , the variable that can play the @xmath225 role is actually the time @xmath58 , whose variation domain @xmath226 $ ] is fixed and then obviously independent of @xmath201 . thus one has to generalize the relation @xmath227 , valid only on the periodic orbit . this is done by writing explicitly the coordinate transformation @xmath228 : @xmath229 where @xmath230 are @xmath231 orthogonal unit vectors lying in the plane perpendicular to the periodic orbit at time @xmath232 . the jacobian of the transformation reads @xmath233}\\ & = \|\dot{\mathbf{q}}^{\mathrm{cl}}\|-\frac{1}{\|\dot{\mathbf{q}}^{\mathrm{cl}}\| } \bm{\xi}_0^{\perp}\cdot\ddot{\mathbf{q}}^{\mathrm{cl}}. \end{aligned}\ ] ] inserting the volume element in eq . ( [ etfpi ] ) , the contribution @xmath204 of the periodic orbit now reads , keeping only terms giving rise to @xmath0 corrections , @xmath234,\end{gathered}\ ] ] where we have seen that @xmath235 and we have introduced the index @xmath236 for @xmath237 . as explained in ref . @xcite , the imaginary gaussian integrals can be expressed in terms of another classical green s functions @xmath24 , whose boundary conditions are extracted when comparing the detailed expression of @xmath212 with the discrete version of the jacobi - hill operator @xmath27 , see eq . ( [ jho ] ) . especially , it can be shown that , in the large @xmath211 limit , they become @xmath238,\ ] ] where we have introduced @xmath239 the projector along the periodic orbit at time @xmath232 and @xmath240 . in ref . @xcite , only the @xmath241 boundary conditions corresponding to the first two lines were given , whereas the @xmath242 ones corresponding to the last line were missing . gathering all results and taking the large @xmath211 limit in eq . ( [ etfpi3 ] ) , the contribution of the given periodic orbit to the trace of the propagator reads as follows : @xmath243 @xmath244 being the usual semiclassical leading order @xcite @xmath245},\end{gathered}\ ] ] where @xmath246 is the classical action of the periodic orbit and @xmath247 its maslov index . the first @xmath0 correction @xmath101 to @xmath244 is then obtained by averaging over the time @xmath232 ( i.e. over the full periodic orbit ) the coefficient @xmath248 , given by @xmath249,\end{gathered}\ ] ] where @xmath232 represents thus the position @xmath11 on the periodic orbit at which boundary conditions ( [ bcp ] ) on the classical green s function @xmath24 are applied . @xmath11 is also the initial ( and final ) position on the periodic orbit for classical motions corresponding to times @xmath58 and @xmath73 entered in the preceding expression . as in sec . [ cgfkqq0 ] , where expressions for classical green s functions for the propagator @xmath2 where derived , we introduce the @xmath250 notations and @xmath251 , @xmath252 matrices . using all boundary conditions ( at times @xmath33 , @xmath42 and @xmath43 ) gives rise to the following equation : @xmath253\left(m(t)-\openone_{2f}\right ) \left(\begin{array}{c } a_-(t ' ) \\ b_-(t ' ) \end{array}\right)=\\ \left[\begin{array}{cc } \openone & 0 \\ 0 & \mathcal{q}_{t_0 } \end{array}\right]m(t ) \left(\begin{array}{r } -j^{\top}_1(t')\\ j^{\top}_2(t ' ) \end{array}\right).\end{gathered}\ ] ] the preceding set of linear equations , formally written @xmath254 , can not be solved directly because the @xmath38 matrix @xmath255 is obviously singular . more precisely , existence and number of solutions for the system @xmath256 are determined by the two following properties : * solutions exists if for all vectors @xmath257 such that @xmath258 , then @xmath259 . * if the preceding condition is fulfilled , and if @xmath260 is a solution , then for all vectors @xmath261 such that @xmath262 , @xmath263 is also a solution , showing that the dimension of the solution space is that of the nullspace of @xmath255 . in the present case , equation @xmath264 leads to , either @xmath265\mathbf{y}=\mathbf{0 } \qquad \rightarrow\qquad \mathbf{y}\propto \left(\begin{array}{c } \mathbf{0 } \\ \dot{\mathbf{q}}(t_0 ) \end{array}\right)\ ] ] or @xmath266\mathbf{y}\neq\mathbf{0}.\ ] ] for a generic unstable periodic orbit , the eigenspace associated with the eigenvalue @xmath80 of @xmath55 ( t being the period ) , is of dimension one and is spanned by the vector parallel to the flow ( see the appendix ) @xmath267 , so that , in the second case , one gets @xmath268 and @xmath257 is a solution of : @xmath265\mathbf{y}\propto \left(\begin{array}{r } -\dot{\mathbf{p}}(t_0 ) \\ \dot{\mathbf{q}}(t_0 ) \end{array}\right),\ ] ] which is impossible unless @xmath269 , which , for hamiltonian separating into kinetic and potential energies , corresponds to a self - retracing periodic orbit , for which a slightly modified approach should be developed @xcite . nevertheless , this case is peculiar , and we will suppose in the rest of the section that @xmath270 never vanishes along the periodic orbit in consideration . thus , the nullspace of @xmath271 being one - dimensional and spanned by the vector @xmath272 , eq . ( [ lsab ] ) immediately shows that for any column of matrix @xmath273 , we get @xmath274 , thus fulfilling the first condition . denoting @xmath275 as a solution of eq . ( [ lsab ] ) , which can be easily obtained using singular value decomposition ( svd ) of matrix @xmath255 , and the nullspace of @xmath276 being spanned by @xmath277 , the general solution of eq . ( [ lsab ] ) reads @xmath278,\ ] ] where @xmath279 are unknown real parameters still to be determined . actually , in eq . ( [ lsab ] ) one boundary condition has not been taken into account , namely that @xmath280 which , using that the projector @xmath239 reads @xmath281 allows us to get @xmath279 values and , from that , the final expression @xmath282\mathcal{x}_0,\ ] ] which , of course , is now independent of the particular solution @xmath275 . whereas in the case of the propagator @xmath2 , for which we were able to give an explicit expression ( [ gttp ] ) , the classical green s function associated with the trace of the propagator @xmath192 is only defined trough a linear system ( [ lsab ] ) , which nevertheless allows us to obtain its numerical value for any @xmath65 . although it clearly appears that matrix @xmath283 expression ( see ref . @xcite ) is symmetric , meaning that the classical green s function must fulfill the property @xmath284 , getting the later directly from eq . ( [ lsab ] ) is not obvious . however , in the case of the 2d hydrogen in a magnetic field ( see sec . [ bidihy ] for all details ) , we have numerically checked that the property holds . for example , in fig . [ figtrg ] the four coefficients of classical green s function @xmath24 ( for @xmath285 ) of the periodic orbit @xmath286 are plotted with respect to time @xmath58 . the starting point @xmath232 on the periodic orbit is depicted by the cross . actually , we have plotted the coefficient of the rotated matrix @xmath287 , such that its first row corresponds to the direction parallel to the orbit ; @xmath288 ( continuous line ) and @xmath289 ( dotted line ) are thus equal to zero for initial ( @xmath42 ) and final ( @xmath43 ) points . the other boundary conditions can also be verified in the figure : the dashed line ( @xmath290 ) ( respectively the long dashed line , @xmath291 ) has not only the same value at initial and final time , but also the same slope , which means that @xmath290 ( respectively @xmath291 ) and its time derivative fulfill the periodic boundary conditions ( [ bcp ] ) . finally , for @xmath33 , the off - diagonal coefficients @xmath292 ( dotted line ) and @xmath293 ( dashed line ) are equal , as expected from the symmetry property . as seen previously ( see sec . [ c1qq0 ] ) , we will explain how the numerical value of coefficients @xmath248 can be obtained by integrating a set of differential equation , using the standard runge - kutta method . there are now four contributions to @xmath248 , namely @xmath294 the two main difficulties now are that @xmath24 does not factorize anymore in a product of matrix at time @xmath58 and a matrix at time @xmath73 , nor does the symmetric property @xmath284 explicitly appear ( even if we have numerically checked that it is fulfilled ) . nevertheless , as seen previously , separating @xmath295 and @xmath296 contributions in @xmath83 expressions and introducing four quantities @xmath297 for @xmath298 , allows us to compute @xmath299 , by integrating the following set of differential equations from @xmath42 to @xmath13 ( besides equations for @xmath85 and @xmath37 ) : @xmath300 with vanishing initial conditions for @xmath297 and @xmath299 . for each time step , one must compute matrices @xmath301 and @xmath302 ( and from there matrices @xmath303 and @xmath304 ) , solving the linear system described in the previous section , using singular value decomposition of matrix @xmath255 , which , being independent of @xmath58 , is done before starting the runge - kutta integration . skipping intermediate steps , the differential equations leading to @xmath305 computation reads as follow , introducing another eight quantities @xmath306 : @xmath307 with vanishing initial conditions for @xmath306 and @xmath308 finally , one must add equations leading to @xmath309 and @xmath90 computation , namely @xmath310 where we have used @xmath311 . taking into account equations for @xmath85 and @xmath37 , this gives rise to a total of @xmath312 equations , that is @xmath313 for a 2d system . in practice , having found a periodic orbit and for a given @xmath232 along this orbit , the coefficient @xmath248 is computed in two steps : * achieve the svd decomposition of the matrix @xmath255 , appearing on the left - hand side of eq . ( [ lsab ] ) , and compute the projector matrix appearing on the right - hand side of eq . ( [ mproj ] ) . * integrate the differential set ( [ dsettrk ] ) along the periodic orbit ( starting at point depicted by @xmath232 ) . at any time @xmath58 , use the preceding svd decomposition to obtain a solution @xmath275 and the projector matrix to get the true solution @xmath314 and thus @xmath315 , using eq . ( [ mproj ] ) . finally , the coefficient @xmath248 , being a smooth function of @xmath232 , the average over time @xmath232 , leading to the @xmath0 correction term @xmath101 , can be handled by any conventional integrator . steps leading to the semiclassical contribution @xmath317 from a given periodic orbit to the trace of the green s function @xmath316 are identical to those giving the @xmath105 expression , so that @xmath317 reads @xmath318 } \\ \times\bigl\{1+i\hbar\left[c_1(t_0)+c_1^{t\rightarrow e}(t_0)\right ] + \mathcal{o}(\hbar^2)\bigr\},\end{gathered}\ ] ] where @xmath319 is given by @xmath320\\ -\frac{w_l^{(3)}c_0^{(1)}}{2\left(w_l^{(2)}\right)^2 } -\frac{w_l^{(4)}}{8\left(w_l^{(2)}\right)^2}+ \frac{5}{24}\frac{\left(w_l^{(3)}\right)^2}{\left(w_l^{(2)}\right)^3}.\end{gathered}\ ] ] @xmath321 ( respectively @xmath322 ) are the taylor coefficients of the @xmath246 ( resppectively @xmath323 ) expansion around @xmath94 . computation of @xmath321 is much the same as in the green s function case , because the functional relation @xmath324 still holds for a given periodic orbit , @xmath325 being its energy as function of its period , which is still given by the value of the hamiltonian @xmath115 taken at any point on the corresponding phase space trajectory @xmath116 . thus , the taylor expansion of @xmath128 around the periodic orbit @xmath120 , will lead to the same expressions for @xmath321 coefficients ( eq . [ dwdt ] ) and for @xmath127 equations ( eq . [ dxdt ] ) . the only differences with the preceding section arise from the boundary conditions fulfilled by @xmath127 , deduced from the equation @xmath326 , i.e. , @xmath128 is a periodic orbit of period @xmath13 . the taylor expansion of this relation leads to the following conditions : @xmath327 solutions of the differential set ( [ dxdt ] ) still have the following formal expressions ( [ deltaxt ] ) , which , inserted in the boundary conditions ( [ bcxp ] ) , leads to equations on @xmath137 only @xmath328 the matrix @xmath329 being singular , solving the preceding linear equations need additional discussion , which , for simplicity , will focus on @xmath330 only . first , the nullspace of is spanned by @xmath331 , which is obviously orthogonal to @xmath332 , the right hand - side of the equation for @xmath330 , thus showing that this equation admits solutions . then , the nullspace of being spanned by @xmath332 , the whole set of solutions reads @xmath333 where @xmath334 is a particular solution of the equation . actually , the term @xmath335 corresponds to an displacement of the initial conditions along the flow , which , of course , gives back the same periodic orbit ( at first order in @xmath336 ) . we thus expect that this term has a vanishing contribution to @xmath337 , which is easily verified when inserting the general solution in the @xmath337 expression ( taken at time @xmath338 ) : @xmath339 because of the hamilton s equations @xmath340 . these two properties also hold in the cases of @xmath341 and @xmath342 , but are slightly more complicated to establish because the right - hand sides of the equations involve @xmath343 and derivatives of @xmath160 . thus , integrating the same differential sets that were used for @xmath92 , one is able to compute the first four derivatives of the action , @xmath321 , with respect to the period . starting from the @xmath344 expression @xmath345 and using the fact that @xmath346 , one obtains @xmath347 which means that one is left with the calculation of derivatives of @xmath348 with respect to the period @xmath13 . as shown in the appendix , @xmath349 is given by the determinant of the @xmath350 matrix @xmath351 defined as follows : @xmath352 where we have introduced @xmath353 ( respectively @xmath354 ) the projector on the direction parallel to the flow ( respectively perpendicular to the energy shell ) , more precisely , the @xmath353 and @xmath354 expressions are @xmath355 where @xmath356 is the unit vector tangent to the flow at initial ( and thus final ) time and @xmath357 . now , using again formula ( [ dlndet ] ) , derivatives of @xmath349 with respect to the period read @xmath358 with @xmath359 as seen previously ( sec . [ c1te ] ) , @xmath360 and @xmath361 are expressed in terms of the coefficients @xmath181 of the taylor expansion of the monodromy matrix @xmath174 ( associated with the periodic orbit @xmath128 of period @xmath13 ) around the periodic orbit @xmath126 of period @xmath94 , see eq . ( [ dmdt ] ) . inserting the taylor expansion of @xmath362 around @xmath94 in the @xmath353 expression , namely , @xmath363 one obtains the derivatives of @xmath353 with respect to @xmath13 : @xmath364 where all @xmath365 are evaluated at time @xmath42 . gathering the preceding expressions into eq . ( [ dndt ] ) allows us to compute @xmath366 derivatives , which , inserted together with derivatives of the action , in eq . ( [ dc0dttr ] ) gives the numerical values for @xmath367 and @xmath368 , which finally leads to the additional @xmath0 correction @xmath369 . the hydrogen atom is one example of quantum system whose classical counterpart depicts a chaotic behavior and has been widely studied ( see for , e.g. , ref . @xcite for a complete review ) . it has now become a very useful tool for testing new ideas and tools in the quantum chaos area , both on the semiclassical @xcite or universality @xcite points of view , especially because computing very highly excited states has become a standard task on a regular workstation , allowing the semiclassical regime to be reached easily . even if one would have preferred to work with the real hydrogen atom ( i.e. , the three - dimensional one ) , in this paper we will focus on the two dimensional hydrogen atom in a magnetic field , because taking into account invariance by rotation around the magnetic field , gives rise to centrifugal terms in the hamiltonian ( typically @xmath370 ) which would also contribute to @xmath0 corrections and would need a study on its own . one must also notice that , even if the classical dynamics are identical for both cases , the fact that the magnetic field axis is no longer a rotation axis in the 2d case gives rise to slight modifications in the maslov indices @xcite . in atomic units the hamiltonian of the 2d hydrogen in magnetic field reads @xmath371 where @xmath372 , with @xmath373 . the classical counterpart of this hamiltonian has a scaling property , that is , if we define new variables by @xmath374 we obtain a new hamiltonian @xmath375 given by @xmath376 which does not depend on @xmath377 anymore . the classical dynamics of this hamiltonian is entirely fixed by the scaled energy @xmath378 given by : @xmath379 all properties of the classical trajectories of the original hamiltonian can be deduced from the scaled dynamics using the scaling transformation . from the quantum point of view , this scaling introduces an effective @xmath0 value , which is easily seen on the scaled schrdinger equation , @xmath380 , for a fixed scaled energy @xmath378 : @xmath381\psi=\epsilon\psi.\ ] ] thus , the effective @xmath0 is given by @xmath382 and so at a fixed value of the scaled energy @xmath378 , the semiclassical limit is obtained when @xmath377 tends to 0 . the singularity in the classical equations of motion due to the divergence of the coulomb potential at @xmath383 is regularized using the semiparabolic coordinates @xmath384 , giving rise to the following effective classical hamiltonian @xcite : @xmath385 the trajectories corresponding to the original problem are obtained when fixing total energy @xmath386 . the associated quantum hamiltonian reads @xmath387 which separates into kinetic and potential energy , so that the semiclassical formula derived in the preceding sections applied to the associated quantum green s function @xmath388 , the hydrogen in a magnetic field being recovered for @xmath389 ( actually @xmath390 corresponds to the nucleus charge ) @xmath391 where @xmath392 is an ( normalized ) eigenvector of @xmath393 for the eigenenergy @xmath394 , @xmath395 representing the set of quantum labels , i.e. , level number and symmetry properties ( see below ) , describing @xmath392 . the matrix element @xmath396 , where @xmath397 then reads @xmath398 where @xmath399 has been supposed to be real , with @xmath393 being invariant under @xmath400 . taking @xmath401 on the real axis , the imaginary part of @xmath396 , becomes @xmath402 to which any classical path going from @xmath12 to @xmath11 at energy @xmath403 , gives the following contribution ( see eq . ( [ gqqofin ] ) ) : @xmath404 provided it is far enough from any bifurcation and that @xmath12 and @xmath11 are not conjugate points for this trajectory . amplitudes and phases being defined by @xmath405 neglecting @xmath0 corrections in eq . ( [ imgqq0sc ] ) , the fourier transform with respect to the variable @xmath406 of the following function : @xmath407 will depict peaks at the classical actions @xmath408 , with complex amplitude @xmath409 , which has been extensively used to compare the exact quantum green s function with its semiclassical estimation at the leading order in @xmath0 . in the same way , the fourier transform of the following function : @xmath410 will also depict peaks at the classical actions @xmath408 , whose complex amplitude , given by @xmath411 allows us to extract the numerical value of the @xmath0 correction @xmath412 . the energy @xmath403 being fixed , the @xmath413 function selects the values @xmath414 of @xmath415 for which @xmath403 is an eigenvalue , transforming eqs . ( [ scaled0 ] ) and ( [ scaled1 ] ) into @xmath416 stepping to the case of the trace of the green s function , the preceding relations ( [ imgqq0 ] ) and ( [ imgqq0sc ] ) becomes @xmath417 and , see eq . ( [ trgreen ] ) : @xmath418 where @xmath419 is the action of the periodic orbit and @xmath420 so that the classical quantities @xmath419 , @xmath421 , and the @xmath0 correction @xmath422 can be obtained by taking the fourier transform of the following expressions with respect to the variable @xmath415 : @xmath423 focusing on the @xmath424 value , the 2d hydrogen in a magnetic field case , one has to find effective @xmath0 values for which @xmath425 is an eigenvalue of the schrdinger equation @xmath426 , which is conveniently written as follows : @xmath427\psi(u , v)\\ = \hbar^2\left[-\frac{1}{2}\left(\frac{\partial^2}{\partial u^2}+ \frac{\partial^2}{\partial v^2}\right)\right]\psi(u , v)\end{gathered}\ ] ] such that @xmath428 appears to be a solution of a generalized eigenvalue problem @xmath429 , with @xmath430 the preceding operators @xmath431 , @xmath432 , and thus @xmath393 are invariant under all transformations belonging to the symmetry group @xmath433 , leading to four nondegenerate series of energy levels , labeled eee , eeo , ooe and ooo according to ref . @xcite and a twofold degenerate series eo and oe , where e means even and o means odd , the first two letters referring to the @xmath434 and @xmath435 symmetries , the third letter to @xmath436 . actually , because of the definition of the semiparabolic coordinates @xmath60 , only eigenvectors invariant under the parity symmetry @xmath437 correspond to eigenvectors of the 2d hydrogen in a magnetic field , allowing us , in principle , to drop the oe and eo series @xcite . however , from the semiclassical point of view , one would have to extend all preceding sections to symmetry - projected propagator and green s function @xcite , and thus to take into account symmetry properties of the classical green s function , which is beyond the scope of this paper . for this reason , we also include the oe and eo series in the remainder of this paper . finally , eigenvalues and eigenvectors are obtained by solving the matrix representation of the generalized eigenvalue problem @xmath429 in sturmian bases ( one for each symmetry class ) @xcite , using the lanczos algorithm . typically , we have computed effective @xmath0 values ranging from @xmath438 to @xmath439 , which for scaled energy @xmath440 corresponds to roughly @xmath441 eigenvalues in total . one must notice that the generalized eigenvectors @xmath442 , for a fixed @xmath0 value , are actually orthogonal for the scalar product defined by operator @xmath443 : latexmath:[\[\langle\widetilde{\tau,\hbar}\|\frac{\mathbf{p}^2}{2 } @xmath442 and @xmath392 relations read @xmath445 giving rise to @xmath446 ( [ scaled3 ] ) and @xmath447 ( [ scaledtr ] ) expressions in terms of the computed eigenvectors : @xmath448 as explained previously , the fourier transform of the two functions @xmath449 and @xmath450 will depicts peaks at classical actions and @xmath0 corrections are obtained from the amplitude of these peaks . however , in the case of signal given by @xmath451 , it is now well known that the harmonic inversion method is very well suited and is much more powerful than the conventional fourier transform to extract unknown frequencies @xmath452 and amplitudes @xmath453 @xcite . in our case the signals are the two functions @xmath454 and @xmath455 , which are of the form @xmath456 besides contributions from all other types of orbits ( ghost , continuous family , etc . ) . orbits having initial and final points at the nucleus ( i.e. , @xmath458 ) are of special interest because they are involved in semiclassical estimation of the photoionization cross - section @xcite , which can be directly compared to experimental results @xcite . even if the full @xmath0 expansion of the cross - section does not reduce to @xmath459 contributions , all closed orbits are well known and classified , so that this case remains a nice example of @xmath0 corrections for @xmath457 . .[tabclosed]classical properties of closed orbits involved in the semiclassical expansion of the quantum green s function @xmath457 of the 2d hydrogen atom in a magnetic field , for the case @xmath458 . because each closed orbit corresponds either to a half - periodic orbit or a periodic orbit , we have labeled them with the four - disk code of the corresponding periodic orbit @xcite . their trajectories in the @xmath60 plane are shown in fig . [ g00tf ] . @xmath460 is the reduced action ( i.e. , @xmath461 ) , @xmath462 is the period , @xmath463 is the leading semiclassical amplitude , @xmath464 is the maslov index , @xmath465 is the first order @xmath0 correction , given by the sum @xmath466 , see eq . ( [ cpgqq0 ] ) . [ cols="<,^,^,^,^ " , ] in summary , we have explained in this paper how to effectively compute @xmath0 corrections in the semiclassical expansions of the propagator @xmath2 , its trace @xmath192 , the quantum green s function @xmath92 and its trace @xmath316 for chaotic systems with smooth potential . the method is based on the classical green s functions associated to the relevant trajectories , that is either going from @xmath12 to @xmath11 in the propagator case or periodic orbits for @xmath192 , together with adapted boundary conditions . we have shown how all quantities can be obtained by integrating , using the standard runge - kutta method , sets of differential equations . we have also shown that in the derivation of the semiclassical expansion for @xmath192 ( and thus @xmath316 ) , starting from the feynman path integral , one must take into account additional terms , which affect only @xmath0 correction coefficients . this is emphasized by the excellent agreement observed when comparing , in the case of the 2d hydrogen atom in a magnetic field , our theoretical results with the numerical coefficients extracted from exact quantum data , using the harmonic inversion . obviously , there are still many points to be developed . besides the few cases , such as self - retracing orbits or continuous families of orbits , needing specific extensions , it would be very interesting to understand how to include continuous and discrete symmetries . also , going into the extended phase space @xmath467 @xcite , it would be possible to get a better understanding of similarities observed between the differential sets leading , on one side to the @xmath0 corrections for the propagator and its trace and , on the other side to the additional terms arising in the @xmath0 corrections for the quantum green s function and its trace . the author thanks d. delande for fruitful discussions and for his kind support during this work , especially for numerous suggestions which led to the present form . the author also thanks m. ku for useful discussions . laboratoire kastler brossel is laboratoire de luniversit pierre et marie curie et de lecole normale suprieure , unit mixte de recherche 8552 du cnrs . in this appendix , we consider an isolated unstable periodic orbit of period @xmath13 . we shall use the notations @xmath356 and @xmath468 for the units vectors which are , respectively , parallel to the flow and perpendicular to the energy shell at the initial point . from hamilton s equations , we have that @xmath469 , i.e. , @xmath356 is an eigenvector of the matrix @xmath55 for the eigenvalue 1 . the symplectic equation fulfilled by @xmath55 , namely , @xmath470 , implies that , if @xmath471 and @xmath472 are two eigenvectors for the eigenvalues @xmath473 and @xmath474 , we have the following properties : @xmath475 showing thus that @xmath476 is an eigenvalue of @xmath477 and , from that , of @xmath55 . in addition , @xmath55 being a real matrix , @xmath478 and @xmath479 are also eigenvalues of @xmath55 , so that the nontrivial eigenvalues ( i.e. , @xmath480 ) either fall in the @xmath481 pair or in quadruplet @xmath482 . in the case of @xmath483 , the two preceding equations ( [ emt ] ) imply that @xmath468 is an eigenvector of @xmath477 ( but not necessarily of @xmath55 ) for the eigenvalue 1 and that for every @xmath484 , @xmath472 is orthogonal to @xmath468 . in the basis @xmath485 , @xmath55 entries then read @xmath486,\ ] ] where we have supposed that all eigenvalues are simple . for degenerated eigenvalues , @xmath55 would be block diagonal . for a generic periodic orbit , @xmath487 and @xmath279 are nonvanishing emphasizing thus that @xmath468 is not an eigenvector of @xmath55 . introducing the vector @xmath488 defined as follows : @xmath489 one immediately gets that @xmath490 in the case @xmath491 , we have thus found another eigenvector for the eigenvalue 1 , which means that a small displacement of initial conditions in the @xmath488 direction leads to another periodic motion with the same period @xmath13 , and thus that the periodic orbit is actually embedded in a continuous family . indeed , using notations from sec . [ sec : tg ] , one can show that @xmath492 so that we have @xmath493 in sec . [ sec : tg ] , one needs to compute derivatives with respect to the period @xmath13 of @xmath349 , whose expression in terms of the nontrivial eigenvalues of the monodromy matrix reads @xmath494 introducing @xmath495 and @xmath496 the projectors on the directions @xmath356 and @xmath468 , more precisely @xmath497 ones defines the matrix @xmath351 as follows : @xmath498 in the basis @xmath485 , using orthogonality between @xmath468 and @xmath472 , entries of @xmath351 read @xmath499,\ ] ] where @xmath500 , which actually could be related to the @xmath501 , but this is not necessary in our case . this shows that the determinant of @xmath351 is exactly @xmath502 . the main advantage of the matrix @xmath351 is that its expression ( [ nt ] ) does not involve the eigenvectors or the eigenvalues of @xmath55 , so that its determinant can be directly computed , without the diagonalization stage required when getting @xmath349 through the eigenvalues @xmath474 . furthermore , derivatives of @xmath503 with respect to the period @xmath13 are also straightforward to obtain , knowing derivatives of @xmath55 and of @xmath362 , whereas derivatives of @xmath474 would require the knowledge of those of the eigenvectors @xmath472 .	the validity of semiclassical expansions in the power of @xmath0 for the quantum green s function have been extensively tested for billiards systems , but in the case of chaotic dynamics with smooth potential , even if formula are existing , a quantitative comparison is still missing . in this paper , extending the theory developed by gaspard et _ al . _ , adv . chem . phys . * xc * 105 ( 1995 ) , based on the classical green s functions , we present an efficient method allowing the calculation of @xmath0 corrections for the propagator , the quantum green s function , and their traces . especially , we show that the previously published expressions for @xmath0 corrections to the traces are incomplete .
the authors thank cris moore for helpful discussions , as well as sidney redner and antoine allard for comments on the manuscript . this work has been supported by the santa fe institute , by the university of vermont , by the james s. mcdonnell foundation postdoctoral fellowship ( lhd ) , by grant dmr-1623243 from the national science foundation ( ub ) , and grant no . 2012145 from the united states israel binational science foundation ( ub ) .	the giant @xmath0-core maximal connected subgraph of a network where each node has at least @xmath0 neighbors is important in the study of phase transitions and in applications of network theory . unlike erds - rnyi graphs and other random networks where @xmath0-cores emerge discontinuously for @xmath1 , we show that transitive linking ( or triadic closure ) leads to 3-cores emerging through single or double phase transitions of both discontinuous and continuous nature . we also develop a @xmath0-core calculation that includes clustering and provides insights into how high - level connectivity emerges . the emergence of high - level or large - scale connectivity patterns on a network influences its macroscopic behaviour . it is at times essential , as it is for traffic on transportation networks or communication on the internet , and at times detrimental , such as epidemics on social networks or cascades of failure on technological networks . one way or the other , identifying _ when _ and _ how _ strong connectivity appears or disappears is critical . this explains the explosion of models to produce phase transitions of rich and diverse nature on networks @xcite . we here discuss how single or double phase transitions of both discontinuous and continuous nature can occur solely due to the structure of a network , without ad hoc rules or complicated models . to do so , we focus on random networks with clustering and investigate the emergence of @xmath0-cores . giant @xmath0-cores are the largest maximal connected sub - graphs of the network where each node has at least @xmath0 neighbors within the subgraph . we use the word ` giant ' when the largest @xmath0-core component scales linearly with the size of the network . network @xmath0-cores are important for network analysis @xcite , diffusion of information @xcite , and the spread of diseases @xcite . a well - known result is that the giant @xmath0-core of the erds - rnyi graphs @xcite emerges discontinuously @xcite if @xmath2 , whereas giant @xmath3-core ( the largest connected component ) and @xmath4-core emerge continuously . however , many empirical studies @xcite have shown that most real - world networks are highly clustered , i.e. they have a large number of triangles and higher order motifs , which can not be accounted for by simple random network models like erds - rnyi ( er ) networks @xcite , block models @xcite , or the configuration model @xcite.we here consider the simplest way in which we can extend the er random network model to incorporate clustering by using transitive linking or triadic closure @xcite . we find that despite its simplicity , this ensemble of networks shows remarkably rich features in its @xmath0-core structure . in particular , we focus on the emergence of giant 3-cores ( henceforth we omit the word ` giant ' ) and calculate their sizes in more realistic networks with clustering . considering that real - world networks are often highly clustered @xcite , our results challenge the typical assumption of simple and discontinuous emergence of @xmath0-core structure . our conclusions therefore have important implications on resilience @xcite , outbreaks of diseases @xcite or social contagions @xcite , jamming @xcite and failures of multiple dependent networks @xcite . all of these processes can be interpreted as generalizations to higher order constraints of classic bond percolation . they thus all depend , either directly or indirectly , on the @xmath0-core structure and clustering of connections in a network . our model , which we call the @xmath5-process , is a one - step extension to the classic er random network . it has two parameters , @xmath6 , the average degree of the er substrate , and @xmath7 the probability of transitive links . we generate an ensemble of random networks with the parameters @xmath6 and @xmath7 as follows . first , draw an er random network with average degree @xmath6 . on this network , identify all pairs of nodes joined by a mutual neighbor ( equivalently all motifs with three nodes and two links ) . connect each of these pairs independently with probability @xmath7 . note that @xmath7 controls the level of clustering in the network . henceforth we call links from the er substrate as ` er - links ' and links from the transitive linking process as ` clustering ' links . a node with er degree @xmath0 , has @xmath8 pairs of neighbors . each pair is linked with probability @xmath7 . hence , the total number of clustering links is given by @xmath9 yielding a total average degree of @xmath10 . a more careful , but similar calculation gives us the full degree - distribution . each node with er degree @xmath11 , has second - neighbors distributed as the sum of @xmath11 poisson random variables with average excess degree @xmath6 ( average number of links from a node reached by a random link , not counting the link you came from ) @xcite . the sum of @xmath11 poisson random variables with average @xmath6 is again a poisson random variable with average @xmath12 . each of these second - neighbors is linked to the node with probability @xmath7 , yielding a poisson distributed number of clustering links with average @xmath13 . thus the total degree distribution is given by , @xmath14 where the individual terms within the summation are the joint probability distribution of a node having @xmath11 er links and @xmath15 clustering links . in this paper , we are only interested in the @xmath0-core structure of this model . note that the transitive linking process does not change the size of the 1-core as it only links nodes that are already in the same component . however , the @xmath5-process can upgrade nodes in the 1-core component to a 2-core nodes by connecting the nodes of degree-1 to nodes of higher degree , but since in the er substrate both the giant 1-core and 2-core emerge at exactly @xmath16 , we expect no difference after the transitive linking process . the emergence of the 3-core as a function of @xmath6 exhibits four distinct regimes depending on @xmath7 , 1 . a regime of sudden discontinuous appearance of a 3-core when @xmath7 is lesser than some bound @xmath17 ( we notate @xmath18 ) , 2 . a double phase transition regime consisting of a continuous appearance of a 3-core followed by a discontinuous jump in size when @xmath19 . 3 . another double phase transition regime consisting of two continuous transitions when @xmath20 , 4 . a single sudden but continuous appearance of a 3-core for @xmath21 . to understand the effect of @xmath7 on the size of the 3-core , we develop the following analytical approximations : we calculate the 3-core sizes in random networks with the same degree distribution as our @xmath5-process and in random networks with the same joint degree - triangle distribution as our @xmath5-process . these calculations lead us to conclude that the degree - distribution and clustering by themselves are not a sufficient to explain the regimes of double transition . we also highlight the role of higher order motifs in the @xmath5-process . + [ [ configuration - model . ] ] * configuration model . * + + + + + + + + + + + + + + + + + + + + + + the configuration model defines the ensemble of all random networks constrained by a given degree distribution . we can find the size of the 3-core for this ensemble through a self - consistent equations @xcite . the probability @xmath22 , that a randomly chosen link does _ not _ lead to the 3-core is given by , @xmath23 where @xmath24 is the generating function for the degree distribution @xcite and @xmath25 is its @xmath26-th derivative . from eq . , we can find that @xmath24 is given by , @xmath27 \ ; .\ ] ] equation means that for the link to not lead to the 3-core ( left - hand side ) , the node it leads to must have at most one excess neighbor in the 3-core ( right - hand side ) . we then calculate the fraction @xmath28 of nodes in the 3-core by summing the probability that a randomly chosen node has at least three links leading to the @xmath29-core ( or one minus the probability of having less than three ) : @xmath30 this approach always leads to discontinuous transitions ( fig . [ clustered - and - config - model ] ) , which shows that the degree distribution alone is not sufficient to explain the existence of the three other regimes . hence , we develop the following approach to incorporate clustering . [ [ clustered - configuration - model . ] ] * clustered configuration model . * + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + the clustered configuration model defines the ensemble of all random networks constrained by a given joint degree - triangle distribution @xmath31 . we wish to calculate the size of the @xmath29-core given this probability @xmath31 of a randomly chosen node having @xmath0 links and @xmath32 triangles . this ensemble of networks was introduced to study the impact of clustering on bond percolation @xcite . in networks without clustering , all links were considered independent ( as they are indeed independent in a fully random network ) . however , this assumption fails when a network contains triangles and we must therefore explicitly account for triangles through additional self - consistent equations . let @xmath33 and @xmath34 be the excess degree - triangle distribution when following a link or a triangle , respectively @xcite . we can write self - consistent equations for the probability @xmath22 of following a link to a node _ not _ in the 3-core , the probability @xmath35 that following a triangle does _ not _ lead to two nodes in the 3-core and the probability @xmath36 that following a triangle leads to exactly _ zero _ node in the 3-core . figure [ cartoon - kore ] gives a visual representation of these probabilities and motifs . mathematically , we write 0.98@xx@ @xmath37 \label{eqn : u}\ ] ] & @xmath38 ^ 2 \label{eqn : v}\ ] ] @xmath39 \label{eqn : w}\end{aligned}\ ] ] where @xmath40 . equation ( [ eqn : u ] ) is equivalent to eq . ( [ uconfig ] ) , a regular link will not lead to the 3-core if the node it reaches has at _ most _ one link leading to the 3-core . this potential link can either be a regular link ( subscript @xmath41 for the binomial term @xmath42 ) or be one link of a triangle ( subscript @xmath6 for the binomial term @xmath43 ) , but no triangles can contribute more than a single link ( the term @xmath44 ) . equation ( [ eqn : v ] ) corresponds to following a triangle from one node and finding that both links lead to the 3-core , hence both nodes reached must have at _ least _ one other link leading to the 3-core . we thus write @xmath45 has the square of one minus the probability of having zero link leading to the 3-core , since the other links of these two nodes are independent . equation ( [ eqn : w ] ) is more involved but similar . when following a triangle , if neither of the two links lead to the 3-core , both nodes must have at _ most _ one link leading to the 3-core , but they can not both have one as there is also a link between the two . the size of the 3-core , @xmath28 , is then given by , @xmath46 we evaluate this formula on our @xmath5-process as follows . as we can not calculate the @xmath31 matrix analytically , we empirically obtain @xmath31 from generated networks and use this in eq . to obtain fig . [ clustered - and - config - model ] . we compare this to the 3-core sizes obtained from rewired networks preserving @xmath31 and from our @xmath5-process simulations . we see that there is remarkable agreement except at the onset of the transition . it is important to note that the rewired networks show a systematic quantitative difference as compared to our model . this is due to the high density of adjacent triangles that share a link . under a configuration model rewiring , these adjacent triangles are very unlikely , which means that the rewired networks have a higher average degree from the same @xmath31 joint distribution . every triangle contributes two links in the clustered configuration model whereas , in the @xmath5-process , we see many pairs of adjacent triangles that account for only three links as they overlap . more importantly , the clustered configuration model captures how high clustering drastically diminishes the size of the discontinuous jump . however , the transition is still discontinuous and we do not see regimes of double phase transitions . we can thus conclude that regimes ii through iv are caused the high number of adjacent triangles in the @xmath5-process and correlations between neighboring nodes . [ [ discussion . ] ] * discussion . * + + + + + + + + + + + + + our model is a simple extension of the classic er network to include clustering through triadic closure . the degree distribution still has a gaussian tail , but remarkably we find four distinct regimes shaping how 3-cores can emerge : through a single discontinuous phase transition ( @xmath18 ) , through a double transition of hybrid nature ( @xmath19 ) , through two consecutive continuous transitions ( @xmath20 ) , or through a single continuous phase transition ( @xmath21 ) . the first and last regime were previously obtained in heterogeneous @xmath0-core percolation where different nodes have different thresholds @xcite . similarly , the regime of hybrid phase transitions ( @xmath19 ) was observed in the completely different context of percolation on interdependent networks @xcite . these analogies might hold the key to understanding the mechanisms at play behind these different regimes : they all use a mixture of nodes with different sensitivity to `` criticality '' . criticality here means whether a given node is in the giant component or not ( either a giant percolating component , or a giant @xmath0-core ) . interdependent networks use a mixture of nodes that are either independent or dependent on a given node in a different networks . heterogeneous @xmath0-core percolation uses a mixture of nodes with different thresholds ; such that the giant component is a mixture of , for example , the 2-core and the 3-core . however we show that realistic networks can have very similar properties even using the simplest definition of @xmath0-cores where all nodes obey the same rules . in this context , these different sensitivities are structural ( e.g. is a node part of an important motif ) rather than _ ad hoc _ conditions . finally , we showed a simple way to extend the configuration model to incorporate clustering in calculations for the size of the 3-core . this way of evaluating motifs can be extended straightforwardly to higher @xmath0-cores , and even to higher order motifs by introducing more self - consistent equations . we believe that this new analytical @xmath0-core calculation will provide useful estimates for the aforementioned applications as well as for the emergence and growth of @xmath0-core structure in real - world networks .
planetary nebulae ( pne ) represent the final phases of the stellar evolution of low- and intermediate - mass stars . the shape of a pn is thought to be determined by the interaction of the current fast , tenuous stellar wind ( e.g. , * ? ? ? * ) of its central star with the previous slow , dense wind of the asymptotic giant branch ( agb ) phase . in the so - called interacting stellar winds ( isw ) model of pn formation @xcite , the critical parameter is the density distribution of the agb wind @xcite : a spherical symmetric agb wind will result in a spherical pn , while a polar density gradient in the agb wind will result in an elliptical or bipolar pn , depending on the degree of the density gradient . this simplified view of pn formation has been challenged by the overwhelming body of observations of pne showing small - scale features ( e.g. , ngc7662 , * ? ? ? * ) , collimated bipolar outflows ( e.g. , ngc7009 , * ? ? ? * ) , and point - symmetric morphologies , including multiple point - symmetric bubbles ( e.g. , hen2 - 47 and m1 - 37 , * ? ? ? * ) and point - symmetric collimated outflows ( e.g. , ngc6884 , * ? ? ? all these features are difficult to interpret in the framework of the isw model and have inspired new theoretical ideas to explain the shaping of pne . new scenarios of pne formation and shaping include magnetic collimation @xcite , the action of collimated fast winds @xcite , or the blowing up of a warped circumstellar disk by the fast stellar wind @xcite . @xcite recognized a new morphological class of pne characterized by the point - symmetry with respect to their centers . the morphology of these pne suggests that they are shaped by a mechanism involving precession , most likely the action of precessing or rotating collimated outflows . an investigation of pne with point - symmetric morphologies is , thus , essential to evaluate the action of precessing collimated outflows in the formation of pne and to assess the origin of the collimating mechanisms . among the pne with point - symmetric morphology , ic4634 is an outstanding case . ic4634 is an archetypical point - symmetric pn as @xcite found that its remarkable double - s - shaped morphology is reflected in its kinematics : each component of the two s - shaped features expands radially with its counterpart on the other side of the star expanding in opposite direction . the collimated outflows of ic4634 are further revealed by bow shock features indicating their interaction with the surrounding medium . therefore , both the morphology and kinematics of ic4634 suggest the presence of highly symmetric fast collimated outflows , and the precession or rotation of the source that produced them . since the study of ic4634 by @xcite , several studies have further investigated the kinematical and morphological properties of ic4634 @xcite , but a detailed study is lacking . in this paper , we present a spatio - kinematical study of ic4634 using narrow - band archival _ hst _ wfpc2 and ground - based images and complementary long - slit echelle and medium - dispersion spectroscopic observations . the observations , described in @xmath22 , have allowed us to investigate the morphology , kinematics , and physical conditions of ic4634 and the possible shock excitation of the bow shock features . a thorough description of the different structural components of ic4634 derived from these observations is given in @xmath23 , and the action of precessing collimated outflows is examined and modeled in @xmath24 . the results are discussed in @xmath25 . narrow - band _ hst _ wfpc2 images of ic4634 in the h@xmath3 , h@xmath1 , [ n ii ] @xmath46584 , and [ o iii ] @xmath45007 lines were retrieved from the _ hst _ archive ( proposal i d 6856 , pi : j.t . trauger ) . the total integration times were 800 s for the h@xmath3 image and 1,000 s for the h@xmath1 , [ n ii ] , and [ o iii ] images . these images were calibrated via the pipeline procedure . cosmic rays were removed by combining different exposures obtained with the same filter , using standard iraf routines . the _ hst _ h@xmath1 , [ n ii ] , and [ o iii ] images of ic4634 are displayed in figure [ f1]-_left_. the h@xmath3 image , not shown here , has been used in conjunction with the h@xmath1 image to investigate the amount of extinction towards ic4634 and its spatial distribution . the mosaiced _ hst _ wfpc2 h@xmath1 and [ n ii ] images have been added together to produce the deep h@xmath1+[n ii ] image shown in fig . [ f1]-_top - right_. the _ hst _ h@xmath1 , [ n ii ] , and [ o iii ] images of ic4634 have been used to derive [ n ii]/h@xmath1 , [ o iii]/h@xmath1 , and [ n ii]/[o iii ] ratio maps . in fig . [ f1]-_right - center _ and _ -bottom _ , the [ n ii]/h@xmath1 and [ o iii]/h@xmath1 ratio maps are shown , respectively . the [ n ii]/[o iii ] ratio map is not presented here because it is very similar to the [ n ii]/h@xmath1 ratio map . additional narrow - band images of ic4634 were obtained at the 2.5 m nordic optical telescope ( not ) of the roque de los muchachos observatory ( la palma , spain ) on 2005 august 1 . the images were acquired through filters that included the h@xmath1 ( @xmath5=6563 , @xmath6=9 , @xmath7=900 s ) and h@xmath1+[n ii ] ( @xmath5=6562 , @xmath6=46 , @xmath7=1800 s ) emission lines using the alfosc ( andalucia faint object spectrograph and camera ) camera in imaging mode . the detector was an e2v 2k@xmath82k ccd with a pixel size of 13.5@xmath9 m , providing a plate scale of 0@xmath1019 pixel@xmath11 and a field of view of @xmath126@xmath135 . the angular resolution during the observations , as derived from the @xmath6 of stars in the field of view , was 0@xmath1080@xmath109 . high - dispersion spectroscopic observations of the h@xmath1 and [ n ii ] @xmath146548,6584 lines of ic4634 were obtained using the echelle spectrograph on the 4 m blanco telescope of the cerro tololo inter - american observatory ( ctio ) on 2002 june 22 . the spectrograph was used in the long - slit mode to obtain single - order observations of the h@xmath1 and [ n ii ] @xmath146548,6584 lines . the unvignetted slit length is @xmath123 . the 79 line mm@xmath11 echelle grating and the long - focus red camera were used , resulting in a reciprocal dispersion of 3.4 mm@xmath11 . the data were recorded with the site 2k no . 6 ccd with a pixel size of 24 @xmath9 m . this configuration provides a spatial scale of 026 pixel@xmath11 and a sampling of 3.7 pixel@xmath11 along the dispersion direction . the slit width was set to 09 , and the resultant instrumental resolution ( @xmath6 ) was 8 . the angular resolution , determined by the seeing , was @xmath1210 . the echelle observations were made with the slit placed at the central star of ic4634 and oriented along the position angles ( pa ) 60@xmath15 ( @xmath7=1050 s ) , 120@xmath15 and 180@xmath15 ( @xmath7=900 s ) , and 150@xmath15 and 166@xmath15 ( @xmath7=1200 s ) . high - dispersion spectroscopic observations of the h@xmath1 , [ n ii ] @xmath46584 , and [ o iii ] @xmath45007 lines were also obtained using the iacub spectrograph @xcite on the 2.5 m not of the roque de los muchachos observatory ( la palma , spain ) on 2004 june 29 . the spectrograph was used in the long - slit mode to obtain single - order observations of the h@xmath1 and [ n ii ] @xmath46584 lines in the 9@xmath16 order , and of the [ o iii ] @xmath45007 line in the 11@xmath16 order . iacub provides a reciprocal dispersion of 1.74 mm@xmath11 in the 9@xmath16 order and 1.43 mm@xmath11 in the 11@xmath16 order . the data were recorded with a thomson ccd with a pixel size of 19 @xmath9 m . this configuration provides a spatial scale of 0139 pixel@xmath11 , and a sampling along the dispersion direction of 1.5 pixel@xmath11 for the spectrum of the h@xmath1 and [ n ii ] @xmath46584 lines , and 1.6 pixel@xmath11 for the spectrum of the [ o iii ] @xmath45007 line . the unvignetted slit length is 40 and the slit width was set to 065 , resulting in an instrumental resolution ( @xmath6 ) of 9.5 for the h@xmath1 and [ n ii ] @xmath46584 lines , and 8.5 for the [ o iii ] @xmath45007 line . the angular resolution , determined by the seeing , was @xmath1208 . the echelle observations of the h@xmath1 and [ n ii ] @xmath46584 lines was made along a slit offset 2 west and 464 north of the central star of ic4634 and oriented along pa=0@xmath15 ( @xmath7=1200 s ) . the echelle observations of the [ o iii ] @xmath45007 line was made along a slit placed at the central star of ic4634 and oriented along pa=150@xmath15 ( @xmath7=1800 s ) . ccrrc[b ! ] 2002 june 16 & central star & 153 & 900 & 1 + & central star & 153 & 300 & 4 + & central star & 153 & 60 & 3 + & central star & 153 & 20 & 3 + 2004 may 24 & central star & 153 & 300 & 3 + & central star & 153 & 10 & 3 + 2004 august & central star & 90 & 120 & 3 + & central star & 90 & 30 & 1 + & central star & 90 & 20 & 1 + & central star & 90 & 10 & 4 + & a & 90 & 300 & 3 + & a@xmath17 & 90 & 600 & 3 + & d & 90 & 30 & 4 + & d@xmath17 & 90 & 300 & 3 lrrrrrrrrrr[t ! ] h i @xmath44340 & 40@xmath182 & 44 & 41@xmath183 & 45 & 42@xmath185 & 46 & 41@xmath183 & 45 & 37@xmath186 & 41 + [ o iii ] @xmath44363 & [email protected] & 6.3 & [email protected] & 8.3 & [email protected] & 6.6 & [email protected] & 6.6 & [email protected] & 5.5 + he i @xmath44387 & [email protected] & 1.3 & @xmath19 & @xmath19 & @xmath19 & @xmath19 & @xmath19 & @xmath19 & @xmath19 & @xmath19 + he i @xmath44471 & [email protected] & 5.6 & [email protected] & 5.8 & [email protected] & 5.6 & [email protected] & 5.6 & [email protected] & 5.0 + n iii@xmath44641 + o ii@xmath44649 & & & & & & & & & & + + c iv@xmath44658 & [email protected] & 1.1 & @xmath19 & @xmath19 & @xmath19 & @xmath19 & @xmath19 & @xmath19 & @xmath19 & @xmath19 + he ii @xmath44686 & [email protected] & 0.5 & @xmath19 & @xmath19 & @xmath19 & @xmath19 & @xmath19 & @xmath19 & @xmath19 & @xmath19 + [ ar iv ] @xmath44714 & [email protected] & 1.5 & @xmath19 & @xmath19 & @xmath19 & @xmath19 & [email protected] & 1.0 & [email protected] & 1.2 + [ ar iv ] @xmath44740 & [email protected] & 1.2 & @xmath19 & @xmath19 & @xmath19 & @xmath19 & [email protected] & 0.5 & [email protected] & 0.8 + h i @xmath44861 & 100 & 100 & 100 & 100 & 100 & 100 & 100 & 100 & 100 & 100 + he i @xmath44920 & [email protected] & 1.7 & [email protected] & 2.0 & [email protected] & 1.8 & [email protected] & 1.9 & [email protected] & 1.8 + [ o iii ] @xmath44959 & 330@xmath1810 & 322 & 291@xmath188 & 284 & 260@xmath18 20 & 254 & 290@xmath18 15 & 283 & 285@xmath1820 & 278 + [ o iii ] @xmath45007 & 995@xmath18 30 & 963 & 876@xmath1818 & 847 & 850@xmath18 25 & 822 & 870@xmath18 30 & 841 & 835@xmath1830 & 808 + [ n i ] @xmath45199 & [email protected] & 0.13 & @xmath19 & @xmath19 & @xmath19 & @xmath19 & @xmath19 & @xmath19 & [email protected] & 1.6 + [ fe iii ] @xmath45270 & [email protected] & 0.14 & @xmath19 & @xmath19 & [email protected] & 0.3 & @xmath19 & @xmath19 & @xmath19 & @xmath19 + [ cl iii ] @xmath45517 & [email protected] & 0.35 & [email protected] & 0.6 & [email protected] & 0.5 & [email protected] & 0.5 & [email protected] & 0.4 + [ cl iii ] @xmath45535 & [email protected] & 0.51 & [email protected] & 0.7 & [email protected] & 0.5 & [email protected] & 0.6 & [email protected] & 0.5 + [ n ii ] @xmath45755 & [email protected] & 0.30 & [email protected] & 1.3 & [email protected] & 1.9 & [email protected] & 1.0 & [email protected] & 0.9 + c iv @xmath45801 - 12 & [email protected] & 0.46 & @xmath19 & @xmath19 & @xmath19 & @xmath19 & @xmath19 & @xmath19 & @xmath19 & @xmath19 + he i @xmath45876 & 17@xmath181 & 14.1 & [email protected] & 14.0 & [email protected] & 13.4 & [email protected] & 14.1 & [email protected] & 16.4 + [ o i ] @xmath46300 & [email protected] & 0.6 & [email protected] & 7.4 & [email protected] & 7.5 & [email protected] & 3.4 & [email protected] & 3.8 + [ s iii ] @xmath46312 & [email protected] & 1.0 & [email protected] & 1.8 & [email protected] & 2.1 & [email protected] & 1.5 & [email protected] & 1.8 + [ o i ] @xmath46363 & [email protected] & 0.23 & [email protected] & 2.7 & [email protected] & 2.5 & [email protected] & 1.3 & [email protected] & 1.3 + [ n ii ] @xmath46548 & [email protected] & 3.2 & 37@xmath184 & 28 & 42@xmath184 & 31 & 20@xmath182 & 15.0 & 22@xmath185 & 16.5 + h i @xmath46563 & 355@xmath1820 & 264 & 379@xmath1810 & 283 & 365@xmath1810 & 272 & 390@xmath1810 & 290 & 373@xmath188 & 278 + [ n ii ] @xmath46584 & [email protected] & 8.4 & 110@xmath186 & 82 & 107@xmath186 & 80 & 56@xmath185 & 42 & 65@xmath186 & 49 + he i @xmath46678 & [email protected] & 3.8 & [email protected] & 4.3 & [email protected] & 3.8 & [email protected] & 3.7 & [email protected] & 4.3 + [ s ii ] @xmath46717 & [email protected] & 0.5 & [email protected] & 8.2 & [email protected] & 7.8 & [email protected] & 3.4 & [email protected] & 3.5 + [ s ii ] @xmath46731 & [email protected] & 1.0 & [email protected] & 12.7 & [email protected] & 10.2 & [email protected] & 5.7 & [email protected] & 6.4 + he i @xmath47065 & [email protected] & 5.6 & [email protected] & 3.7 & [email protected] & 3.6 & [email protected] & 3.9 & [email protected] & 5.1 + [ ar iii ] @xmath47135 & [email protected] & 9.4 & [email protected] & 12.5 & [email protected] & 11.5 & [email protected] & 12.7 & [email protected] & 14.0 + he i @xmath47281 & [email protected] & 0.8 & @xmath19 & @xmath19 & @xmath19 & @xmath19 & [email protected] & 0.8 & [email protected] & 0.8 + [ o ii ] @xmath47320 & [email protected] & 2.0 & [email protected] & 6.3 & [email protected] & 6.1 & @xmath19 & @xmath19 & [email protected] & 6.7 + [ o ii ] @xmath47330 & [email protected] & 1.9 & [email protected] & 5.0 & [email protected] & 5.5 & @xmath19 & @xmath19 & [email protected] & 5.6 + & & & & & & & & & & + f(h@xmath3 ) ( ergs @xmath20 s@xmath11 ) & 1.4@xmath810@xmath21 & & 4.0@xmath810@xmath22 & & 3.5@xmath810@xmath22 & & 2.6@xmath810@xmath22 & & 2.1@xmath810@xmath22 & medium - dispersion long - slit spectroscopic observations of ic4634 were obtained using the boller & chivens spectrograph on the 2.1 m telescope of the observatorio astronmico nacional de san pedro mrtir ( baja california , mexico ) on 2002 june 16 and 2004 may 20 and august 12 . multiple observations were obtained at several slit positions as listed in table 1 . in all cases , the 400 line mm@xmath11 grating was used , the slit length was 5@xmath17 , and its width was set to 150 @xmath9 m , projecting to 2 on the sky . the data were recorded on a site3 1k ccd with pixel size of 24@xmath9 m . this configuration provides a spatial scale of 105 pixel@xmath11 and a spectral scale of 3 pixel@xmath11 with a spectral coverage from 4240 to 7310 . the long - slit spectra were reduced and calibrated following standard procedures using the xvista package . for the wavelength calibration , we used a he - ar lamp . the spectral resolution was @xmath126.8 . for the flux calibration , several standard stars were observed each night . the _ hst _ narrow - band images and ratio maps of ic4634 shown in fig . [ f1 ] reveal a wealth of structural components in this nebula . in fig . [ f1 ] , we have labeled the most prominent structures in ic4634 : * a bright inner shell ( _ is _ ) , surrounded by a dim outer shell ( os ) , * an inner s - shaped feature formed by two knotty point - symmetric arcs ( _ d - d@xmath17 _ ) that present distinct properties along the minor axis of _ is _ ( _ dm - dm@xmath17 _ ) , * an outer s - shaped feature formed by two pairs of bow shock - like structures ( _ a - a@xmath17 _ and _ b - b@xmath17 _ ) along different pas and located at different distances from the central star , * a pair of bow shock - like structures , _ c - c@xmath17 _ , along pa@xmath12120 , * a skin of enhanced [ o iii]/h@xmath1 ratio ( _ s _ ) that encloses the inner shell and bow shock - like structures , * an outer envelope ( _ env _ ) surrounding all previous components and where some individual features ( _ f - f@xmath17 _ and _ g _ ) can be distinguished , * a string of knots ( _ h _ ) distributed along an arc - like feature at @xmath1240@xmath0 north of ic4634 and accompanied by diffuse emission ( _ hd _ ) . in the following , we will describe in greater detail the morphology , kinematics , excitation , and physical conditions of the different structural components present in ic4634 . plasma diagnostics & & & & & + @xmath23 [ s ii ] ( @xmath24 ) & 10,600 & 2,800 & 1,500 & 3,800 & 5,200 + @xmath25 [ n ii ] ( k ) & 13,150 & 9,600 & 11,600 & 11,550 & 10,250 + @xmath25 [ o iii ] ( k ) & 9,850 & 11,350 & 10,700 & 10,550 & 10,050 + ionic abundances & & & & & + he@xmath26/h@xmath27 & 0.096 & 0.097 & 0.098 & 0.098 & 0.113 + he@xmath28/h@xmath27 & 0.001 & 0.000 & 0.000 & 0.000 & 0.000 + o@xmath29/h@xmath27 ( @xmath810@xmath30 ) & 6.2 & 150 & 95 & 44 & 70 + o@xmath27/h@xmath27 ( @xmath810@xmath31 ) & 7.1 & 110 & 60 & @xmath19 & 77 + o@xmath28/h@xmath27 ( @xmath810@xmath32 ) & 3.7 & 2.0 & 2.4 & 2.6 & 2.8 + n@xmath27/h@xmath27 ( @xmath810@xmath31 ) & 1.3 & 16 & 12 & 6.6 & 9.9 + s@xmath27/h@xmath27 ( @xmath810@xmath33 ) & 7.1 & 70 & 41 & 27 & 43 + cl@xmath28/h@xmath27 ( @xmath810@xmath33 ) & 6.5 & 6.6 & 6.2 & 6.7 & 6.0 + ar@xmath34/h@xmath27 ( @xmath810@xmath30 ) & 2.1 & 8.7 & 9.2 & 10 & 13 + elemental abundances & & & & & + he / h & 0.097 & 0.097 & 0.098 & 0.098 & 0.113 + o / h ( @xmath810@xmath32 ) & 3.7 & 3.1 & 3.0 & 2.6 & 3.6 + n / h ( @xmath810@xmath32 ) & 0.67 & 0.47 & 0.62 & @xmath19 & 0.46 + s / h ( @xmath810@xmath32 ) & 0.025 & 0.059 & 0.043 & @xmath19 & 0.053 + cl / h ( @xmath810@xmath32 ) & 0.0007 & 0.0007 & 0.0006 & 0.0007 & 0.0006 + ar / h ( @xmath810@xmath32 ) & 0.011 & 0.009 & 0.012 & 0.021 & 0.018 the h@xmath3 fluxes , and measured and intrinsic line intensity ratios of the emission lines detected in the medium - dispersion spectra in regions _ is _ , _ a - a@xmath17 _ , and _ d - d@xmath17 _ are listed in table 2 . the h@xmath1 to h@xmath3 ratio measured in the medium dispersion spectra and in the _ hst _ images imply an extinction coefficient , @xmath35 . h@xmath1/h@xmath3 ratio map indicates that there are no significant extinction variations across ic4634 . the measured line intensity ratios have been dereddened accordingly using the wavelength - dependent extinction law from @xcite . the plasma diagnostics ( @xmath25 and @xmath23 ) , and ionic and elemental abundances listed in table 3 were derived from the intrinsic line intensity ratios listed in tab . 2 for the different regions of ic4634 using the nebular abundance package elsa @xcite . from the center outward , the first component of ic4634 is the inner shell , _ is _ , around the central star . this inner shell has a size of 25@xmath852 and a major axis along pa@xmath12150@xmath15 . the edge of the inner shell is sharp and its surface brightness is enhanced with respect to the shell interior . this morphology suggests a thin shell , although it must be noted that an intricate pattern of filaments is superimposed on the shell . the shell brightness is also notably enhanced along its minor axis , marking a bright equatorial belt . the shape of the inner shell can be roughly described as elliptical ( figure [ f2 ] ) , but , at the tips of its major axis , several protrusions of different sizes extend outwards . further distortion of the shell shape is introduced by the bending of the shell edge at its equator , making the edge of the inner shell along its major axis to show a subtle s - shape . it is worthwhile to note that the central star of ic4634 is not located at the exact center of the shell , but displaced by @xmath12020 towards the south . the echellogram of the [ o iii ] emission line along the major axis of the inner shell at pa 150 ( figure [ f3 ] ) reveals a prolate expanding shell with an expansion velocity at the position of the central star of 15.3 km s@xmath11 . the systemic radial velocity in the local standard of rest ( lsr ) system is 20.1 km s@xmath11 , in agreement with other estimates ( e.g. , * ? ? ? the line tilt is small , indicating a low inclination of the shell with respect to the plane of the sky , but the s - shaped pattern of the line suggests accelerating features at the tips of the shell major axis . these blowout structures correspond to the protrusions observed at the tips of the major axis of the shell , with the northwest protrusion receding from the observer and the southeast protrusion approaching . the echellogram of the [ n ii ] emission line along the major ( pa=150@xmath15 ) and minor ( pa=60@xmath15 ) axes ( fig . [ f3 ] ) also reveal an expanding shell , but its kinematics can not be studied in detail in this line because the bright emission from the adjacent _ d - d@xmath17 _ and _ dm - dm@xmath17 _ regions overwhelms the emission from _ is_. the expansion velocity at the position of the central star is illustrated in the [ n ii ] echellogram along the minor axis by an ellipse with spatial semi - axis @xmath1213 and expansion velocity of 18 km s@xmath11 . in the h@xmath1 echellogram along the major axis of ic4634 , also shown in fig . [ f3 ] , the thermal broadening of the line is larger than the line split and no shell can be clearly identified . the inner shell of ic4634 is the region where the [ n ii]/h@xmath1 ratio map ( fig . [ f1 ] ) shows its lowest values , @xmath120.04 . the high excitation is confirmed by the [ n ii]/h@xmath1 line ratio map presented in fig . [ f3 ] that shows this ratio to be nearly zero in the central regions of ic4634 . the [ o iii]/h@xmath1 ratio ( fig . [ f1 ] ) is very flat in this region , with a typical dereddened value of @xmath123.0 . this is also the only region where emission lines of he ii are detected in the low - dispersion spectra ( tab . 1 ) , although the intensities of these lines are low with @xmath36(he ii 4686 )/@xmath36(h@xmath3 ) @xmath37 . besides the [ ar iv ] @xmath144715,4740 lines , no other high - excitation lines are present in the spectrum of the inner shell of ic4634 . the [ fe iii ] @xmath45270 emission line and the wolf - rayet complex of c iv @xmath45801 - 12 are weakly detected in this spectrum . the electron density and temperature of the inner shell of ic4634 can be derived using the density sensitive [ s ii ] @xmath146716,6731 doublet line ratio and the temperature sensitive [ n ii ] and [ o iii ] auroral to nebular line ratios . using the line intensities listed in tab . 2 , we have computed an electron density of @xmath1210,600 @xmath24 from the [ s ii ] line ratios , and electron temperatures of 13,000 k from the [ n ii ] lines and of 10,000 k from the [ o iii ] lines . these values are consistent with previous estimates of the physical conditions in the central region of ic4634 @xcite that also find higher electron temperature from the [ n ii ] lines than from the [ o iii ] lines . the nebula does not show any evidence of n or he enrichment , thus it can be classified as a type ii pn based on its chemical abundances @xcite . the inner shell of ic4634 is surrounded by a faint shell of radius 3@xmath0 best seen at pas @xmath38150@xmath15 and @xmath39330@xmath15 ( fig . [ f2 ] ) , the outer shell _ os _ , as marked in fig . [ f1 ] . at other pas , the low surface brightness of this shell is overwhelmed by the bright _ d - d@xmath17 _ and _ dm - dm@xmath17 _ features . the visible portion of _ os _ suggests that it is round or , at least , less aspherical than the innermost _ is_. one of the most prominent features in ic4634 is the pair of low - excitation arcs that we named _ d - d@xmath17 _ in fig . [ f1 ] . a close - up of these features is presented in figure [ f4 ] . _ d - d@xmath17 _ have a point - symmetric morphology , bending around _ is _ in arc - like features that extend from the _ is _ equator , at pa @xmath4060@xmath41 ( 240@xmath41 ) , to the _ is _ tips , at pa @xmath40330@xmath41 ( 150@xmath41 ) . these features do not look like filaments , but rather like broad ( @xmath40 1@xmath42 ) , clumpy strips , particularly in the [ n ii ] image . the morphology is reminiscent of the low - excitation _ polar caps _ of ngc6543 @xcite and of the _ south fuzz _ of ngc2392 @xcite . the surface brightness of these structures does not change significantly in the [ n ii ] line , but there is a clear enhancement of the h@xmath1 and [ o iii ] emission in the equatorial regions that we denote as _ dm - dm@xmath17_. to investigate these changes in more detail , we have used the well established [ o iii]/h@xmath1 vs. [ n ii]/h@xmath1 line diagnostic diagram . to construct this diagram , we have selected the rectangular regions encompassing _ d - d@xmath17 _ and _ dm - dm@xmath17 _ shown in figure [ f5]-_right _ and computed the dereddened [ o iii]/h@xmath1 and [ n ii]/h@xmath1 ratios at every pixel with surface brightness above a threshold value of 3@xmath43 . the resulting values are plotted in the [ o iii]/h@xmath1 vs. [ n ii]/h@xmath1 diagrams shown in fig . [ f5]-_left_. in these diagrams , _ d _ and _ d@xmath17 _ present very similar distributions in their excitation conditions , and the same applies between _ dm _ and _ dm@xmath17_. these show slightly higher [ o iii]/h@xmath1 values than _ d - d@xmath17 _ , but very different [ n ii]/h@xmath1 distributions , with _ dm - dm@xmath17 _ values lower than @xmath120.15 , while _ d - d@xmath17 _ present values of this ratio up to 0.6 . therefore , there is a notable segregation in the [ o iii]/h@xmath1 vs. [ n ii]/h@xmath1 plane between _ d - d@xmath17 _ and _ dm - dm@xmath17 _ , indicating different excitation conditions . the different nature of _ d - d@xmath17 _ and _ dm - dm@xmath17 _ is further substantiated by their kinematics in the [ n ii ] echellogram at pa 60@xmath41 ( fig.[f3 ] ) . in this echellogram , four distinct compact knots are detected at the location of _ d - d@xmath17 _ and _ dm - dm@xmath17_. of these four knots , the pair of knots located at the same side of the nebula have opposite velocities , i.e. , _ d _ and _ dm _ have very similar locations , but very different radial velocities , and the same applies to _ d@xmath17 _ and _ dm@xmath17_. the position - velocity ( pv ) map of the [ n ii]/h@xmath1 ratio ( fig . [ f3 ] ) shows that the brighter two [ n ii ] knots , corresponding to _ d - d@xmath17 _ , have lower excitation , while the two fainter knots show higher excitation and can be identified with _ dm - dm@xmath17_. therefore , _ d _ and _ dm@xmath17 _ are blueshifted , with systemic radial velocities of 21 km s@xmath11 and 27 km s@xmath11 , respectively , while _ d@xmath17 _ and _ dm _ are redshifted , with systemic radial velocities of + 21 km s@xmath11 and + 27 km s@xmath11 , respectively . the kinematics along _ d - d@xmath17 _ has been further mapped at pas 150@xmath41 , 166@xmath41 and 180@xmath41 . in the corresponding echellograms , _ d - d@xmath17 _ appear as bright , compact knots with a velocity @xmath6 of @xmath1218 km s@xmath11 in the [ n ii ] @xmath46584 line . _ d _ is blueshifted and its radial velocity shows little variations along this structure as shown in fig . _ d@xmath17 _ is redshifted and shows velocity variations symmetric to those presented by _ d_. it must be noted that these velocities are opposed to the line tilt shown by the inner shell , whose northwest protrusion is redshifted , while its southeast tip is blueshifted . the physical conditions at the tips of _ d - d@xmath17 _ have been derived using temperature and density sensitive line ratios obtained from the medium dispersion spectroscopy . using the line intensities listed in tab.2 , we have computed an electron density of 3,800 @xmath24 for _ d _ and 5,200 @xmath24 for _ d@xmath17 _ , i.e. , the density of _ d - d@xmath44 _ is 23 times lower than the density of the inner shell . the [ n ii ] electron temperature is found to be 11,500 k and 10,200 k for _ d _ and _ d@xmath17 _ , respectively . the [ o iii ] electron temperature is 10,500 k for _ d _ and 10,000 k for _ d@xmath17_. the system of bow shock - like structures observed at pas @xmath12150@xmath15 ( _ a - a@xmath17 _ ) , @xmath12166@xmath15 ( _ b - b@xmath17 _ ) , and @xmath12120@xmath15 ( _ c - c@xmath17 _ ) is one of the most remarkable features of ic4634 . there are very few examples of bow shock structures among pne ( e.g. , ic4593 , * ? ? ? * ) . in the following subsections , we describe in further detail the structure , excitation , physical conditions , and kinematics of these features . the h@xmath1 , [ n ii ] and [ o iii ] composite pictures of _ a _ and _ a@xmath17 _ shown in figure [ f6 ] reveal a wealth of small - scale structures in these features . the bow shock - like feature _ a _ lies 10@xmath0 nw of the central star . the counter bow shock ( feature _ a@xmath17 _ ) is precisely symmetric on the opposite side of the central star ( see fig . [ f1 ] ) . a chain of knots extends towards the se from structure _ a _ and towards the nw from structure _ a@xmath17 _ , ending in a smaller sized bow shock labeled _ b _ and _ b@xmath17 _ in the nw and se , respectively , that lie at @xmath127@xmath0 from ic4634 central star . both in _ a _ and _ a@xmath17 _ there are well - defined bow - shaped structures with sharp edges that emit predominantly in [ o iii ] ( shown in blue in fig . [ f6]-_top _ ) . fainter [ o iii ] , [ n ii ] and h@xmath1 diffuse emission is observed in the region between the bow shock and the inner nebula . in the brightest regions of _ a _ and _ a@xmath17 _ , several knots can be resolved appearing distinct in [ n ii ] ( fig . [ f6 ] ) , but surrounded by diffuse emission in [ o iii ] . the twisted distribution of these [ n ii ] knots is indicative of the development of instabilities at this location . the bow shock structures _ b _ and _ b@xmath17 _ show faint wings that are also more prominent and extended in the [ o iii ] image . only one knot , located at the cap of the bow shock structure , is resolved within _ b_. the [ o iii ] image shows that the spatial distribution of this knot is considerably more extended than the corresponding h@xmath1 and [ n ii ] intensity peaks . in the emission line images of the counter bow shock _ b@xmath17 _ , three knots are resolved . _ c _ and _ c@xmath17 _ ( fig . [ f1 ] ) are a pair of low - emissivity features located along an axis at pa @xmath12120@xmath15 . in the [ n ii ] image , they show pointed , triangular tips , that are surrounded by round , cap - like structures in the [ o iii ] and h@xmath1 images . at the outermost skin of these cap - like structures , the [ o iii]/h@xmath1 line ratio is enhanced . the [ o iii]/h@xmath1 and [ n ii]/h@xmath1 ratio maps show abrupt changes in these line ratios throughout _ a - a@xmath17 _ , _ b - b@xmath17 _ , and _ c - c@xmath17 _ ( fig . [ f6 ] ) . at the leading edge of the bow shock structures , the [ o iii ] emission is enhanced , with [ o iii]/h@xmath1 values of 23 . at the brightest regions of _ a - a@xmath17 _ and _ b - b@xmath17 _ ( i.e. , at the emitting knots of the bow shock structure ) , the [ o iii]/h@xmath1 ratio declines to 11.5 , while the [ n ii ] emission is enhanced , with [ n ii]/h@xmath1 values of 0.80.9 , significantly raised from the value of 0.050.15 behind the bow shock . similarly , there are abrupt changes in these ratios across the _ b - b@xmath17 _ features ( fig . [ f6 ] ) , with the lowest [ o iii]/h@xmath1 ratios and the highest [ n ii]/h@xmath1 ratios at the location of the knots ( i.e. , at the head of the bow shock ) . the spatial offset between the [ o iii ] enhanced cap - shaped bow shocks and the low - ionization [ n ii ] bright knots is @xmath1203 . we note that the wings at the _ a - a@xmath17 _ bow shocks show the highest excitation among these structures , with [ o iii]/h@xmath1 values of @xmath124 , and [ n ii]/h@xmath1 values lower than @xmath120.2 . as in 3.2 , we have computed the [ o iii]/h@xmath1 vs. [ n ii]/h@xmath1 diagrams for the different rectangular regions encompassing _ a - a@xmath17 _ , _ b - b@xmath17 _ , and _ c - c@xmath17 _ shown in fig . [ f5]-_right_. in these diagrams , shown in fig . [ f5]-_left _ , it is noticeable the strong similarity between the distribution of the data points of each structure in the nw and se regions . data points from structures _ a - a@xmath17 _ and _ b - b@xmath17 _ tend to occupy an extended region in fig . [ f5]-_left _ , implying large variations in the excitation conditions on small scales . most points fall in the region of the diagram with [ o iii]/h@xmath1 values @xmath1223 and [ n ii]/h@xmath1 @xmath120.050.20 , corresponding to the gas behind the bow shock . there is a locus in this diagram characterized by low values of [ o iii]/h@xmath1 and large values of [ n ii]/h@xmath1 , which correspond to the bright emitting knots . on the other hand , there are also points characterized by large values of the [ o iii]/h@xmath1 ratio , @xmath453 , and very low values of the [ n ii]/h@xmath1 ratio ; these points correspond to the skin of [ o iii ] observed at the leading edge of the knots . the excitation of the bow shock - like structures _ b - b@xmath17 _ is similar to that of _ a - a@xmath17_. most of the data points of _ b - b@xmath17 _ fall in the region having [ o iii]/h@xmath1 in the range 23 and [ n ii]/h@xmath1 ranging from 0.05 to 0.20 . the points with low [ o iii]/h@xmath1 values and high [ n ii]/h@xmath1 values corresponds to the [ n ii ] bright knots . note the presence of several data points with large [ o iii]/h@xmath1 values ( ranging from 3 to 4 ) , which correspond to the leading edge of the bow shocks . the data points from features _ c - c@xmath17 _ occupy a small region in the [ o iii]/h@xmath1 vs. [ n ii]/h@xmath1 diagram ( fig . [ f5]-_right _ ) . most points fall in the region with large [ o iii]/h@xmath1 ratio ( @xmath12 2.25 to 3.5 ) and low [ n ii]/h@xmath1 ratio ( @xmath46 0.15 ) . data points corresponding to the outer `` skin '' of [ o iii ] show [ o iii]/h@xmath1 larger than 3.5 . there are noticeable differences between the distribution of the data points of _ a - a@xmath17 _ and _ b - b@xmath17 _ in the [ o iii]/h@xmath1 vs. [ n ii]/h@xmath1 diagram , and this of _ c - c@xmath17_. in particular , the data points of _ c - c@xmath17 _ have a smaller scatter in the [ o iii]/h@xmath1 vs. [ n ii]/h@xmath1 diagram than those of _ a - a@xmath17 _ and _ b - b@xmath17 _ , implying small variations in the excitation conditions across _ c - c@xmath17_. moreover , _ c - c@xmath17 _ have higher excitation than _ a - a@xmath17 _ and _ b - b@xmath17_. _ _ ( open squares ) _ , and the bow shock - like features _ a - a@xmath17 _ _ ( stars ) _ as derived from the undereddened line ratios measured in the low - dispersion spectra . , width=340 ] our ground - based medium - dispersion spectroscopy ( tab . 2 ) can not resolve the rapidly changing ionization structure seen in the _ hst _ images , but the use of additional line diagnostic diagrams from a variety of emission line ratios can provide further information on the excitation and physical conditions of _ a - a@xmath17_. to better compare the overall excitation of _ a - a@xmath17 _ with that of the inner shell , _ is _ , and _ d - d@xmath17 _ , the line diagnostic diagrams shown in figure [ f7 ] also include emission line ratios from these regions . the spectra of _ a - a@xmath17 _ show enhanced emission of low - excitation lines : the relative intensities of [ o i ] , [ o ii ] , [ n ii ] , and [ s ii ] are up to 10 times higher than those in the spectrum of _ is _ , while the relative intensities of [ o i ] , [ n ii ] , and [ s ii ] are larger than these of _ d - d@xmath17 _ by a factor of @xmath122 . the relative intensity of the [ o iii ] lines of both _ a - a@xmath17 _ and _ d - d@xmath17 _ is only 10@xmath4715@xmath47 lower than these in _ is _ , while the he i and [ ar iii ] lines have similar intensities among these different regions , indicating that the radiation field reaching both _ a - a@xmath17 _ and _ d - d@xmath17 _ , modified by the absorption in the is , is similar in both cases and capable of ionizing species with ionization potential @xmath46 30 ev . we have obtained the [ o iii]/[o ii ] intensity ratio , which is indicative of the ionizing parameter for a photoionized gas , for the three different regions . the largest value ( i.e. , the largest ionizing parameter ) corresponds to the inner shell , as expected , with a value of @xmath12250 . _ a - a@xmath17 _ and _ d - d@xmath17 _ show lower values in the narrow range from @xmath1270 to @xmath1280 , thus indicating that these structures have similar local ionizing conditions ( somewhat lower in _ a - a@xmath17 _ ) . even though the ionizing parameters are similar in _ d - d@xmath17 _ and _ a - a@xmath17 _ , the latter shows stronger emission of low excitation lines ( see above ) , indicating shock excitation . fig . [ f7 ] also reveals the progressive change in the [ n ii ] and [ o iii ] temperatures with distance from the central star of ic4634 : the [ n ii ] 6584/5755 emission line ratio increases ( and so the temperature inferred from this line ratio decreases ) from _ is _ to _ a - a@xmath44 _ , while the [ o iii ] 5007/4363 emission line ratio decreases ( and so the temperature derived from this line ratio increases ) as we move outwards from regions of low [ n ii]/h@xmath1 values at _ is _ to regions with larger values of this ratio . the electron temperatures of _ a _ and _ a@xmath44 _ are , respectively , 10,200 k and 12,200 k from the [ n ii ] lines , and 10,800 k and 11,400 k from the [ o iii ] lines . similarly , the electron density from the [ s ii ] doublet line ratio is found to be 2,800 @xmath24 in _ a _ and 1,500 @xmath24 in _ a@xmath17 _ , i.e. , 37 times lower than in _ is_. the large opening angle of these bow shock - like structures ( _ a _ and _ a@xmath17 _ have almost flat morphologies ) suggests a low inclination angle with the plane of the sky . this is confirmed by the pv diagrams of the bow shock - like structures _ a - a@xmath17 _ and _ b - b@xmath17 _ displayed in figure [ f8 ] that show low radial velocities with respect to the systemic velocity : @xmath1820 km s@xmath11 for _ a - a@xmath17 _ , and @xmath183 km s@xmath11 for _ b - b@xmath17_. the h@xmath1 and [ n ii ] @xmath46584 echellogram along pa=120@xmath41 ( not shown in fig . [ f8 ] ) shows faint emission from _ c _ and _ c@xmath17 _ moving at radial velocities with respect to the systemic velocity of @xmath1830 km s@xmath11 . we note that _ a _ , _ b _ , and _ c _ ( _ a@xmath17 _ , _ b@xmath17 _ , and _ c@xmath17 _ ) are receding from ( approaching ) us , opposite to the motion of _ d _ ( _ d@xmath17 _ ) . it is also interesting to note the detailed line shape in the pv diagrams of fig . [ f8 ] at _ a - a@xmath17 _ and _ b - b@xmath17_. at the location of _ a - a@xmath17 _ , the h@xmath1 and [ o iii ] emission lines are wedge - shaped with the widest side at the line tip ( the leading edge of the bow shock ) , while the [ n ii ] line shape is nearly round . the fwhm of the h@xmath1 emission line of both _ a - a@xmath17 _ increases from @xmath1230 km s@xmath11 at the narrow end of this wedge - shaped feature to @xmath1270 km s@xmath11 at its wide end . a close examination of the h@xmath1 and [ o iii ] 5007 pv diagrams reveals a velocity gradient at the leading edge of _ a - a@xmath17 _ , whereas the velocities in the region behind the bow shock seem to be rather constant . the lowest velocities are found at the leading edge of the bow shock , facing away from the central star of ic4634 , where the radial velocity of the h@xmath1 line decreases @xmath1210 km s@xmath11 within 1@xmath0 . across _ b - b@xmath17 _ , the radial velocity is roughly constant . the h@xmath1 emission line profiles show a mean fwhm of @xmath1240 km s@xmath11 . the emission ends in a sharp drop at the outer edge of _ b - b@xmath17_. finally , the string of [ n ii ] knots that arises from the edge of the _ a _ and _ a@xmath17 _ bow shocks may be suggestive of a trail of material left in their motion forward . indeed , the radial velocity in the outer low - ionization arcs shows an abrupt change from the bow shock structure to the linear string of knots , suggesting a deceleration . we will show in @xmath48 that the interaction between a precessing collimated outflow with time - dependent velocity and the surrounding medium offers an alternative explanation to the string of [ n ii ] knots . it has been mentioned that the [ o iii]/h@xmath1 ratio is clearly enhanced in cap - like structures just outside the bow shocks _ a - a@xmath17 _ , _ b - b@xmath17 _ , and _ c - c@xmath17_. the [ o iii]/h@xmath1 ratio map ( fig . [ f1]-_right - center _ ) reveals that these [ o iii ] enhanced cap - like structures extend on a strip that inscribe the inner regions of ic4634 , like a thin shell or skin of enhanced [ o iii ] emission . only the outer envelope ( _ env _ ) and helical structures ( _ h _ and _ hd _ ) are located outside this skin of enhanced [ o iii]/h@xmath1 ratio . the presence of a skin of enhanced [ o iii]/h@xmath1 ratio is rare among pne ; only in ngc6543 a structure of this kind has been reported @xcite . in his detailed analysis of narrow - band _ hst _ wfpc2 images of ngc6543 , balick noticed that there is no observational artifact ( contamination of [ n ii ] emission in the h@xmath1 image or incorrect correction for wavelength - dependent geometric distortions in the camera optics ) , nor ionization effects or local variations of the o / h abundances or the electronic temperature in ngc6543 able to produce the observed increase of the [ o iii]/h@xmath1 ratio at the nebular edge . observational artifacts can also be dismissed as the origin of the observed morphology in the [ o iii]/h@xmath1 ratio map of ic4634 , as the emission in [ n ii ] is much weaker than in h@xmath1 , as well as a local variation of the o / h abundances which are rather constant throughout the nebula . emission from the main nebula scattered in the pn outer layers can also be excluded as the [ o iii]/h@xmath1 ratio in this skin is significantly different from the ratio value measured in the bright , innermost regions of the nebula . the only plausible cause of the observed enhancement in the [ o iii]/h@xmath1 ratio is a local increase of o@xmath28/h@xmath27 or @xmath25 . the origin of the skin of enhanced [ o iii]/h@xmath1 ratio in ngc6543 was concluded to be uncertain @xcite . in ic4634 , this structure is related to the bow shock structures _ a - a@xmath17 _ , _ b - b@xmath17 _ , and _ c - c@xmath17 _ and , therefore , it can be speculated that it originates in the interaction of fast collimated outflows with the nebular material . the shocks produced by fast collimated outflows would excavate a cavity in the low density nebular envelope and propagate outwards , inducing a marginal increase of the electronic temperature in a forward shock that enhances the [ o iii ] as also observed in wind - blown bubbles around wr stars @xcite . a similar origin can be attributed to such structure in ngc6543 , as suggested by the caps of enhanced [ o iii]/h@xmath1 associated with the jet - like features of ngc6543 ( see figure 2 of balick 2004 ) . the _ hst _ images of ic4634 reveal an envelope of faint emission surrounding its central regions that we have labeled as _ env _ in fig . this envelope has a patchy appearance , with a distinct arc towards the east , and individual filamentary ( _ f - f@xmath17 _ ) and arc - like ( _ g _ ) features towards the west . some of these features are best seen in the [ o iii]/h@xmath1 ratio map shown in fig . [ f1]-_right - bottom_. overall , the morphology of the envelope can be classified as elliptical , with its major axis along pa@xmath12120@xmath15 . since the envelope is more clearly detected in the h@xmath1 and [ o iii ] images than in the [ n ii ] image , it suggests a high excitation . its surface brightness is low , up to @xmath121,000 lower than that of the inner shell in the h@xmath1 line . if we assume the same electron temperature as in the inner shell , its density can be scaled down from that of the inner shell to @xmath49100 @xmath24 . finally , the envelope is detected in the high - dispersion spectra , especially in the echellograms of the h@xmath1 line . its emission shows a broad , unresolved line with a radial velocity similar to the systemic velocity . this structure seems to be an inert irregularly shaped , low - density outer envelope . the existence of a distant string of faint knots associated with ic4634 was first reported by @xcite . this feature , labeled _ h _ and _ hd _ in fig . [ f1]-_right - top _ , is shown in greater detail in the deep not h@xmath1+[n ii ] image presented in figure [ f9 ] . in this figure , _ h _ and _ hd _ are composed of several knots detected from pa @xmath40@xmath505@xmath41 to 20@xmath41 at angular distances @xmath51@xmath52 from the central star of ic4634 . in the _ hst _ images , the knots in the structure _ h _ are resolved into compact cores surrounded by clumpy emission or show the appearance of a string of knots . morphologically , component _ h _ forms an arc - like string of knots whose orientation is different from that of the major nebular axis , but similar to this of components _ b - b@xmath17 _ and _ d - d@xmath17_. component _ hd _ consists of several faint diffuse knots , including a large ( @xmath12 25@xmath42 ) corkscrew - shaped structure oriented along the south - north direction . we have obtained deep narrow - band images to search for a possible southern counterpart of _ h _ and _ hd _ , but our search has yielded negative results . + [ n ii ] image of the string of faint knots forming component _ h _ ( see fig . [ f1 ] ) . the arrows and closed curves overlaid on the image mark the nebulosities associated with this component , while the white stars indicate the position of field stars overimposed onto some of the line emission knots , as identified by comparing this narrow - band image with an _ hst _ wfpc2 broad - band @xmath53 image . the origin ( 0,0 ) is at the position of the central star of ic4634 . , width=321 ] the h@xmath1 and [ n ii ] emission from _ h _ and _ hd _ is detected in the high - dispersion spectroscopic observations . the emission lines can be fit with a single gaussian profile with @xmath6 of @xmath1225 km s@xmath11 in the h@xmath1 line and @xmath1220 km s@xmath11 in the [ n ii ] @xmath46584 line . the systemic radial velocity derived from these lines is very similar to the nebula systemic radial velocity . the morphology of the bow - shaped features _ a - a@xmath17 _ , _ b - b@xmath17 _ , and _ c - c@xmath17 _ resembles that of herbig - haro ( hh ) jets . these jets are morphologically characterized by chains of aligned knots with bow shock - like appearance , and a leading bow shock , the `` head '' of the jet , where the supersonic flow slams into the surrounding material . observations of the knots of hh jets have been interpreted by several authors as the result of time - dependent variations in the velocity of the flow ( see , e.g. , * ? ? ? such a jet model with variable ejection velocity has been shown to reproduce the overall morphology and kinematical properties of the high - velocity jets of the proto - planetary nebula hen3 - 1475 @xcite . the bow shock structure of _ a - a@xmath17 _ , _ b - b@xmath17 _ , and _ c - c@xmath17 _ is consistent with those features being the result of a jet , ejected with variable velocity , interacting with the agb remnant . the suggestion drawn from the morphology of ic4634 is reinforced by the kinematics of _ a - a@xmath17_. its h@xmath1 and [ o iii ] pv diagrams clearly show a wedge - shaped feature analogous to those observed in several hh objects ( e.g. , * ? ? ? a wedge - shaped feature in the pv diagram is expected for a bow shock moving nearly on the plane of the sky , as predicted by the traditional `` 3/2''-dimension bow shock model which is a reasonable approximation to the leading bow shock of an hh jet @xcite . in such case , the radial velocity dispersion increases suddenly at the stagnation point of the bow shock , producing a line broadening that looks like an horizontal edge in the pv diagram . the amount of broadening of the emission line profiles can be related to the velocity of the bow shock @xcite . in the case of _ a - a@xmath17 _ , the observed width of the line would indicate a bow shock velocity of @xmath12 100 km s@xmath11 . we conclude that the pv diagrams of _ a - a@xmath17 _ are , at least qualitatively , compatible with the predictions of a leading bow shock ( i.e. , the `` head '' of a jet ) nearly on the plane of the sky . further support for this scenario is provided by the spectral properties of _ a - a@xmath17_. the dereddened emission line ratios in these regions with intense emission in low - ionization lines are reminiscent of shock - excited nebula such as hh objects . while shocks might be collisionally exciting the low ionization emission lines of _ a - a@xmath17 _ , a detailed comparison of the observed emission line ratios with spectra predicted by photoionized shock models ( e.g. , * ? ? ? * ) is not straightforward , since the bow shocks in ic4634 are illuminated by the stellar ionizing flux from the post - shock direction . a more realistic comparison is enabled by the axisymmetric numerical simulations of a shocked , dense cloudlet moving away from a source of ionizing photons through a uniform and photoionized environment that have been recently obtained by @xcite and @xcite . these simulations not only produce synthetic spectra that are qualitatively consistent with the spectra of _ a - a@xmath17 _ , but also reproduce the ionization stratification of _ a - a@xmath17 _ , with the gas in the downstream region being more highly ionized than the outward - facing edge and the [ o iii ] emission showing a larger extension towards the photoionizing source . it is thus tempting to interprept the point - symmetry morphology of ic4634 as the result of moderate velocity jets with a time - dependent ejection velocity . furthermore , the point - symmetric morphology of ic4634 , which gives its s - shaped appearance to the nebula , and the varying sign of the radial velocity are commonly interpreted as the direct result of a time - dependent direction of ejection of the source . in the next section , we explore into further detail this supposition that can , in general , be of interest to other point - symmetric pne @xcite . the morphology and kinematics derived for ic4634 place important constrains and clues on some of the parameters describing the motion of such a precessing outflow . the bow shock morphology of _ a - a@xmath17 _ , and the spatial distribution of the [ o iii ] and h@xmath1 emission , preceding the [ n ii ] emission , reveals shock excitation . the observed radial velocities are low , suggesting that the angle between the precession axis and the plane of the sky , the inclination angle @xmath54 , is small . we will thus assume that the precession axis is on the plane of the sky . finally , the semi - angle of the aperture of the precession cone , @xmath1 , can be derived by measuring the angle subtended by _ a _ and _ b _ with respect to the central star . this angle is estimated to be @xmath1211 . numerical simulations were carried out with the 3d code yguaz - a @xcite , using a 5-level binary adaptive grid . the dimensions of the computational domain are 2.2@xmath810@xmath55 cm , in the x- and y - directions , and 4.4@xmath810@xmath55 cm in the z - direction , i.e. , 63@xmath8127 at a distance of 2.3 kpc , with a maximum resolution of 8.6@xmath810@xmath56 cm , i.e. , @xmath120025 . this code integrates the gas - dynamic equations by using the flux vector splitting scheme of @xcite . together with the gas - dynamic equations , several rate equations for the atomic / ionic species are also integrated . these species are : h i , h ii , he i , he ii , he iii , c ii , c iii , c iv , n i , n ii , n iii , o i , o ii , o iii , o iv , s ii , and s iii ( see details about the reaction and the cooling rates in * ? ? ? these rate equations enable the computation of a non - equilibrium function for the radiative losses . given a set of initial conditions for the jet and surrounding circumstellar medium ( csm ) , yguaz - a determines the temperature and density distributions at a given time . the temperature and density distributions allow us to compute the emission line coefficients of the h@xmath1 , [ n ii ] @xmath46584 , and [ o iii ] @xmath45007 emission lines . the intensity of the h@xmath1 line is computed considering the contributions from the recombination cascade and from @xmath57 collisional excitations . the intensity of the forbidden lines [ n ii ] @xmath46584 and [ o iii ] @xmath45007 are calculated by solving 5-level atom problems , using the parameters of @xcite . these intensities can be integrated along the line of sight to produce synthetic emission maps and pv diagrams for a wide slit covering completely the working surfaces . a jet of number density @xmath58 is injected in the base of the computational domain ( @xmath59 ) at position ( 1.1@xmath810@xmath55 cm , 1.1@xmath810@xmath55 cm ) in the @xmath60-plane , with a radius and length of 2.6@xmath810@xmath61 cm and 4.3@xmath810@xmath61 cm , respectively . the jet density is assumed to be 10@xmath32 @xmath24 . the jet velocity is modeled by the relationship : @xmath62 where @xmath63 is the mean velocity , @xmath64 is the amplitude of the velocity variation , @xmath65 is the time , and @xmath66 is the period of the velocity variation . the symmetry axis of this jet precesses with period @xmath67 . the surrounding csm has been assumed to be produced by an agb wind of constant mass - loss rate , @xmath68 , and expansion velocity , @xmath69 . the dependence of the density of such medium with the radial distance to the central star , @xmath70 , is given by : @xmath71 where the density , @xmath72 , at radius @xmath73 is determined by the mass loss - rate and expansion velocity of the agb wind as @xmath74 . in our simulations , the mass - loss rate and expansion velocity of the agb wind were assumed to be @xmath75 @xmath76 yr@xmath11 and 20 km s@xmath11 , respectively . the temperature of the csm has been set to 100 k. several numerical simulations were carried out varying @xmath77 , @xmath64 , @xmath67 , and @xmath66 , to produce synthetic emission maps and pv diagrams that can be directly compared to the observed images and long - slit echellograms of _ a - a@xmath17 _ , _ b - b@xmath17 _ , and _ d - d@xmath17_. there is a qualitative agreement between the synthetic and observed images and pv diagrams for @xmath63=300 km s@xmath11 , @xmath64=25 km s@xmath11 , @xmath67=460 yr , and @xmath66=110 yr ( figures [ f10 ] and [ f11 ] ) . the jet expansion velocity implies a jet mass - loss rate of @xmath122@xmath810@xmath78 yr@xmath11 . [ f10 ] displays the h@xmath1 , [ n ii ] , and [ o iii ] simulated emission maps obtained for a calculation time of 500 yr . at this time , the jet structure achieves a length of 3.3@xmath810@xmath55 cm , which corresponds to an angular size of 98 at a distance to ic4634 of 2.3 kpc . the upper panels correspond to the @xmath79-projection ( i.e. , the line of sight is along the @xmath80-axis ) , while the bottom panels represent the @xmath81-projection ( i.e. , the line of sight is along the @xmath82-axis ) , basically meaning the initial direction along which the jet is emitted . several working surfaces are observed in these maps at distances of 1.4@xmath810@xmath55 cm , 2.4@xmath810@xmath55 cm , and 3.3@xmath810@xmath55 cm , which are produced by the variability of the jet velocity , when slow gas is swept up by fast material . the observed morphology and spatial distribution of _ a - a@xmath17 _ and _ b - b@xmath17 _ are more closely reproduced by the @xmath79-projection than by the @xmath81-projection . we notice that the spatial separation seen at the head of the _ a - a@xmath17 _ features between the h@xmath1 and [ o iii ] emissions are not reproduced by the synthetic emission maps because the spatial structure of the region behind the shock , where the [ o iii ] emission arises mainly , is poorly resolved in our simulations . [ f11 ] displays the h@xmath1 , [ n ii ] , and [ o iii ] simulated pv diagrams obtained at the same calculation time as fig . [ f10 ] . the centroids of the working surfaces in these pv diagrams do not show exactly the alternate positive and negative velocities observed in _ a - a@xmath17 _ and _ b - b@xmath17 _ , although they display similar sinusoidal variations with respect to the systemic velocity . this indicates that the initial direction along which the jet is emitted is not coincident with neither the @xmath79- and @xmath81-projections , or that the inclination angle differs from the one that has been assumed . on the other hand , a close examination of the pv diagrams in fig . [ f11 ] reveals a good agreement with the detailed kinematics of the working surfaces : the velocity widths or @xmath6 of the synthetic emission lines are comparable to the observed ones , with larger values of the @xmath6 for the h@xmath1 line , and the wedge - shaped tips of the h@xmath1 and [ o iii ] emission line profiles of _ a - a@xmath17 _ are also well reproduced . the data presented in the previous sections reaffirm the known complexity of ic4634 . the nebula is composed of a series of morphological components that clearly reveal the action of different shaping agents . among the prevalent shaping mechanisms in ic4634 , we shall consider the interaction of the current fast stellar wind with the slow agb wind , the ionizing flux of photons from its central star , and a series of fast collimated outflows . we discuss below the effects that the fast stellar wind and fast collimated outflows have had in the shaping of ic4634 . _ iue _ observations of the central star of ic4634 uncovered prominent p cygni profiles in the c iii , c iv , n v , and o v lines superposed on the stellar continuum . these p cygni profiles manifest the presence of a fast stellar wind whose terminal velocity has been derived to be 3,5004,000 km s@xmath11 @xcite . this fast stellar wind has excavated a central cavity within the nebula , compressing the nebular material into a thin shell . this thin shell and the central cavity can be identified with the inner shell _ is _ , while the nebular envelope can be associated with the faint shell _ os _ surrounding this inner shell . in this respect , the double - shell morphology of the inner regions of ic4634 compares well with other pne ( e.g. , ngc6826 ) . _ ( left ) _ , [ n ii ] _ ( center ) _ , and [ o iii ] _ ( right ) _ simulated emission maps of a velocity time - dependent precessing collimated outflow at a time of 500 yr . the uppers panels show the simulated emission maps for the @xmath79projection , while the @xmath81projection is displayed in the bottom panels . the spatial axes show the distance to the central star of ic4634 in units of @xmath83 cm . the maps are displayed in logarithmic scale with the flux level in units of ergs s@xmath11 @xmath20 str@xmath11 shown in the colorbar located at the top of each upper panel . , height=415 ] the bow shock morphology and spatial distribution of _ a - a@xmath17 _ , _ b - b@xmath17 _ , and _ c - c@xmath17 _ are highly indicative of the interaction of precessing collimated outflows with surrounding material . indeed , the detailed morphology and kinematics of _ a - a@xmath17 _ and _ b - b@xmath17 _ are successfully reproduced assuming the interaction of a precessing jet with a time - dependent velocity in the range of 300 km s@xmath11 and a precession period @xmath12460 yrs . this jet is interacting with material in the outer regions of ic4634 that forms the envelope , _ env _ , a region of rough elliptical symmetry that may represent an episode of major mass loss prior to the one that formed ic4634 inner regions . in their interaction , the collimated outflows generate bow - shocks structures and drive a forward shock that produces a high excitation skin . one of the new nebular components revealed in this work is the outermost features _ h _ and _ hd_. the location of _ h _ and _ hd _ with respect to ic4634 and their similar radial velocities make it unlikely that these components are unrelated to the nebula . the morphology of the _ h _ feature , composed of a series of knots distributed along an arc - like structure , can be interpreted as a loop on the surface of an imaginary cone whose vertex is coincident with the central star of ic4634 . therefore , _ h _ and _ hd _ may correspond to an ancient precessing ejection . it is worth emphasizing that the precession axis of this ancient ejection would be almost orthogonal to the axis of the precessing ejection that has been considered for _ a - a@xmath17 _ and _ b - b@xmath17_. _ ( left ) _ , [ n ii ] _ ( center ) _ , and [ o iii ] _ ( right ) _ emission lines of a velocity time - dependent precessing collimated outflow at a time of 500 yr . the bottom panels display the @xmath81projection , while the upper panels show the @xmath79projection . in the panels , the horizontal axis displays the systemic radial velocity in units of , and the vertical axis the distance to the central star of ic4634 in units of @xmath83 cm . the maps are displayed in logarithmic scale with the flux level in units of ergs s@xmath11 @xmath20 str@xmath11 ( cm s@xmath11)@xmath11 shown in the colorbar located at the top of each upper panel . , height=415 ] there are , thus , clear signs that ic4634 innermost shell , _ is _ , and the envelope , _ env _ , formed by the action of the current fast stellar wind on material associated to episodes of mildly asymmetric mass loss produced during the late agb , while the bow shock structures _ a - a@xmath17 _ , _ b - b@xmath17 _ , and _ c - c@xmath17 _ , and the arc - like features _ h _ and _ hd _ have been produced by the action of fast , precessing collimated outflows . these collimated outflows may have also played some role in the shaping of the innermost regions of ic4634 , although the evidence is not overwhelming . the inner shell of ic4634 , _ is _ , is elongated and shows morphological and kinematical signs of blowout along its tips . while the origin of the asymmetry of _ is _ can be attributed to a density enhancement in the equator of the nebular envelope as suggested by the bright emission in the surrounding material along the minor axis in the h@xmath1 and [ o iii ] images , it is worthwhile noting that the orientation and inclination of the collimated outflows associated with _ a - a@xmath17 _ are coincident with those of the blowout along the tips of the inner shell . therefore , it is possible that the collimated outflows responsible of _ a - a@xmath17 _ may have been involved in the emergence of the blowout structures of _ is _ and in the development of the axisymmetry of this shell . this may also be the case of the inner shells of several pne with collimated outflows ( e.g. , ngc6210 and ngc6884 ) that also show the mild point - symmetry exhibited by the inner shell of ic4634 . similarly , the origin of the _ d - d@xmath17 _ features may be associated with the action of collimated outflows . the kinematics and morphology of _ d - d@xmath17 _ are very similar to those of the low - ionization _ polar caps _ described in ngc6543 @xcite , although these structures may have different inclinations . a thorough discussion of the possible origins of knots and low - ionization structures in pne is provided by @xcite . as in ngc6543 , the _ d - d@xmath17 _ arcs of ic4634 can be interpreted as intrusions of low - ionization , high - density knots that are expanding with the shell that surrounds its inner shell @xcite . these features may represent the relic of the interaction of a collimated outflow with the nebular envelope . on the other hand , _ dm - dm@xmath17 _ , that has higher excitation than _ d - d@xmath17 _ , and that are separated from these in the velocity space , can be related to the density enhancement around the equatorial region of _ pne display a large variety of low - ionization structures of which fast collimated outflows deserve special attention because of their outstanding kinematical properties ( e.g. , * ? ? ? the origin and nature of the fast collimated outflows seen in pne has been disputed , but the morphological , kinematical , and physical properties of some of them seem to be suggestive of high - density bullets moving supersonically through the nebular material @xcite . if this were the case , the interaction of the pn fast collimated outflows with the nebular material should lead to the formation of bow shocks whose morphologies would share some similarities with these of hh objects . bow shocks would then be expected to be common among pne with fast collimated outflows ; the number of bow shocks identified in pne , however , is small . in ic4634 , we find one of the rare cases of bow shocks in pne . the detailed morphology and kinematics of the bow shock structures _ a - a@xmath17 _ and _ b - b@xmath17 _ of ic4634 can be unambiguously ascribed to a precessing fast collimated outflow , and the morphology and ionization of the bow shock structures _ c - c@xmath17 _ are reminiscent of the interaction of a fast , dense bullet with the nebular material . in view of these findings , it is worthwhile to revisit other cases of bow shocks in pne . so far , the best studied case of a bow shock in a pn is that of ic4593 @xcite . in this nebula , we find several systems of knots and inward facing tails that have been interpreted as multiple collimated outflows propagating along different directions . these knots are preceded by outward facing caps with bow shock morphologies . the bow shock structures in ic4593 share many similarities with those of ic4634 . both have low radial velocities , show a decrease in the observed radial velocity and the broadening of the emission profile , and display caps of enhanced [ o iii ] emission as expected in bow shock models . all these properties suggest the motion of medium velocity jets near the plane of the sky that are interacting with the surrounding material . the _ ansae _ of ngc7009 are another promising example of fast , dense bullets ramming through the outer shells of the nebula . this interpretation was questioned as the ionization state of the gas declines at the head of the bow shock @xcite , while the opposite is expected in bow shocks @xcite . however , the observed ionization gradient can be reproduced by the numerical simulations of a shocked cloudlet moving away from the central star if the stellar ionizing photon flux , which modifies the ionization and excitation structure of the shock , is included @xcite . in their study of ic4593 bow shocks , @xcite noted that the ionizing flux from the central star can play a predominant role in the ionization stratification of bow shocks in pne . at any rate , an examination of a deep _ hst _ wfpc2 [ o iii]/h@xmath1 ratio map of ngc7009 reveals that the _ ansae _ of ngc7009 are preceded by caps of enhanced [ o iii ] as expected in bow shocks @xcite . besides ic4593 , ic4634 , and ngc7009 , there are very few other cases of bow shock associated with pne . the [ o iii]/h@xmath1 ratio image of ngc6543 @xcite shows a bow shock structure enveloping the precessing - like collimated outflows along the polar directions of this nebula @xcite , although this structure is not discussed in detail . a bow shock structure has also been described in ngc6572 , although the limited spatial resolution of ground - based images do not allow the authors to study the ionization stratification at the bow shock @xcite . a comprehensive investigation of the occurrence of bow shocks associated with the expansion of fast collimated outflows or bullets in pne is in progress @xcite . our spatio - kinematical study of ic4634 has revealed new structural components in this nebula , including what seems to be the relic of ancient precessing collimated ejections , a triple shell morphology , and a thin skin of enhanced [ o iii]/h@xmath1 enveloping the nebula and the bow shock structures at the tip of the collimated outflows . furthermore , the _ hst _ images and ground - based echelle spectroscopy provide a detailed view of the physical structure of the bow shock associated with a collimated outflow in a pn . their morphology and kinematics have been successfully modeled using hydro - dynamical simulations in which a precessing fast collimated outflow interacts with nebular material . ic4634 seems to have experienced a series of mildly asymmetric mass loss episodes that have removed the stellar envelope , yielding to the current fast stellar wind . these periods of mass loss are interspersed with two episodic ejections of collimated outflows that have interacted with the nebular material , having important consequences in the shaping of ic4634 . the most recent episode of ejection of collimated outflows is directly responsible of the formation of the bow shock structures _ a - a@xmath17 _ , _ b - b@xmath17 _ , and _ c - c@xmath17 _ within the envelope , and has played a significant role in the asymmetry and orientation of the inner shell . the probable interaction of this collimated outflow with circumstellar material might be linked to the formation of _ d - d@xmath17_. finally , the oldest episode of ejection of collimated outflows formed _ h _ and _ hd _ in the outermost regions . the two ejections of collimated outflows took place at different times during the pn formation . the large misalignment between the ejection axes of the collimated outflows giving raise to _ a - a@xmath17 _ , _ b - b@xmath17 _ , and _ c - c@xmath17 _ , on one hand , and _ h _ and _ hd _ , on the other , imply that the collimating source had experienced important changes between the ejections . the data presented here have been taken using alfosc , which is owned by the instituto de astrofsica de andaluca ( iaa ) and operated at the nordic optical telescope under agreement between iaa and the nbifafg of the astronomical observatory of copenhagen . part of this work was supported by the spanish mcyt projects number aya 2002 - 00376 and aya 2005 - 01495 cofunded by feder funds . the work of a.r . was supported by the spanish mcyt grants aya 2005 - 08523-c03 - 01 and aya 2005 - 08013-c03 - 01 cofunded by feder funds . pfv and acr acknowledge support from conacyt ( mexico ) grant 46628-f , and dgapa - unam grants in108207 and in100407 . the work of pfv and acr is also supported by the `` macroproyecto de tecnologas para la universidad de la informacin y la computacin '' ( secretara de desarrollo institucional de la unam , programa transdisciplinario en investigacin y desarrollo para facultades y escuelas , unidad de apoyo a la investigacin en facultades y escuelas ) . pfv and acr thank to enrique palacios and martn cruz ( icn ) for maintaining and supporting our linux servers and their assistance provided . lo , rv , and gb acknowledge kind support from conacyt ( mexico ) grant 45848 and dgapa - papiit - unam grant in111903 . aller , l. h. , & czyzak , s. j. 1983 , , 51 , 211 balick , b. 1987 , , 94 , 671 balick , b. 2004 , , 127 , 2262 balick , b. , alexander , j. , hajian , a. r. , terzian , y. , perinotto , m. , & patriarchi , p. 1998 , , 116 , 360 balick , b. , & hajian , a. r. 2004 , , 127 , 2269 balick , b. , perinotto , m. , maccioni , a. , terzian , y. , & hajian , a. r. 1994 , , 424 , 800 bhm , k .- h . & solf , j. 1985 , , 294 , 533 cliffe , j. a. , frank , a. , livio , m. , & jones , t. w. 1995 , , 447 , l49 corradi , r. l. m. , guerrero , m. a. , manchado , a. , & mampaso , a. 1996 , new astr , 2 , 461 de freitas pacheco , j. a. , maciel , w. j. , & costa , r. d. d. 1992 , , 261 , 579 dopita , m. a. 1997 , , 485 , l41 durand , s. , acker , a. , & zijlstra , a. 1998 , , 132 , 13 fernndez , r. , monteiro , h. , & schwarz , h. e. 2004 , , 603 , 595 garca - arredondo , f. , & frank , a. 2004 , , 600 , 992 garca - segura , g. , & lpez , j. a. 2000 , , 544 , 336 gonalves , d.r . , corradi , r.l.m . , & mampaso , a. 2001 , , 547 , 302 gruendl , r. a. , chu , y .- h . , dunne , b. c. , & points , s. d. 2000 , , 120 , 2670 guerrero , m.a . , medina , j.j . , luridiana , v. , miranda , l.f . , riera , a. , & velzquez , p.f . 2008 , in preparation guerrero , m. a. , miranda , l. f. , & chu , y .- h . 2004 , asymmetrical planetary nebulae iii : winds , structure and the thunderbird , 313 , 30 hajian , a. r. , balick , b. , terzian , y. , & perinotto , m. 1997 , , 487 , 304 hartigan , p. , raymond , j. & hartmann , l. 1987 , , 316 , 323 hyung , s. , aller , l. h. , & feibelman , w. a. 1999 , , 525 , 294 johnson , m. d. , levitt , j. s. , henry , r. b. c. , & kwitter , k. b. 2006 , planetary nebulae in our galaxy and beyond , 234 , 439 kwok , s. , purton , c. r. , & fitzgerald , p. m. 1978 , , 219 , l125 lee , c .- f . , & sahai , r. 2003 , , 586 , 319 m@xmath84keith , c. d. , garca - lpez , r. j. , rebolo , r. , barnett , e. w. , beckman , j. e. , martn , e. l. , & trapero , j. 1993 , , 273 , 331 medina , j.j . , guerrero , m.a . , luridiana , v. , miranda , l.f . , riera , a. , & velzque , p.f . 2007 , proceedings of the asymmetrical planetary nebulae iv conference , eds . r.l.m . corradi , a. manchado and n. soker , in press mendoza , c. 1983 , in planetary nebulae , iau symp . , 103 , 143 miranda , l. f. , guerrero , m. a. , & torrelles , j. m. 1999 , , 117 , 1421 miranda , l. f. , & solf , j. 1992 , , 260 , 397 miranda , l. f. , vzquez , r. , corradi , r.l.m . , guerrero , m. a. , lpez , j.a . , & torrelles , j. m. 1999 , , 520 , 714 odell , c. r. , balick , b. , hajian , a. r. , henney , w. j. , & burkert , a. 2002 , , 123 , 3329 patriarchi , p. , & perinotto , m. 1991 , , 91 , 325 peimbert , m. 1978 , iau symp . 76 : planetary nebulae , 76 , 215 perinotto , m. , patriarchi , p. , balick , b. , & corradi , r. l. m. 2004 , , 422 , 963 raga , a. c. , binette , l. , canto , j. , & calvet , n. 1990 , , 364 , 601 raga , a.c . & bhm , k .- h . 1985 , , 58 , 201 raga , a.c . & bhm , k .- h . 1986 , , 308 , 829 raga , a.c . , de gouveia dal pino , e.m . , noriega - crespo , a. , mininni , p.d . , & velzquez , p.f . 2002 , , 392 , 267 raga , a.c . , navarro - gonzlez , r. , & villagran - muniz , m. 2000 , revista mexicana de astr . y astrof . , 36 , 67 raga , a.c . , riera , a. , mellema , g. , esquivel , a. & velzquez , p.f . 2007 , ( submitted ) riera , a. & raga , a.c . 2007 , proceedings of the asymmetrical planetary nebulae iv conference , eds . corradi , a. manchado and n. soker , in press rijkhorst , e .- j . , mellema , g. , & icke , v. 2005 , , 444 , 849 sahai , r. 2000 , , 537 , l43 savage , b.d . , & mathis , j.s . 1979 , , 17 , 23 schwarz , h. e. 1993 , mass loss on the agb and beyond , second eso / ctio workshop , eds . h. e. schwarz ( eso ; garching ) , 223 soker , n. , & bisker , g. 2006 , , 369 , 1115 stanghellini , l. , corradi , r. l. m. , & schwarz , h. e. 1993 , , 279 , 521 toledano , o. , riesgo - tirado , h. , lpez , j. a. , meaburn , j. , bryce , m. , & holloway , a. j. 2003 , iau symposium , 209 , 543 van leer , b. 1982 , icase report nos . 82 - 30 velzquez , p. f. , riera , a. & raga , a. c. 2004 , , 419 , 991	with its remarkable double - s shape , ic4634 is an archetype of point - symmetric planetary nebulae ( pn ) . in this paper , we present a detailed study of this pn using archival _ hst _ wfpc2 and ground - based narrow - band images to investigate its morphology , and long - slit spectroscopic observations to determine its kinematics and to derive its physical conditions and excitation . the data reveal new structural components , including a distant string of knots distributed along an arc - like feature 40@xmath060@xmath0 from the center of the nebula , a skin of enhanced [ o iii]/h@xmath1 ratio enveloping the inner shell and the double - s feature , and a triple - shell structure . the spatio - kinematical study also finds an equatorial component of the main nebula that is kinematically independent from the bright inner s - shaped arc . we have investigated in detail the bow shock - like features in ic4634 and found that their morphological , kinematical and emission properties are consistent with the interaction of a collimated outflow with surrounding material . indeed , the morphology and kinematics of some of these features can be interpreted using a 3d numerical simulation of a collimated outflow precessing at a moderate , time - dependent velocity . apparently , ic4634 has experienced several episodes of point - symmetric ejections oriented at different directions with the outer s - shaped feature being related to an earlier point - symmetric ejection and the outermost arc - like string of knots being the relic of an even much earlier point - symmetric ejection . there is tantalizing evidence that the action of these collimated outflows has also taken part in the shaping of the innermost shell and inner s - shaped arc of ic4634 .

Subsets and Splits

No saved queries yet

Save your SQL queries to embed, download, and access them later. Queries will appear here once saved.